{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 4436,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.9850746268656716e-07,
      "loss": 2.4984,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.970149253731343e-07,
      "loss": 2.4375,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.955223880597015e-07,
      "loss": 2.3496,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1940298507462686e-06,
      "loss": 2.3019,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4925373134328358e-06,
      "loss": 2.2681,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.791044776119403e-06,
      "loss": 2.307,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.08955223880597e-06,
      "loss": 2.3224,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3880597014925373e-06,
      "loss": 2.2319,
      "step": 16
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.686567164179105e-06,
      "loss": 2.2864,
      "step": 18
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.9850746268656716e-06,
      "loss": 2.2184,
      "step": 20
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.283582089552239e-06,
      "loss": 2.2243,
      "step": 22
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.582089552238806e-06,
      "loss": 2.1806,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8805970149253735e-06,
      "loss": 2.1346,
      "step": 26
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.17910447761194e-06,
      "loss": 2.1112,
      "step": 28
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.477611940298508e-06,
      "loss": 2.1375,
      "step": 30
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.7761194029850745e-06,
      "loss": 2.0972,
      "step": 32
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.074626865671642e-06,
      "loss": 2.1996,
      "step": 34
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.37313432835821e-06,
      "loss": 2.115,
      "step": 36
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.671641791044776e-06,
      "loss": 2.1727,
      "step": 38
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.970149253731343e-06,
      "loss": 2.0205,
      "step": 40
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.2686567164179116e-06,
      "loss": 2.0463,
      "step": 42
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.567164179104478e-06,
      "loss": 2.1016,
      "step": 44
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.865671641791045e-06,
      "loss": 2.1054,
      "step": 46
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.164179104477612e-06,
      "loss": 2.0415,
      "step": 48
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.46268656716418e-06,
      "loss": 2.0726,
      "step": 50
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.761194029850747e-06,
      "loss": 2.0446,
      "step": 52
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.059701492537314e-06,
      "loss": 2.0737,
      "step": 54
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.35820895522388e-06,
      "loss": 2.0369,
      "step": 56
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.656716417910447e-06,
      "loss": 2.0708,
      "step": 58
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.955223880597016e-06,
      "loss": 2.0231,
      "step": 60
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.253731343283582e-06,
      "loss": 2.0473,
      "step": 62
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.552238805970149e-06,
      "loss": 2.0222,
      "step": 64
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.850746268656717e-06,
      "loss": 2.0204,
      "step": 66
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.0149253731343284e-05,
      "loss": 2.0266,
      "step": 68
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.0447761194029851e-05,
      "loss": 1.9896,
      "step": 70
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.074626865671642e-05,
      "loss": 1.9435,
      "step": 72
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.1044776119402986e-05,
      "loss": 1.9961,
      "step": 74
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.1343283582089553e-05,
      "loss": 2.0354,
      "step": 76
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.1641791044776121e-05,
      "loss": 2.0071,
      "step": 78
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.1940298507462686e-05,
      "loss": 1.9622,
      "step": 80
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.2238805970149255e-05,
      "loss": 1.9752,
      "step": 82
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.2537313432835823e-05,
      "loss": 1.9363,
      "step": 84
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.2835820895522388e-05,
      "loss": 2.0192,
      "step": 86
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.3134328358208957e-05,
      "loss": 1.9606,
      "step": 88
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.3432835820895525e-05,
      "loss": 1.9499,
      "step": 90
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.373134328358209e-05,
      "loss": 1.9427,
      "step": 92
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.4029850746268658e-05,
      "loss": 1.9274,
      "step": 94
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.4179104477611942e-05,
      "loss": 1.9864,
      "step": 96
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.4477611940298509e-05,
      "loss": 1.899,
      "step": 98
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.4776119402985077e-05,
      "loss": 2.0033,
      "step": 100
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.5074626865671642e-05,
      "loss": 1.953,
      "step": 102
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.537313432835821e-05,
      "loss": 1.9023,
      "step": 104
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.5671641791044777e-05,
      "loss": 2.022,
      "step": 106
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.5970149253731344e-05,
      "loss": 1.9652,
      "step": 108
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.626865671641791e-05,
      "loss": 1.9581,
      "step": 110
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.6567164179104477e-05,
      "loss": 2.0001,
      "step": 112
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.6865671641791048e-05,
      "loss": 1.8767,
      "step": 114
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.7164179104477614e-05,
      "loss": 1.8871,
      "step": 116
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.746268656716418e-05,
      "loss": 1.8394,
      "step": 118
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.7761194029850748e-05,
      "loss": 1.9129,
      "step": 120
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.8059701492537314e-05,
      "loss": 1.9511,
      "step": 122
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.8208955223880598e-05,
      "loss": 1.9331,
      "step": 124
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.8507462686567165e-05,
      "loss": 1.9341,
      "step": 126
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.8805970149253735e-05,
      "loss": 1.8714,
      "step": 128
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9104477611940298e-05,
      "loss": 1.9453,
      "step": 130
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9402985074626868e-05,
      "loss": 1.8974,
      "step": 132
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9701492537313435e-05,
      "loss": 1.9223,
      "step": 134
    },
    {
      "epoch": 0.12,
      "learning_rate": 2e-05,
      "loss": 1.8966,
      "step": 136
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999998933431258e-05,
      "loss": 1.8785,
      "step": 138
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999995733727306e-05,
      "loss": 1.9156,
      "step": 140
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9999904008949705e-05,
      "loss": 1.9364,
      "step": 142
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9999829349456267e-05,
      "loss": 1.8797,
      "step": 144
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9999733358952002e-05,
      "loss": 1.9061,
      "step": 146
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999961603764167e-05,
      "loss": 1.8328,
      "step": 148
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9999477385775543e-05,
      "loss": 1.8222,
      "step": 150
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9999317403649377e-05,
      "loss": 1.8906,
      "step": 152
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9999136091604433e-05,
      "loss": 1.9194,
      "step": 154
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9998933450027483e-05,
      "loss": 1.8939,
      "step": 156
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9998709479350785e-05,
      "loss": 1.7923,
      "step": 158
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.99984641800521e-05,
      "loss": 1.897,
      "step": 160
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9998197552654686e-05,
      "loss": 1.908,
      "step": 162
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.999790959772729e-05,
      "loss": 1.8676,
      "step": 164
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9997600315884166e-05,
      "loss": 1.8565,
      "step": 166
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9997269707785055e-05,
      "loss": 1.871,
      "step": 168
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9996917774135186e-05,
      "loss": 1.8545,
      "step": 170
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.999654451568528e-05,
      "loss": 1.9056,
      "step": 172
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9996149933231554e-05,
      "loss": 1.842,
      "step": 174
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9995734027615702e-05,
      "loss": 1.8387,
      "step": 176
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9995296799724914e-05,
      "loss": 1.9177,
      "step": 178
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9994838250491852e-05,
      "loss": 1.8014,
      "step": 180
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9994358380894665e-05,
      "loss": 1.8049,
      "step": 182
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.999385719195698e-05,
      "loss": 1.8036,
      "step": 184
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9993334684747906e-05,
      "loss": 1.852,
      "step": 186
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9992790860382023e-05,
      "loss": 1.8787,
      "step": 188
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9992225720019377e-05,
      "loss": 1.8519,
      "step": 190
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9991639264865492e-05,
      "loss": 1.8272,
      "step": 192
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9991031496171362e-05,
      "loss": 1.9247,
      "step": 194
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9990402415233436e-05,
      "loss": 1.8138,
      "step": 196
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9989752023393635e-05,
      "loss": 1.8417,
      "step": 198
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9989080322039328e-05,
      "loss": 1.7823,
      "step": 200
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.998838731260335e-05,
      "loss": 1.9182,
      "step": 202
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.998767299656399e-05,
      "loss": 1.7648,
      "step": 204
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9986937375444973e-05,
      "loss": 1.8909,
      "step": 206
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9986180450815485e-05,
      "loss": 1.8438,
      "step": 208
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9985402224290153e-05,
      "loss": 1.7763,
      "step": 210
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9984602697529034e-05,
      "loss": 1.8553,
      "step": 212
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9983781872237634e-05,
      "loss": 1.8461,
      "step": 214
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9982939750166888e-05,
      "loss": 1.8635,
      "step": 216
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9982076333113153e-05,
      "loss": 1.8799,
      "step": 218
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9981191622918217e-05,
      "loss": 1.8216,
      "step": 220
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9980285621469288e-05,
      "loss": 1.9037,
      "step": 222
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9979358330698995e-05,
      "loss": 1.8344,
      "step": 224
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.997840975258538e-05,
      "loss": 1.7403,
      "step": 226
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.997743988915188e-05,
      "loss": 1.8381,
      "step": 228
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9976448742467352e-05,
      "loss": 1.8507,
      "step": 230
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9975436314646052e-05,
      "loss": 1.8299,
      "step": 232
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9974402607847623e-05,
      "loss": 1.7769,
      "step": 234
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9973347624277105e-05,
      "loss": 1.8628,
      "step": 236
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9972271366184922e-05,
      "loss": 1.8459,
      "step": 238
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9971173835866884e-05,
      "loss": 1.8887,
      "step": 240
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.997005503566417e-05,
      "loss": 1.8039,
      "step": 242
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.996891496796334e-05,
      "loss": 1.7956,
      "step": 244
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9967753635196306e-05,
      "loss": 1.7893,
      "step": 246
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.996657103984036e-05,
      "loss": 1.8458,
      "step": 248
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9965367184418138e-05,
      "loss": 1.748,
      "step": 250
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.996414207149762e-05,
      "loss": 1.8475,
      "step": 252
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9962895703692155e-05,
      "loss": 1.8296,
      "step": 254
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9961628083660406e-05,
      "loss": 1.8945,
      "step": 256
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9960339214106388e-05,
      "loss": 1.8396,
      "step": 258
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.995902909777943e-05,
      "loss": 1.8078,
      "step": 260
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9957697737474198e-05,
      "loss": 1.8347,
      "step": 262
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.995634513603066e-05,
      "loss": 1.791,
      "step": 264
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9954971296334102e-05,
      "loss": 1.7971,
      "step": 266
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9953576221315116e-05,
      "loss": 1.8384,
      "step": 268
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9952159913949583e-05,
      "loss": 1.8272,
      "step": 270
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9950722377258692e-05,
      "loss": 1.8127,
      "step": 272
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9949263614308894e-05,
      "loss": 1.763,
      "step": 274
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.994778362821194e-05,
      "loss": 1.8761,
      "step": 276
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9946282422124842e-05,
      "loss": 1.8082,
      "step": 278
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.994475999924987e-05,
      "loss": 1.8545,
      "step": 280
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9943216362834575e-05,
      "loss": 1.8098,
      "step": 282
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.994165151617174e-05,
      "loss": 1.8048,
      "step": 284
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9940065462599394e-05,
      "loss": 1.8017,
      "step": 286
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9938458205500812e-05,
      "loss": 1.8242,
      "step": 288
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9936829748304496e-05,
      "loss": 1.8284,
      "step": 290
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9935180094484164e-05,
      "loss": 1.9135,
      "step": 292
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9933509247558755e-05,
      "loss": 1.8937,
      "step": 294
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.993181721109242e-05,
      "loss": 1.7698,
      "step": 296
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.99301039886945e-05,
      "loss": 1.8058,
      "step": 298
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9928369584019536e-05,
      "loss": 1.7913,
      "step": 300
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9926614000767254e-05,
      "loss": 1.7902,
      "step": 302
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.992483724268255e-05,
      "loss": 1.7558,
      "step": 304
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.99230393135555e-05,
      "loss": 1.7755,
      "step": 306
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9921220217221325e-05,
      "loss": 1.7874,
      "step": 308
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9919379957560413e-05,
      "loss": 1.7697,
      "step": 310
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9917518538498292e-05,
      "loss": 1.7788,
      "step": 312
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.991563596400562e-05,
      "loss": 1.8082,
      "step": 314
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.991373223809819e-05,
      "loss": 1.8246,
      "step": 316
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9911807364836915e-05,
      "loss": 1.7254,
      "step": 318
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9909861348327806e-05,
      "loss": 1.7688,
      "step": 320
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.990789419272199e-05,
      "loss": 1.8298,
      "step": 322
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.990590590221568e-05,
      "loss": 1.7999,
      "step": 324
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.990389648105017e-05,
      "loss": 1.8174,
      "step": 326
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9901865933511834e-05,
      "loss": 1.8054,
      "step": 328
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9899814263932106e-05,
      "loss": 1.8832,
      "step": 330
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9897741476687486e-05,
      "loss": 1.8076,
      "step": 332
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9895647576199507e-05,
      "loss": 1.749,
      "step": 334
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9893532566934747e-05,
      "loss": 1.83,
      "step": 336
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9891396453404817e-05,
      "loss": 1.8311,
      "step": 338
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.988923924016634e-05,
      "loss": 1.8504,
      "step": 340
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9887060931820937e-05,
      "loss": 1.8123,
      "step": 342
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9884861533015255e-05,
      "loss": 1.8611,
      "step": 344
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.988264104844091e-05,
      "loss": 1.7545,
      "step": 346
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9880399482834493e-05,
      "loss": 1.7604,
      "step": 348
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9878136840977577e-05,
      "loss": 1.7608,
      "step": 350
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.987585312769669e-05,
      "loss": 1.7772,
      "step": 352
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9873548347863306e-05,
      "loss": 1.8385,
      "step": 354
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9871222506393834e-05,
      "loss": 1.7471,
      "step": 356
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9868875608249613e-05,
      "loss": 1.8298,
      "step": 358
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9866507658436904e-05,
      "loss": 1.7782,
      "step": 360
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9864118662006868e-05,
      "loss": 1.766,
      "step": 362
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.986170862405556e-05,
      "loss": 1.7897,
      "step": 364
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9859277549723924e-05,
      "loss": 1.7589,
      "step": 366
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9856825444197777e-05,
      "loss": 1.8643,
      "step": 368
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.98543523127078e-05,
      "loss": 1.7779,
      "step": 370
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9851858160529512e-05,
      "loss": 1.8131,
      "step": 372
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9849342992983295e-05,
      "loss": 1.7722,
      "step": 374
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.984680681543434e-05,
      "loss": 1.7034,
      "step": 376
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.984424963329266e-05,
      "loss": 1.7491,
      "step": 378
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9841671452013082e-05,
      "loss": 1.8034,
      "step": 380
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9839072277095222e-05,
      "loss": 1.8034,
      "step": 382
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.983645211408347e-05,
      "loss": 1.7686,
      "step": 384
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9833810968567004e-05,
      "loss": 1.8068,
      "step": 386
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9831148846179743e-05,
      "loss": 1.7915,
      "step": 388
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.982846575260036e-05,
      "loss": 1.7891,
      "step": 390
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9825761693552264e-05,
      "loss": 1.7181,
      "step": 392
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9823036674803585e-05,
      "loss": 1.8185,
      "step": 394
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.982029070216717e-05,
      "loss": 1.8424,
      "step": 396
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.981752378150054e-05,
      "loss": 1.7818,
      "step": 398
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.981473591870593e-05,
      "loss": 1.7919,
      "step": 400
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.981192711973023e-05,
      "loss": 1.7922,
      "step": 402
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9809097390564996e-05,
      "loss": 1.7766,
      "step": 404
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.980624673724643e-05,
      "loss": 1.812,
      "step": 406
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9803375165855366e-05,
      "loss": 1.8114,
      "step": 408
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.980048268251726e-05,
      "loss": 1.7852,
      "step": 410
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9797569293402174e-05,
      "loss": 1.7116,
      "step": 412
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9794635004724774e-05,
      "loss": 1.7899,
      "step": 414
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9791679822744296e-05,
      "loss": 1.8268,
      "step": 416
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9788703753764554e-05,
      "loss": 1.7258,
      "step": 418
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9785706804133906e-05,
      "loss": 1.7856,
      "step": 420
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9782688980245263e-05,
      "loss": 1.7606,
      "step": 422
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9779650288536057e-05,
      "loss": 1.7671,
      "step": 424
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9776590735488238e-05,
      "loss": 1.8111,
      "step": 426
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9773510327628248e-05,
      "loss": 1.8157,
      "step": 428
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.977040907152702e-05,
      "loss": 1.8134,
      "step": 430
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.976728697379996e-05,
      "loss": 1.7108,
      "step": 432
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9764144041106942e-05,
      "loss": 1.7848,
      "step": 434
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.976098028015226e-05,
      "loss": 1.7785,
      "step": 436
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9757795697684656e-05,
      "loss": 1.7781,
      "step": 438
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9754590300497283e-05,
      "loss": 1.8016,
      "step": 440
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9751364095427694e-05,
      "loss": 1.6873,
      "step": 442
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9748117089357827e-05,
      "loss": 1.7731,
      "step": 444
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9744849289213994e-05,
      "loss": 1.7192,
      "step": 446
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.974156070196686e-05,
      "loss": 1.7577,
      "step": 448
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9738251334631435e-05,
      "loss": 1.7699,
      "step": 450
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9734921194267054e-05,
      "loss": 1.7899,
      "step": 452
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.973157028797737e-05,
      "loss": 1.7815,
      "step": 454
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9728198622910313e-05,
      "loss": 1.7487,
      "step": 456
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9724806206258117e-05,
      "loss": 1.7338,
      "step": 458
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9721393045257277e-05,
      "loss": 1.6726,
      "step": 460
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9717959147188527e-05,
      "loss": 1.766,
      "step": 462
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9714504519376846e-05,
      "loss": 1.8309,
      "step": 464
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9711029169191437e-05,
      "loss": 1.7432,
      "step": 466
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9707533104045686e-05,
      "loss": 1.8108,
      "step": 468
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9704016331397192e-05,
      "loss": 1.7305,
      "step": 470
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.970047885874771e-05,
      "loss": 1.8073,
      "step": 472
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9696920693643157e-05,
      "loss": 1.7198,
      "step": 474
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9693341843673586e-05,
      "loss": 1.7378,
      "step": 476
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.968974231647318e-05,
      "loss": 1.78,
      "step": 478
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.968612211972022e-05,
      "loss": 1.8091,
      "step": 480
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.968248126113709e-05,
      "loss": 1.779,
      "step": 482
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9678819748490236e-05,
      "loss": 1.765,
      "step": 484
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9675137589590173e-05,
      "loss": 1.7272,
      "step": 486
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.967143479229145e-05,
      "loss": 1.7872,
      "step": 488
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9667711364492638e-05,
      "loss": 1.7774,
      "step": 490
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9663967314136324e-05,
      "loss": 1.7369,
      "step": 492
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9660202649209088e-05,
      "loss": 1.8046,
      "step": 494
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.965641737774147e-05,
      "loss": 1.7097,
      "step": 496
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9652611507807977e-05,
      "loss": 1.7573,
      "step": 498
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9648785047527052e-05,
      "loss": 1.7586,
      "step": 500
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9644938005061062e-05,
      "loss": 1.735,
      "step": 502
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.964107038861628e-05,
      "loss": 1.7838,
      "step": 504
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9637182206442857e-05,
      "loss": 1.7658,
      "step": 506
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9633273466834826e-05,
      "loss": 1.7907,
      "step": 508
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9629344178130062e-05,
      "loss": 1.738,
      "step": 510
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.962539434871028e-05,
      "loss": 1.7868,
      "step": 512
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9621423987001013e-05,
      "loss": 1.8226,
      "step": 514
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9617433101471582e-05,
      "loss": 1.6927,
      "step": 516
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9613421700635098e-05,
      "loss": 1.7609,
      "step": 518
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.960938979304843e-05,
      "loss": 1.7614,
      "step": 520
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.960533738731219e-05,
      "loss": 1.7703,
      "step": 522
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.960126449207072e-05,
      "loss": 1.7205,
      "step": 524
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.959717111601206e-05,
      "loss": 1.6713,
      "step": 526
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9593057267867945e-05,
      "loss": 1.7879,
      "step": 528
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9588922956413783e-05,
      "loss": 1.8063,
      "step": 530
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9584768190468624e-05,
      "loss": 1.7567,
      "step": 532
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.958059297889516e-05,
      "loss": 1.7511,
      "step": 534
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9576397330599686e-05,
      "loss": 1.7831,
      "step": 536
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.95721812545321e-05,
      "loss": 1.7832,
      "step": 538
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9567944759685873e-05,
      "loss": 1.8,
      "step": 540
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9563687855098027e-05,
      "loss": 1.7094,
      "step": 542
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9559410549849125e-05,
      "loss": 1.778,
      "step": 544
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9555112853063254e-05,
      "loss": 1.6998,
      "step": 546
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9550794773907983e-05,
      "loss": 1.6977,
      "step": 548
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9546456321594374e-05,
      "loss": 1.7396,
      "step": 550
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9542097505376942e-05,
      "loss": 1.7353,
      "step": 552
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9537718334553637e-05,
      "loss": 1.7153,
      "step": 554
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9533318818465837e-05,
      "loss": 1.7359,
      "step": 556
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9528898966498315e-05,
      "loss": 1.7319,
      "step": 558
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9524458788079218e-05,
      "loss": 1.8195,
      "step": 560
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9519998292680062e-05,
      "loss": 1.7492,
      "step": 562
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9515517489815694e-05,
      "loss": 1.733,
      "step": 564
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9511016389044283e-05,
      "loss": 1.7876,
      "step": 566
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9506494999967298e-05,
      "loss": 1.6752,
      "step": 568
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9501953332229482e-05,
      "loss": 1.7158,
      "step": 570
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9497391395518836e-05,
      "loss": 1.6684,
      "step": 572
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.94928091995666e-05,
      "loss": 1.7932,
      "step": 574
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9488206754147223e-05,
      "loss": 1.7774,
      "step": 576
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.948358406907836e-05,
      "loss": 1.7292,
      "step": 578
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9478941154220833e-05,
      "loss": 1.8014,
      "step": 580
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9474278019478614e-05,
      "loss": 1.7008,
      "step": 582
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9469594674798814e-05,
      "loss": 1.7719,
      "step": 584
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9464891130171647e-05,
      "loss": 1.7625,
      "step": 586
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.946016739563042e-05,
      "loss": 1.7655,
      "step": 588
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.945542348125152e-05,
      "loss": 1.7899,
      "step": 590
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9450659397154353e-05,
      "loss": 1.7145,
      "step": 592
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9445875153501375e-05,
      "loss": 1.783,
      "step": 594
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.944107076049804e-05,
      "loss": 1.6651,
      "step": 596
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9436246228392762e-05,
      "loss": 1.7709,
      "step": 598
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.943140156747694e-05,
      "loss": 1.7286,
      "step": 600
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9426536788084905e-05,
      "loss": 1.7013,
      "step": 602
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.94216519005939e-05,
      "loss": 1.7354,
      "step": 604
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9416746915424057e-05,
      "loss": 1.6642,
      "step": 606
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9411821843038387e-05,
      "loss": 1.7365,
      "step": 608
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9406876693942747e-05,
      "loss": 1.6826,
      "step": 610
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9401911478685815e-05,
      "loss": 1.7372,
      "step": 612
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9396926207859085e-05,
      "loss": 1.7176,
      "step": 614
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.939192089209682e-05,
      "loss": 1.7928,
      "step": 616
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.938689554207605e-05,
      "loss": 1.719,
      "step": 618
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9381850168516535e-05,
      "loss": 1.8082,
      "step": 620
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9376784782180747e-05,
      "loss": 1.7874,
      "step": 622
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.937169939387386e-05,
      "loss": 1.7655,
      "step": 624
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.93665940144437e-05,
      "loss": 1.7579,
      "step": 626
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9361468654780748e-05,
      "loss": 1.7333,
      "step": 628
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9356323325818094e-05,
      "loss": 1.7346,
      "step": 630
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9351158038531438e-05,
      "loss": 1.6965,
      "step": 632
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9345972803939046e-05,
      "loss": 1.7706,
      "step": 634
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9340767633101735e-05,
      "loss": 1.7284,
      "step": 636
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9335542537122856e-05,
      "loss": 1.7154,
      "step": 638
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9330297527148246e-05,
      "loss": 1.6853,
      "step": 640
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9325032614366244e-05,
      "loss": 1.6915,
      "step": 642
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9319747810007625e-05,
      "loss": 1.7142,
      "step": 644
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9314443125345606e-05,
      "loss": 1.8119,
      "step": 646
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.930911857169581e-05,
      "loss": 1.6678,
      "step": 648
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.930377416041624e-05,
      "loss": 1.7466,
      "step": 650
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.929840990290726e-05,
      "loss": 1.7159,
      "step": 652
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.929302581061157e-05,
      "loss": 1.7208,
      "step": 654
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9287621895014178e-05,
      "loss": 1.7412,
      "step": 656
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.928219816764238e-05,
      "loss": 1.7257,
      "step": 658
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.927675464006573e-05,
      "loss": 1.6678,
      "step": 660
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9271291323896026e-05,
      "loss": 1.7721,
      "step": 662
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9265808230787265e-05,
      "loss": 1.75,
      "step": 664
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9260305372435643e-05,
      "loss": 1.7501,
      "step": 666
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9254782760579515e-05,
      "loss": 1.7389,
      "step": 668
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9249240406999366e-05,
      "loss": 1.7821,
      "step": 670
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9243678323517803e-05,
      "loss": 1.7417,
      "step": 672
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9238096521999514e-05,
      "loss": 1.7157,
      "step": 674
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9232495014351248e-05,
      "loss": 1.7157,
      "step": 676
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.922687381252179e-05,
      "loss": 1.661,
      "step": 678
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9221232928501937e-05,
      "loss": 1.8155,
      "step": 680
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.921557237432447e-05,
      "loss": 1.7296,
      "step": 682
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.920989216206413e-05,
      "loss": 1.7099,
      "step": 684
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.920419230383759e-05,
      "loss": 1.7449,
      "step": 686
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.919847281180343e-05,
      "loss": 1.7194,
      "step": 688
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.919273369816212e-05,
      "loss": 1.6699,
      "step": 690
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9186974975155967e-05,
      "loss": 1.7437,
      "step": 692
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9181196655069126e-05,
      "loss": 1.7165,
      "step": 694
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.917539875022755e-05,
      "loss": 1.7493,
      "step": 696
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9169581272998964e-05,
      "loss": 1.6878,
      "step": 698
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9163744235792845e-05,
      "loss": 1.7825,
      "step": 700
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9157887651060402e-05,
      "loss": 1.6875,
      "step": 702
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9152011531294524e-05,
      "loss": 1.6695,
      "step": 704
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9146115889029793e-05,
      "loss": 1.743,
      "step": 706
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.914020073684242e-05,
      "loss": 1.7069,
      "step": 708
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9134266087350243e-05,
      "loss": 1.7586,
      "step": 710
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.912831195321268e-05,
      "loss": 1.7369,
      "step": 712
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.912233834713072e-05,
      "loss": 1.7295,
      "step": 714
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.911634528184688e-05,
      "loss": 1.6867,
      "step": 716
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9110332770145198e-05,
      "loss": 1.7016,
      "step": 718
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.910430082485119e-05,
      "loss": 1.6782,
      "step": 720
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9098249458831822e-05,
      "loss": 1.683,
      "step": 722
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9092178684995487e-05,
      "loss": 1.6822,
      "step": 724
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9086088516291986e-05,
      "loss": 1.6561,
      "step": 726
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.907997896571248e-05,
      "loss": 1.662,
      "step": 728
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9073850046289484e-05,
      "loss": 1.7087,
      "step": 730
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9067701771096826e-05,
      "loss": 1.7309,
      "step": 732
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9061534153249622e-05,
      "loss": 1.7506,
      "step": 734
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9055347205904245e-05,
      "loss": 1.7252,
      "step": 736
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9049140942258312e-05,
      "loss": 1.619,
      "step": 738
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.904291537555063e-05,
      "loss": 1.7701,
      "step": 740
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.903667051906119e-05,
      "loss": 1.6819,
      "step": 742
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.903040638611113e-05,
      "loss": 1.6712,
      "step": 744
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9024122990062707e-05,
      "loss": 1.7043,
      "step": 746
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.901782034431927e-05,
      "loss": 1.755,
      "step": 748
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9011498462325226e-05,
      "loss": 1.6906,
      "step": 750
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9005157357566022e-05,
      "loss": 1.7031,
      "step": 752
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8998797043568102e-05,
      "loss": 1.695,
      "step": 754
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8992417533898898e-05,
      "loss": 1.641,
      "step": 756
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.898601884216677e-05,
      "loss": 1.6758,
      "step": 758
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8979600982021014e-05,
      "loss": 1.7431,
      "step": 760
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.897316396715181e-05,
      "loss": 1.6965,
      "step": 762
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8966707811290188e-05,
      "loss": 1.7556,
      "step": 764
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.896023252820802e-05,
      "loss": 1.67,
      "step": 766
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.895373813171798e-05,
      "loss": 1.7312,
      "step": 768
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8947224635673503e-05,
      "loss": 1.6034,
      "step": 770
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.8940692053968773e-05,
      "loss": 1.7197,
      "step": 772
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.8934140400538687e-05,
      "loss": 1.6933,
      "step": 774
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.8927569689358818e-05,
      "loss": 1.7197,
      "step": 776
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.89209799344454e-05,
      "loss": 1.6594,
      "step": 778
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.891437114985528e-05,
      "loss": 1.7776,
      "step": 780
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8907743349685917e-05,
      "loss": 1.657,
      "step": 782
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8901096548075305e-05,
      "loss": 1.6959,
      "step": 784
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8894430759201996e-05,
      "loss": 1.6839,
      "step": 786
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.888774599728503e-05,
      "loss": 1.7109,
      "step": 788
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8881042276583924e-05,
      "loss": 1.6902,
      "step": 790
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8874319611398636e-05,
      "loss": 1.6338,
      "step": 792
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8867578016069535e-05,
      "loss": 1.6391,
      "step": 794
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8860817504977374e-05,
      "loss": 1.6426,
      "step": 796
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.885403809254325e-05,
      "loss": 1.6807,
      "step": 798
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8847239793228572e-05,
      "loss": 1.6501,
      "step": 800
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8840422621535067e-05,
      "loss": 1.6706,
      "step": 802
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.883358659200468e-05,
      "loss": 1.7352,
      "step": 804
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8826731719219615e-05,
      "loss": 1.6896,
      "step": 806
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.881985801780225e-05,
      "loss": 1.743,
      "step": 808
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8812965502415145e-05,
      "loss": 1.7311,
      "step": 810
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8806054187760974e-05,
      "loss": 1.6658,
      "step": 812
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8799124088582523e-05,
      "loss": 1.6806,
      "step": 814
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.879217521966265e-05,
      "loss": 1.6803,
      "step": 816
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8785207595824243e-05,
      "loss": 1.7223,
      "step": 818
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8778221231930204e-05,
      "loss": 1.6689,
      "step": 820
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8771216142883407e-05,
      "loss": 1.6601,
      "step": 822
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.876419234362667e-05,
      "loss": 1.6667,
      "step": 824
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8757149849142724e-05,
      "loss": 1.7026,
      "step": 826
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8750088674454178e-05,
      "loss": 1.7311,
      "step": 828
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8743008834623486e-05,
      "loss": 1.7196,
      "step": 830
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8735910344752925e-05,
      "loss": 1.7032,
      "step": 832
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.872879321998454e-05,
      "loss": 1.7119,
      "step": 834
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8721657475500146e-05,
      "loss": 1.7413,
      "step": 836
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.871450312652126e-05,
      "loss": 1.6728,
      "step": 838
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.87073301883091e-05,
      "loss": 1.7625,
      "step": 840
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8700138676164523e-05,
      "loss": 1.7422,
      "step": 842
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8692928605428016e-05,
      "loss": 1.6934,
      "step": 844
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8685699991479648e-05,
      "loss": 1.6893,
      "step": 846
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.867845284973905e-05,
      "loss": 1.6688,
      "step": 848
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.8671187195665373e-05,
      "loss": 1.6917,
      "step": 850
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.866390304475725e-05,
      "loss": 1.7303,
      "step": 852
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.865660041255278e-05,
      "loss": 1.7043,
      "step": 854
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.8649279314629484e-05,
      "loss": 1.6968,
      "step": 856
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.8641939766604264e-05,
      "loss": 1.7125,
      "step": 858
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.863458178413339e-05,
      "loss": 1.7805,
      "step": 860
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.862720538291245e-05,
      "loss": 1.7479,
      "step": 862
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.861981057867632e-05,
      "loss": 1.6536,
      "step": 864
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.8612397387199133e-05,
      "loss": 1.6844,
      "step": 866
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.8604965824294253e-05,
      "loss": 1.7561,
      "step": 868
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.8597515905814215e-05,
      "loss": 1.6265,
      "step": 870
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8590047647650724e-05,
      "loss": 1.6159,
      "step": 872
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8582561065734602e-05,
      "loss": 1.7576,
      "step": 874
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8575056176035758e-05,
      "loss": 1.7615,
      "step": 876
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8567532994563155e-05,
      "loss": 1.7376,
      "step": 878
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8559991537364767e-05,
      "loss": 1.7,
      "step": 880
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.855243182052757e-05,
      "loss": 1.8079,
      "step": 882
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8544853860177466e-05,
      "loss": 1.7259,
      "step": 884
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8537257672479293e-05,
      "loss": 1.6671,
      "step": 886
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.852964327363677e-05,
      "loss": 1.6526,
      "step": 888
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8522010679892443e-05,
      "loss": 1.7153,
      "step": 890
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8514359907527693e-05,
      "loss": 1.6349,
      "step": 892
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.850669097286267e-05,
      "loss": 1.6579,
      "step": 894
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.849900389225626e-05,
      "loss": 1.6429,
      "step": 896
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8491298682106066e-05,
      "loss": 1.7093,
      "step": 898
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8483575358848363e-05,
      "loss": 1.7222,
      "step": 900
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.847583393895806e-05,
      "loss": 1.6711,
      "step": 902
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8468074438948664e-05,
      "loss": 1.6748,
      "step": 904
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8460296875372267e-05,
      "loss": 1.6802,
      "step": 906
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.845250126481947e-05,
      "loss": 1.7504,
      "step": 908
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8444687623919388e-05,
      "loss": 1.685,
      "step": 910
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8436855969339592e-05,
      "loss": 1.6664,
      "step": 912
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8429006317786073e-05,
      "loss": 1.6405,
      "step": 914
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.842113868600322e-05,
      "loss": 1.6919,
      "step": 916
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8413253090773773e-05,
      "loss": 1.7471,
      "step": 918
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8405349548918792e-05,
      "loss": 1.6325,
      "step": 920
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8397428077297622e-05,
      "loss": 1.6538,
      "step": 922
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.838948869280784e-05,
      "loss": 1.7362,
      "step": 924
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8381531412385253e-05,
      "loss": 1.6641,
      "step": 926
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.837355625300383e-05,
      "loss": 1.6573,
      "step": 928
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8365563231675693e-05,
      "loss": 1.748,
      "step": 930
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8357552365451038e-05,
      "loss": 1.6665,
      "step": 932
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.834952367141816e-05,
      "loss": 1.6976,
      "step": 934
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8341477166703356e-05,
      "loss": 1.6153,
      "step": 936
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8333412868470934e-05,
      "loss": 1.6839,
      "step": 938
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8325330793923146e-05,
      "loss": 1.7304,
      "step": 940
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.831723096030017e-05,
      "loss": 1.6834,
      "step": 942
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8309113384880065e-05,
      "loss": 1.7582,
      "step": 944
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8300978084978736e-05,
      "loss": 1.6958,
      "step": 946
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8292825077949895e-05,
      "loss": 1.6799,
      "step": 948
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.828465438118503e-05,
      "loss": 1.6473,
      "step": 950
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8276466012113358e-05,
      "loss": 1.6568,
      "step": 952
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.826825998820179e-05,
      "loss": 1.635,
      "step": 954
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8260036326954914e-05,
      "loss": 1.7147,
      "step": 956
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8251795045914922e-05,
      "loss": 1.6833,
      "step": 958
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.824353616266161e-05,
      "loss": 1.7482,
      "step": 960
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8235259694812296e-05,
      "loss": 1.7112,
      "step": 962
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8226965660021836e-05,
      "loss": 1.6922,
      "step": 964
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.821865407598254e-05,
      "loss": 1.7317,
      "step": 966
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8210324960424163e-05,
      "loss": 1.645,
      "step": 968
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8201978331113855e-05,
      "loss": 1.6681,
      "step": 970
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8193614205856118e-05,
      "loss": 1.7141,
      "step": 972
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8185232602492785e-05,
      "loss": 1.6721,
      "step": 974
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.817683353890297e-05,
      "loss": 1.6618,
      "step": 976
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8168417033003026e-05,
      "loss": 1.6555,
      "step": 978
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8159983102746522e-05,
      "loss": 1.6415,
      "step": 980
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8151531766124186e-05,
      "loss": 1.7108,
      "step": 982
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8143063041163885e-05,
      "loss": 1.6733,
      "step": 984
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.813457694593057e-05,
      "loss": 1.6924,
      "step": 986
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8126073498526254e-05,
      "loss": 1.7121,
      "step": 988
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.811755271708995e-05,
      "loss": 1.6959,
      "step": 990
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8109014619797667e-05,
      "loss": 1.6739,
      "step": 992
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8100459224862336e-05,
      "loss": 1.7133,
      "step": 994
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.809188655053379e-05,
      "loss": 1.6585,
      "step": 996
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8083296615098724e-05,
      "loss": 1.6682,
      "step": 998
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8074689436880643e-05,
      "loss": 1.6857,
      "step": 1000
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8066065034239854e-05,
      "loss": 1.711,
      "step": 1002
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8057423425573385e-05,
      "loss": 1.7064,
      "step": 1004
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.804876462931498e-05,
      "loss": 1.6845,
      "step": 1006
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8040088663935037e-05,
      "loss": 1.6651,
      "step": 1008
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.803139554794059e-05,
      "loss": 1.6555,
      "step": 1010
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8022685299875245e-05,
      "loss": 1.7049,
      "step": 1012
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8013957938319158e-05,
      "loss": 1.6926,
      "step": 1014
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8005213481888994e-05,
      "loss": 1.6573,
      "step": 1016
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.799645194923788e-05,
      "loss": 1.6412,
      "step": 1018
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7987673359055372e-05,
      "loss": 1.6951,
      "step": 1020
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7978877730067403e-05,
      "loss": 1.6705,
      "step": 1022
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7970065081036266e-05,
      "loss": 1.6474,
      "step": 1024
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.796123543076055e-05,
      "loss": 1.6242,
      "step": 1026
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.7952388798075112e-05,
      "loss": 1.6494,
      "step": 1028
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.7943525201851038e-05,
      "loss": 1.6096,
      "step": 1030
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.79346446609956e-05,
      "loss": 1.6974,
      "step": 1032
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.7925747194452204e-05,
      "loss": 1.5565,
      "step": 1034
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.7916832821200375e-05,
      "loss": 1.6754,
      "step": 1036
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7907901560255697e-05,
      "loss": 1.6594,
      "step": 1038
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.789895343066978e-05,
      "loss": 1.7223,
      "step": 1040
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7889988451530208e-05,
      "loss": 1.663,
      "step": 1042
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.788100664196052e-05,
      "loss": 1.7279,
      "step": 1044
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7872008021120146e-05,
      "loss": 1.6634,
      "step": 1046
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7862992608204384e-05,
      "loss": 1.7776,
      "step": 1048
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.7853960422444346e-05,
      "loss": 1.6886,
      "step": 1050
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.784491148310693e-05,
      "loss": 1.6632,
      "step": 1052
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.783584580949477e-05,
      "loss": 1.6862,
      "step": 1054
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.782676342094619e-05,
      "loss": 1.6353,
      "step": 1056
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.781766433683517e-05,
      "loss": 1.6897,
      "step": 1058
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.7808548576571314e-05,
      "loss": 1.6054,
      "step": 1060
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.7799416159599786e-05,
      "loss": 1.7094,
      "step": 1062
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.7790267105401295e-05,
      "loss": 1.6769,
      "step": 1064
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.7781101433492026e-05,
      "loss": 1.6944,
      "step": 1066
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.7771919163423617e-05,
      "loss": 1.7149,
      "step": 1068
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.776272031478311e-05,
      "loss": 1.6351,
      "step": 1070
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.7753504907192923e-05,
      "loss": 1.648,
      "step": 1072
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.7744272960310782e-05,
      "loss": 1.7506,
      "step": 1074
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.7735024493829696e-05,
      "loss": 1.6794,
      "step": 1076
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.7725759527477923e-05,
      "loss": 1.6343,
      "step": 1078
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.7716478081018897e-05,
      "loss": 1.7295,
      "step": 1080
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.7707180174251232e-05,
      "loss": 1.6723,
      "step": 1082
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.769786582700864e-05,
      "loss": 1.7363,
      "step": 1084
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.7688535059159896e-05,
      "loss": 1.7436,
      "step": 1086
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.767918789060882e-05,
      "loss": 1.6605,
      "step": 1088
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.7669824341294203e-05,
      "loss": 1.7059,
      "step": 1090
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.766044443118978e-05,
      "loss": 1.607,
      "step": 1092
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7651048180304196e-05,
      "loss": 1.5689,
      "step": 1094
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7641635608680942e-05,
      "loss": 1.7319,
      "step": 1096
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.763220673639833e-05,
      "loss": 1.6465,
      "step": 1098
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7622761583569438e-05,
      "loss": 1.5836,
      "step": 1100
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7613300170342073e-05,
      "loss": 1.6483,
      "step": 1102
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.7603822516898735e-05,
      "loss": 1.6409,
      "step": 1104
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.759432864345656e-05,
      "loss": 1.724,
      "step": 1106
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.7584818570267287e-05,
      "loss": 1.6458,
      "step": 1108
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.757529231761721e-05,
      "loss": 1.6573,
      "step": 1110
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.756574990582713e-05,
      "loss": 1.7138,
      "step": 1112
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.755619135525233e-05,
      "loss": 1.5559,
      "step": 1114
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.754661668628251e-05,
      "loss": 1.7281,
      "step": 1116
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.753702591934175e-05,
      "loss": 1.6373,
      "step": 1118
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7527419074888483e-05,
      "loss": 1.6487,
      "step": 1120
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7517796173415423e-05,
      "loss": 1.6042,
      "step": 1122
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7508157235449547e-05,
      "loss": 1.6037,
      "step": 1124
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.749850228155203e-05,
      "loss": 1.5873,
      "step": 1126
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7488831332318224e-05,
      "loss": 1.5981,
      "step": 1128
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7479144408377583e-05,
      "loss": 1.5995,
      "step": 1130
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7469441530393652e-05,
      "loss": 1.6619,
      "step": 1132
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7459722719064004e-05,
      "loss": 1.6442,
      "step": 1134
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.74499879951202e-05,
      "loss": 1.6343,
      "step": 1136
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7440237379327745e-05,
      "loss": 1.6415,
      "step": 1138
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.743047089248604e-05,
      "loss": 1.6876,
      "step": 1140
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7420688555428348e-05,
      "loss": 1.6451,
      "step": 1142
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7410890389021737e-05,
      "loss": 1.6764,
      "step": 1144
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7401076414167045e-05,
      "loss": 1.6495,
      "step": 1146
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7391246651798822e-05,
      "loss": 1.6299,
      "step": 1148
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7381401122885316e-05,
      "loss": 1.6415,
      "step": 1150
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.737153984842838e-05,
      "loss": 1.6169,
      "step": 1152
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7361662849463477e-05,
      "loss": 1.5897,
      "step": 1154
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7351770147059604e-05,
      "loss": 1.6446,
      "step": 1156
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.734186176231925e-05,
      "loss": 1.6889,
      "step": 1158
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7331937716378363e-05,
      "loss": 1.6264,
      "step": 1160
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7321998030406303e-05,
      "loss": 1.6497,
      "step": 1162
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7312042725605778e-05,
      "loss": 1.6199,
      "step": 1164
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7302071823212828e-05,
      "loss": 1.6627,
      "step": 1166
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.729208534449676e-05,
      "loss": 1.6577,
      "step": 1168
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.7282083310760097e-05,
      "loss": 1.6438,
      "step": 1170
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.7272065743338565e-05,
      "loss": 1.6896,
      "step": 1172
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.7262032663601003e-05,
      "loss": 1.5625,
      "step": 1174
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.725198409294935e-05,
      "loss": 1.6492,
      "step": 1176
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.7241920052818592e-05,
      "loss": 1.6704,
      "step": 1178
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.723184056467671e-05,
      "loss": 1.6208,
      "step": 1180
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.722174565002464e-05,
      "loss": 1.6938,
      "step": 1182
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.721163533039621e-05,
      "loss": 1.6304,
      "step": 1184
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.7201509627358143e-05,
      "loss": 1.6077,
      "step": 1186
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.719644101380906e-05,
      "loss": 1.6241,
      "step": 1188
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.718629227616585e-05,
      "loss": 1.5494,
      "step": 1190
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.7176128209173214e-05,
      "loss": 1.6192,
      "step": 1192
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7165948834512498e-05,
      "loss": 1.5993,
      "step": 1194
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7155754173897718e-05,
      "loss": 1.5976,
      "step": 1196
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7145544249075473e-05,
      "loss": 1.627,
      "step": 1198
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7135319081824945e-05,
      "loss": 1.6417,
      "step": 1200
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7125078693957817e-05,
      "loss": 1.577,
      "step": 1202
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7114823107318245e-05,
      "loss": 1.6091,
      "step": 1204
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7104552343782814e-05,
      "loss": 1.6435,
      "step": 1206
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.709426642526046e-05,
      "loss": 1.5965,
      "step": 1208
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7083965373692476e-05,
      "loss": 1.6498,
      "step": 1210
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.707364921105241e-05,
      "loss": 1.6697,
      "step": 1212
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.706331795934606e-05,
      "loss": 1.6733,
      "step": 1214
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7052971640611406e-05,
      "loss": 1.6135,
      "step": 1216
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7042610276918566e-05,
      "loss": 1.624,
      "step": 1218
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7032233890369755e-05,
      "loss": 1.6201,
      "step": 1220
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7021842503099236e-05,
      "loss": 1.6666,
      "step": 1222
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.701143613727326e-05,
      "loss": 1.6264,
      "step": 1224
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.700101481509004e-05,
      "loss": 1.7085,
      "step": 1226
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.699057855877969e-05,
      "loss": 1.5828,
      "step": 1228
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.6980127390604176e-05,
      "loss": 1.563,
      "step": 1230
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.6969661332857278e-05,
      "loss": 1.5967,
      "step": 1232
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.6959180407864535e-05,
      "loss": 1.6464,
      "step": 1234
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.6948684637983202e-05,
      "loss": 1.6169,
      "step": 1236
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.6938174045602203e-05,
      "loss": 1.6654,
      "step": 1238
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.692764865314207e-05,
      "loss": 1.6086,
      "step": 1240
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.6917108483054917e-05,
      "loss": 1.626,
      "step": 1242
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.6906553557824372e-05,
      "loss": 1.656,
      "step": 1244
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.6895983899965546e-05,
      "loss": 1.5495,
      "step": 1246
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.6885399532024965e-05,
      "loss": 1.7189,
      "step": 1248
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.6874800476580553e-05,
      "loss": 1.6218,
      "step": 1250
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.6864186756241544e-05,
      "loss": 1.6006,
      "step": 1252
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.685355839364846e-05,
      "loss": 1.6826,
      "step": 1254
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.684291541147307e-05,
      "loss": 1.6151,
      "step": 1256
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.683225783241831e-05,
      "loss": 1.5954,
      "step": 1258
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.682158567921826e-05,
      "loss": 1.6481,
      "step": 1260
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.6810898974638098e-05,
      "loss": 1.5727,
      "step": 1262
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.6800197741474028e-05,
      "loss": 1.6376,
      "step": 1264
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.6789482002553252e-05,
      "loss": 1.6167,
      "step": 1266
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.677875178073392e-05,
      "loss": 1.6303,
      "step": 1268
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.6768007098905058e-05,
      "loss": 1.5653,
      "step": 1270
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.6757247979986564e-05,
      "loss": 1.6156,
      "step": 1272
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.674647444692911e-05,
      "loss": 1.576,
      "step": 1274
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.6735686522714126e-05,
      "loss": 1.5843,
      "step": 1276
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.6724884230353737e-05,
      "loss": 1.6331,
      "step": 1278
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.6714067592890713e-05,
      "loss": 1.5994,
      "step": 1280
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.6703236633398436e-05,
      "loss": 1.5567,
      "step": 1282
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.6692391374980827e-05,
      "loss": 1.5959,
      "step": 1284
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.6681531840772314e-05,
      "loss": 1.6335,
      "step": 1286
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.667065805393778e-05,
      "loss": 1.7096,
      "step": 1288
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.66597700376725e-05,
      "loss": 1.5952,
      "step": 1290
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.664886781520212e-05,
      "loss": 1.6733,
      "step": 1292
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.6637951409782568e-05,
      "loss": 1.6305,
      "step": 1294
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.662702084470005e-05,
      "loss": 1.6955,
      "step": 1296
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.661607614327095e-05,
      "loss": 1.5969,
      "step": 1298
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.660511732884183e-05,
      "loss": 1.6004,
      "step": 1300
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.659414442478935e-05,
      "loss": 1.5801,
      "step": 1302
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6583157454520214e-05,
      "loss": 1.6222,
      "step": 1304
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6572156441471146e-05,
      "loss": 1.566,
      "step": 1306
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6561141409108825e-05,
      "loss": 1.6098,
      "step": 1308
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6550112380929814e-05,
      "loss": 1.5743,
      "step": 1310
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6539069380460556e-05,
      "loss": 1.5831,
      "step": 1312
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.6528012431257292e-05,
      "loss": 1.6536,
      "step": 1314
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.651694155690601e-05,
      "loss": 1.6429,
      "step": 1316
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.6505856781022413e-05,
      "loss": 1.5991,
      "step": 1318
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.649475812725184e-05,
      "loss": 1.6168,
      "step": 1320
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.6483645619269256e-05,
      "loss": 1.5556,
      "step": 1322
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.6472519280779164e-05,
      "loss": 1.6329,
      "step": 1324
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.6461379135515575e-05,
      "loss": 1.603,
      "step": 1326
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.645022520724195e-05,
      "loss": 1.6533,
      "step": 1328
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.6439057519751155e-05,
      "loss": 1.6451,
      "step": 1330
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.6427876096865394e-05,
      "loss": 1.6017,
      "step": 1332
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.641668096243619e-05,
      "loss": 1.5915,
      "step": 1334
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.6405472140344294e-05,
      "loss": 1.592,
      "step": 1336
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.6399862603895563e-05,
      "loss": 1.5916,
      "step": 1338
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.638863329514992e-05,
      "loss": 1.6036,
      "step": 1340
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.637739035857112e-05,
      "loss": 1.6597,
      "step": 1342
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.6366133818141893e-05,
      "loss": 1.647,
      "step": 1344
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.635486369787399e-05,
      "loss": 1.6059,
      "step": 1346
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.634358002180812e-05,
      "loss": 1.583,
      "step": 1348
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.633228281401392e-05,
      "loss": 1.5999,
      "step": 1350
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.6320972098589883e-05,
      "loss": 1.5893,
      "step": 1352
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.6309647899663332e-05,
      "loss": 1.6109,
      "step": 1354
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.6298310241390326e-05,
      "loss": 1.6077,
      "step": 1356
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.6286959147955656e-05,
      "loss": 1.668,
      "step": 1358
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.627559464357276e-05,
      "loss": 1.5566,
      "step": 1360
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.6264216752483697e-05,
      "loss": 1.5587,
      "step": 1362
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.625282549895907e-05,
      "loss": 1.5886,
      "step": 1364
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.624142090729798e-05,
      "loss": 1.5838,
      "step": 1366
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.6230003001828e-05,
      "loss": 1.5925,
      "step": 1368
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.6218571806905088e-05,
      "loss": 1.5977,
      "step": 1370
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.620712734691355e-05,
      "loss": 1.5983,
      "step": 1372
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.6195669646266003e-05,
      "loss": 1.6395,
      "step": 1374
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.6184198729403288e-05,
      "loss": 1.5128,
      "step": 1376
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.6172714620794455e-05,
      "loss": 1.5381,
      "step": 1378
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.616121734493668e-05,
      "loss": 1.5826,
      "step": 1380
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.6149706926355237e-05,
      "loss": 1.5407,
      "step": 1382
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.6138183389603427e-05,
      "loss": 1.5566,
      "step": 1384
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.6126646759262548e-05,
      "loss": 1.5748,
      "step": 1386
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.6115097059941807e-05,
      "loss": 1.6607,
      "step": 1388
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.610353431627831e-05,
      "loss": 1.6268,
      "step": 1390
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.609195855293697e-05,
      "loss": 1.5607,
      "step": 1392
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.6080369794610486e-05,
      "loss": 1.6097,
      "step": 1394
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.606876806601928e-05,
      "loss": 1.6064,
      "step": 1396
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.6057153391911422e-05,
      "loss": 1.7139,
      "step": 1398
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.6045525797062614e-05,
      "loss": 1.563,
      "step": 1400
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.6033885306276117e-05,
      "loss": 1.6391,
      "step": 1402
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.6022231944382693e-05,
      "loss": 1.6241,
      "step": 1404
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.601056573624057e-05,
      "loss": 1.707,
      "step": 1406
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.599888670673537e-05,
      "loss": 1.5843,
      "step": 1408
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.598719488078007e-05,
      "loss": 1.5681,
      "step": 1410
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.5975490283314946e-05,
      "loss": 1.6174,
      "step": 1412
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5963772939307508e-05,
      "loss": 1.6542,
      "step": 1414
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5952042873752463e-05,
      "loss": 1.5881,
      "step": 1416
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5940300111671652e-05,
      "loss": 1.6124,
      "step": 1418
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5928544678114007e-05,
      "loss": 1.5708,
      "step": 1420
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5916776598155478e-05,
      "loss": 1.6167,
      "step": 1422
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.5904995896899004e-05,
      "loss": 1.4885,
      "step": 1424
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.5893202599474434e-05,
      "loss": 1.5958,
      "step": 1426
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.5881396731038493e-05,
      "loss": 1.625,
      "step": 1428
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.5869578316774724e-05,
      "loss": 1.5973,
      "step": 1430
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.585774738189343e-05,
      "loss": 1.6034,
      "step": 1432
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.584590395163162e-05,
      "loss": 1.5719,
      "step": 1434
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.583404805125296e-05,
      "loss": 1.5804,
      "step": 1436
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.5822179706047717e-05,
      "loss": 1.6417,
      "step": 1438
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.5810298941332696e-05,
      "loss": 1.6222,
      "step": 1440
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.5798405782451216e-05,
      "loss": 1.5941,
      "step": 1442
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.5786500254773005e-05,
      "loss": 1.6113,
      "step": 1444
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.5774582383694196e-05,
      "loss": 1.5371,
      "step": 1446
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.5762652194637247e-05,
      "loss": 1.6054,
      "step": 1448
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.575070971305089e-05,
      "loss": 1.5252,
      "step": 1450
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.5738754964410084e-05,
      "loss": 1.5853,
      "step": 1452
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.5726787974215948e-05,
      "loss": 1.6497,
      "step": 1454
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.5714808767995716e-05,
      "loss": 1.5488,
      "step": 1456
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.5702817371302684e-05,
      "loss": 1.678,
      "step": 1458
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.5690813809716153e-05,
      "loss": 1.6052,
      "step": 1460
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.5678798108841366e-05,
      "loss": 1.601,
      "step": 1462
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.5666770294309467e-05,
      "loss": 1.6847,
      "step": 1464
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.5654730391777434e-05,
      "loss": 1.5923,
      "step": 1466
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.5642678426928036e-05,
      "loss": 1.59,
      "step": 1468
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.5630614425469776e-05,
      "loss": 1.5686,
      "step": 1470
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.561853841313682e-05,
      "loss": 1.6082,
      "step": 1472
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.5606450415688968e-05,
      "loss": 1.6408,
      "step": 1474
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.5594350458911586e-05,
      "loss": 1.6029,
      "step": 1476
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.558223856861553e-05,
      "loss": 1.5787,
      "step": 1478
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.557011477063714e-05,
      "loss": 1.6162,
      "step": 1480
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.5557979090838136e-05,
      "loss": 1.5908,
      "step": 1482
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.5545831555105598e-05,
      "loss": 1.6125,
      "step": 1484
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.553367218935188e-05,
      "loss": 1.6002,
      "step": 1486
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.55215010195146e-05,
      "loss": 1.6325,
      "step": 1488
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.5509318071556513e-05,
      "loss": 1.6437,
      "step": 1490
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.5497123371465537e-05,
      "loss": 1.6507,
      "step": 1492
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.5484916945254642e-05,
      "loss": 1.6168,
      "step": 1494
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.547269881896181e-05,
      "loss": 1.6525,
      "step": 1496
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.546046901864999e-05,
      "loss": 1.599,
      "step": 1498
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.5448227570407012e-05,
      "loss": 1.601,
      "step": 1500
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.543597450034559e-05,
      "loss": 1.681,
      "step": 1502
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.542370983460319e-05,
      "loss": 1.6409,
      "step": 1504
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.5411433599342038e-05,
      "loss": 1.5691,
      "step": 1506
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.539914582074903e-05,
      "loss": 1.594,
      "step": 1508
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.538684652503569e-05,
      "loss": 1.7021,
      "step": 1510
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.5374535738438105e-05,
      "loss": 1.5613,
      "step": 1512
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.536221348721687e-05,
      "loss": 1.5947,
      "step": 1514
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.5349879797657045e-05,
      "loss": 1.6185,
      "step": 1516
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.5337534696068088e-05,
      "loss": 1.6027,
      "step": 1518
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.5325178208783793e-05,
      "loss": 1.6229,
      "step": 1520
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.5312810362162253e-05,
      "loss": 1.6062,
      "step": 1522
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.5300431182585777e-05,
      "loss": 1.5713,
      "step": 1524
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.5288040696460863e-05,
      "loss": 1.5456,
      "step": 1526
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.5275638930218122e-05,
      "loss": 1.63,
      "step": 1528
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.5263225910312222e-05,
      "loss": 1.5683,
      "step": 1530
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.5250801663221843e-05,
      "loss": 1.7049,
      "step": 1532
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.5238366215449611e-05,
      "loss": 1.628,
      "step": 1534
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.5225919593522049e-05,
      "loss": 1.5985,
      "step": 1536
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.521346182398951e-05,
      "loss": 1.6315,
      "step": 1538
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.520099293342613e-05,
      "loss": 1.6427,
      "step": 1540
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.5188512948429765e-05,
      "loss": 1.6735,
      "step": 1542
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.5176021895621944e-05,
      "loss": 1.5767,
      "step": 1544
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.5163519801647795e-05,
      "loss": 1.6415,
      "step": 1546
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5151006693176005e-05,
      "loss": 1.6607,
      "step": 1548
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5138482596898754e-05,
      "loss": 1.618,
      "step": 1550
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5125947539531663e-05,
      "loss": 1.5899,
      "step": 1552
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5113401547813732e-05,
      "loss": 1.6436,
      "step": 1554
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5100844648507284e-05,
      "loss": 1.6373,
      "step": 1556
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.5088276868397915e-05,
      "loss": 1.6599,
      "step": 1558
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.5075698234294424e-05,
      "loss": 1.5582,
      "step": 1560
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.5063108773028771e-05,
      "loss": 1.6258,
      "step": 1562
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.5050508511456006e-05,
      "loss": 1.6312,
      "step": 1564
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.5037897476454219e-05,
      "loss": 1.5919,
      "step": 1566
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.5025275694924481e-05,
      "loss": 1.5423,
      "step": 1568
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.5012643193790788e-05,
      "loss": 1.6427,
      "step": 1570
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 1.5718,
      "step": 1572
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.4987346140521792e-05,
      "loss": 1.6259,
      "step": 1574
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.4974681642348583e-05,
      "loss": 1.5403,
      "step": 1576
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.496200653249549e-05,
      "loss": 1.5996,
      "step": 1578
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.4949320838000261e-05,
      "loss": 1.4791,
      "step": 1580
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.493662458592323e-05,
      "loss": 1.6647,
      "step": 1582
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.492391780334725e-05,
      "loss": 1.608,
      "step": 1584
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.4911200517377628e-05,
      "loss": 1.618,
      "step": 1586
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.4898472755142095e-05,
      "loss": 1.5103,
      "step": 1588
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.4885734543790707e-05,
      "loss": 1.5614,
      "step": 1590
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4872985910495822e-05,
      "loss": 1.6294,
      "step": 1592
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4860226882452033e-05,
      "loss": 1.6391,
      "step": 1594
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4847457486876097e-05,
      "loss": 1.6182,
      "step": 1596
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4834677751006891e-05,
      "loss": 1.623,
      "step": 1598
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4821887702105348e-05,
      "loss": 1.6026,
      "step": 1600
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.4809087367454402e-05,
      "loss": 1.6163,
      "step": 1602
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.4796276774358927e-05,
      "loss": 1.5841,
      "step": 1604
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.4783455950145675e-05,
      "loss": 1.5542,
      "step": 1606
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.4770624922163233e-05,
      "loss": 1.6349,
      "step": 1608
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.4757783717781942e-05,
      "loss": 1.6687,
      "step": 1610
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.4744932364393863e-05,
      "loss": 1.7306,
      "step": 1612
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4732070889412693e-05,
      "loss": 1.6088,
      "step": 1614
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4719199320273729e-05,
      "loss": 1.5419,
      "step": 1616
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4706317684433802e-05,
      "loss": 1.6691,
      "step": 1618
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4693426009371203e-05,
      "loss": 1.6314,
      "step": 1620
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4680524322585656e-05,
      "loss": 1.5798,
      "step": 1622
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.4667612651598229e-05,
      "loss": 1.6116,
      "step": 1624
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.4654691023951289e-05,
      "loss": 1.5479,
      "step": 1626
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.4641759467208448e-05,
      "loss": 1.61,
      "step": 1628
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.4628818008954492e-05,
      "loss": 1.5859,
      "step": 1630
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.4615866676795334e-05,
      "loss": 1.5711,
      "step": 1632
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.4602905498357944e-05,
      "loss": 1.567,
      "step": 1634
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4589934501290297e-05,
      "loss": 1.5867,
      "step": 1636
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4576953713261313e-05,
      "loss": 1.6344,
      "step": 1638
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4563963161960799e-05,
      "loss": 1.6663,
      "step": 1640
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4550962875099385e-05,
      "loss": 1.659,
      "step": 1642
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4537952880408472e-05,
      "loss": 1.6495,
      "step": 1644
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.4524933205640168e-05,
      "loss": 1.5322,
      "step": 1646
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.4511903878567229e-05,
      "loss": 1.5643,
      "step": 1648
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.4498864926982996e-05,
      "loss": 1.5709,
      "step": 1650
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.448581637870135e-05,
      "loss": 1.6062,
      "step": 1652
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.4472758261556639e-05,
      "loss": 1.5866,
      "step": 1654
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.4459690603403623e-05,
      "loss": 1.603,
      "step": 1656
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.444661343211741e-05,
      "loss": 1.6153,
      "step": 1658
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.4433526775593404e-05,
      "loss": 1.563,
      "step": 1660
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.4420430661747245e-05,
      "loss": 1.6109,
      "step": 1662
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.4407325118514743e-05,
      "loss": 1.5876,
      "step": 1664
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.4394210173851824e-05,
      "loss": 1.6287,
      "step": 1666
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.4381085855734468e-05,
      "loss": 1.5973,
      "step": 1668
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4367952192158652e-05,
      "loss": 1.6117,
      "step": 1670
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4354809211140284e-05,
      "loss": 1.559,
      "step": 1672
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4341656940715147e-05,
      "loss": 1.5539,
      "step": 1674
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4328495408938847e-05,
      "loss": 1.5392,
      "step": 1676
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4315324643886737e-05,
      "loss": 1.5979,
      "step": 1678
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.4302144673653875e-05,
      "loss": 1.604,
      "step": 1680
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.4288955526354942e-05,
      "loss": 1.5481,
      "step": 1682
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.4275757230124207e-05,
      "loss": 1.6369,
      "step": 1684
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.426254981311545e-05,
      "loss": 1.6026,
      "step": 1686
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.4249333303501906e-05,
      "loss": 1.5704,
      "step": 1688
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.4236107729476209e-05,
      "loss": 1.6457,
      "step": 1690
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.4222873119250325e-05,
      "loss": 1.5152,
      "step": 1692
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.4209629501055497e-05,
      "loss": 1.5963,
      "step": 1694
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.4196376903142186e-05,
      "loss": 1.5865,
      "step": 1696
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.4183115353780001e-05,
      "loss": 1.6366,
      "step": 1698
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.4169844881257655e-05,
      "loss": 1.5734,
      "step": 1700
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.415656551388289e-05,
      "loss": 1.562,
      "step": 1702
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.4143277279982415e-05,
      "loss": 1.5946,
      "step": 1704
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.4129980207901867e-05,
      "loss": 1.5339,
      "step": 1706
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.4116674326005731e-05,
      "loss": 1.6555,
      "step": 1708
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.4103359662677276e-05,
      "loss": 1.6325,
      "step": 1710
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.4090036246318513e-05,
      "loss": 1.5931,
      "step": 1712
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.4076704105350117e-05,
      "loss": 1.5118,
      "step": 1714
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.406336326821138e-05,
      "loss": 1.6545,
      "step": 1716
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.4050013763360144e-05,
      "loss": 1.6025,
      "step": 1718
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.4036655619272736e-05,
      "loss": 1.5861,
      "step": 1720
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.4023288864443915e-05,
      "loss": 1.604,
      "step": 1722
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.4009913527386808e-05,
      "loss": 1.5718,
      "step": 1724
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.3996529636632843e-05,
      "loss": 1.675,
      "step": 1726
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.3983137220731702e-05,
      "loss": 1.5882,
      "step": 1728
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.3969736308251252e-05,
      "loss": 1.5851,
      "step": 1730
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.3956326927777478e-05,
      "loss": 1.5425,
      "step": 1732
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.3942909107914431e-05,
      "loss": 1.5696,
      "step": 1734
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.3929482877284168e-05,
      "loss": 1.6312,
      "step": 1736
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.3916048264526687e-05,
      "loss": 1.5701,
      "step": 1738
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.390260529829986e-05,
      "loss": 1.5844,
      "step": 1740
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.3889154007279384e-05,
      "loss": 1.5455,
      "step": 1742
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.3875694420158712e-05,
      "loss": 1.6109,
      "step": 1744
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.3862226565648996e-05,
      "loss": 1.5996,
      "step": 1746
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.3848750472479013e-05,
      "loss": 1.6145,
      "step": 1748
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.3835266169395132e-05,
      "loss": 1.5886,
      "step": 1750
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.3821773685161224e-05,
      "loss": 1.5617,
      "step": 1752
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.3808273048558609e-05,
      "loss": 1.5521,
      "step": 1754
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.3794764288386003e-05,
      "loss": 1.6455,
      "step": 1756
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3781247433459447e-05,
      "loss": 1.6174,
      "step": 1758
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3767722512612251e-05,
      "loss": 1.617,
      "step": 1760
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3754189554694934e-05,
      "loss": 1.5477,
      "step": 1762
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3740648588575156e-05,
      "loss": 1.6121,
      "step": 1764
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3727099643137653e-05,
      "loss": 1.5544,
      "step": 1766
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.3713542747284194e-05,
      "loss": 1.5416,
      "step": 1768
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.3699977929933503e-05,
      "loss": 1.6363,
      "step": 1770
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.3686405220021194e-05,
      "loss": 1.6219,
      "step": 1772
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.3672824646499731e-05,
      "loss": 1.5882,
      "step": 1774
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.3659236238338339e-05,
      "loss": 1.7416,
      "step": 1776
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.3645640024522959e-05,
      "loss": 1.6427,
      "step": 1778
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.363203603405619e-05,
      "loss": 1.5825,
      "step": 1780
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.361842429595721e-05,
      "loss": 1.621,
      "step": 1782
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.3604804839261732e-05,
      "loss": 1.5753,
      "step": 1784
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.3591177693021927e-05,
      "loss": 1.6405,
      "step": 1786
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.3577542886306367e-05,
      "loss": 1.5353,
      "step": 1788
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.3563900448199977e-05,
      "loss": 1.598,
      "step": 1790
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3550250407803951e-05,
      "loss": 1.6229,
      "step": 1792
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3536592794235696e-05,
      "loss": 1.5628,
      "step": 1794
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3522927636628787e-05,
      "loss": 1.5696,
      "step": 1796
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3509254964132883e-05,
      "loss": 1.6368,
      "step": 1798
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3495574805913669e-05,
      "loss": 1.5796,
      "step": 1800
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.3481887191152807e-05,
      "loss": 1.5293,
      "step": 1802
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.346819214904786e-05,
      "loss": 1.57,
      "step": 1804
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.3454489708812237e-05,
      "loss": 1.5036,
      "step": 1806
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.3440779899675128e-05,
      "loss": 1.609,
      "step": 1808
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.3427062750881435e-05,
      "loss": 1.5554,
      "step": 1810
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.3413338291691726e-05,
      "loss": 1.5686,
      "step": 1812
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.3399606551382164e-05,
      "loss": 1.544,
      "step": 1814
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.3385867559244435e-05,
      "loss": 1.6573,
      "step": 1816
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.3372121344585694e-05,
      "loss": 1.591,
      "step": 1818
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.335836793672851e-05,
      "loss": 1.577,
      "step": 1820
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.3344607365010795e-05,
      "loss": 1.5879,
      "step": 1822
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.3330839658785739e-05,
      "loss": 1.5377,
      "step": 1824
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.331706484742175e-05,
      "loss": 1.5187,
      "step": 1826
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3303282960302397e-05,
      "loss": 1.5582,
      "step": 1828
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3289494026826337e-05,
      "loss": 1.6114,
      "step": 1830
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3275698076407268e-05,
      "loss": 1.5021,
      "step": 1832
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3261895138473841e-05,
      "loss": 1.6179,
      "step": 1834
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3248085242469629e-05,
      "loss": 1.5231,
      "step": 1836
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3234268417853027e-05,
      "loss": 1.564,
      "step": 1838
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3220444694097227e-05,
      "loss": 1.577,
      "step": 1840
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3206614100690139e-05,
      "loss": 1.6216,
      "step": 1842
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3192776667134311e-05,
      "loss": 1.5821,
      "step": 1844
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3178932422946892e-05,
      "loss": 1.5774,
      "step": 1846
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3165081397659563e-05,
      "loss": 1.642,
      "step": 1848
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3151223620818463e-05,
      "loss": 1.526,
      "step": 1850
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3137359121984134e-05,
      "loss": 1.5949,
      "step": 1852
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3123487930731464e-05,
      "loss": 1.5635,
      "step": 1854
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3109610076649602e-05,
      "loss": 1.5585,
      "step": 1856
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3095725589341924e-05,
      "loss": 1.6193,
      "step": 1858
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3081834498425952e-05,
      "loss": 1.5287,
      "step": 1860
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3067936833533286e-05,
      "loss": 1.5954,
      "step": 1862
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3054032624309566e-05,
      "loss": 1.5767,
      "step": 1864
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3040121900414371e-05,
      "loss": 1.6328,
      "step": 1866
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3026204691521193e-05,
      "loss": 1.5769,
      "step": 1868
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.3012281027317353e-05,
      "loss": 1.611,
      "step": 1870
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2998350937503939e-05,
      "loss": 1.5665,
      "step": 1872
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2984414451795747e-05,
      "loss": 1.518,
      "step": 1874
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2970471599921222e-05,
      "loss": 1.5733,
      "step": 1876
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2956522411622377e-05,
      "loss": 1.6023,
      "step": 1878
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.2942566916654753e-05,
      "loss": 1.5534,
      "step": 1880
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.292860514478734e-05,
      "loss": 1.6051,
      "step": 1882
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.2914637125802514e-05,
      "loss": 1.6093,
      "step": 1884
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.290066288949598e-05,
      "loss": 1.5965,
      "step": 1886
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.2886682465676707e-05,
      "loss": 1.5038,
      "step": 1888
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.287269588416686e-05,
      "loss": 1.5615,
      "step": 1890
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2858703174801743e-05,
      "loss": 1.5986,
      "step": 1892
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2844704367429723e-05,
      "loss": 1.5638,
      "step": 1894
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2830699491912186e-05,
      "loss": 1.5111,
      "step": 1896
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.2816688578123459e-05,
      "loss": 1.6758,
      "step": 1898
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.280267165595074e-05,
      "loss": 1.6036,
      "step": 1900
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2788648755294056e-05,
      "loss": 1.5597,
      "step": 1902
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2774619906066178e-05,
      "loss": 1.5505,
      "step": 1904
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2760585138192577e-05,
      "loss": 1.6206,
      "step": 1906
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2746544481611336e-05,
      "loss": 1.5424,
      "step": 1908
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.273249796627311e-05,
      "loss": 1.583,
      "step": 1910
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.2718445622141044e-05,
      "loss": 1.6438,
      "step": 1912
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.270438747919072e-05,
      "loss": 1.5231,
      "step": 1914
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.2690323567410094e-05,
      "loss": 1.6052,
      "step": 1916
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.2676253916799421e-05,
      "loss": 1.6286,
      "step": 1918
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.2662178557371198e-05,
      "loss": 1.5095,
      "step": 1920
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.2648097519150107e-05,
      "loss": 1.5288,
      "step": 1922
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.2634010832172932e-05,
      "loss": 1.5961,
      "step": 1924
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.261991852648852e-05,
      "loss": 1.4945,
      "step": 1926
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.2605820632157689e-05,
      "loss": 1.5989,
      "step": 1928
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.2591717179253188e-05,
      "loss": 1.6103,
      "step": 1930
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.2577608197859627e-05,
      "loss": 1.5667,
      "step": 1932
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.25634937180734e-05,
      "loss": 1.5074,
      "step": 1934
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.2549373770002626e-05,
      "loss": 1.6327,
      "step": 1936
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.2535248383767102e-05,
      "loss": 1.5997,
      "step": 1938
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.252111758949822e-05,
      "loss": 1.5462,
      "step": 1940
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.2506981417338903e-05,
      "loss": 1.5862,
      "step": 1942
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.2492839897443554e-05,
      "loss": 1.618,
      "step": 1944
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.2478693059977975e-05,
      "loss": 1.6031,
      "step": 1946
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.2464540935119319e-05,
      "loss": 1.6268,
      "step": 1948
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.2450383553056011e-05,
      "loss": 1.637,
      "step": 1950
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.2436220943987694e-05,
      "loss": 1.6018,
      "step": 1952
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.2422053138125164e-05,
      "loss": 1.5885,
      "step": 1954
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.2407880165690289e-05,
      "loss": 1.5386,
      "step": 1956
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2393702056915977e-05,
      "loss": 1.6509,
      "step": 1958
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2379518842046081e-05,
      "loss": 1.5272,
      "step": 1960
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2365330551335348e-05,
      "loss": 1.5918,
      "step": 1962
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2351137215049353e-05,
      "loss": 1.5838,
      "step": 1964
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2336938863464434e-05,
      "loss": 1.5712,
      "step": 1966
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.2322735526867624e-05,
      "loss": 1.5799,
      "step": 1968
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.2308527235556596e-05,
      "loss": 1.555,
      "step": 1970
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.229431401983959e-05,
      "loss": 1.5856,
      "step": 1972
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.2280095910035343e-05,
      "loss": 1.6309,
      "step": 1974
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.2265872936473044e-05,
      "loss": 1.5952,
      "step": 1976
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.2251645129492252e-05,
      "loss": 1.4671,
      "step": 1978
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.2237412519442828e-05,
      "loss": 1.6362,
      "step": 1980
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.222317513668489e-05,
      "loss": 1.6166,
      "step": 1982
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.2208933011588735e-05,
      "loss": 1.4717,
      "step": 1984
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.2194686174534771e-05,
      "loss": 1.5405,
      "step": 1986
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.2180434655913465e-05,
      "loss": 1.5721,
      "step": 1988
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.2166178486125258e-05,
      "loss": 1.5779,
      "step": 1990
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.2151917695580523e-05,
      "loss": 1.6138,
      "step": 1992
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.2137652314699494e-05,
      "loss": 1.5975,
      "step": 1994
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.2123382373912178e-05,
      "loss": 1.4967,
      "step": 1996
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.2109107903658326e-05,
      "loss": 1.5278,
      "step": 1998
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.209482893438735e-05,
      "loss": 1.46,
      "step": 2000
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2080545496558248e-05,
      "loss": 1.5729,
      "step": 2002
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2066257620639557e-05,
      "loss": 1.6127,
      "step": 2004
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2051965337109284e-05,
      "loss": 1.5214,
      "step": 2006
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2037668676454832e-05,
      "loss": 1.5628,
      "step": 2008
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2023367669172947e-05,
      "loss": 1.6562,
      "step": 2010
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.2009062345769645e-05,
      "loss": 1.5651,
      "step": 2012
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.199475273676014e-05,
      "loss": 1.603,
      "step": 2014
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.1980438872668803e-05,
      "loss": 1.5625,
      "step": 2016
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.1966120784029066e-05,
      "loss": 1.5886,
      "step": 2018
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.195179850138339e-05,
      "loss": 1.5668,
      "step": 2020
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.1937472055283168e-05,
      "loss": 1.6338,
      "step": 2022
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.1923141476288676e-05,
      "loss": 1.51,
      "step": 2024
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.1908806794969015e-05,
      "loss": 1.5857,
      "step": 2026
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.189446804190203e-05,
      "loss": 1.5716,
      "step": 2028
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.188012524767425e-05,
      "loss": 1.5771,
      "step": 2030
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.1865778442880828e-05,
      "loss": 1.5471,
      "step": 2032
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.1851427658125474e-05,
      "loss": 1.5763,
      "step": 2034
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.183707292402038e-05,
      "loss": 1.5639,
      "step": 2036
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.1822714271186173e-05,
      "loss": 1.5648,
      "step": 2038
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.180835173025183e-05,
      "loss": 1.5097,
      "step": 2040
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.1793985331854622e-05,
      "loss": 1.5358,
      "step": 2042
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.1779615106640058e-05,
      "loss": 1.6405,
      "step": 2044
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.1765241085261802e-05,
      "loss": 1.5903,
      "step": 2046
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.1750863298381618e-05,
      "loss": 1.6238,
      "step": 2048
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.1736481776669307e-05,
      "loss": 1.5676,
      "step": 2050
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.172209655080262e-05,
      "loss": 1.5194,
      "step": 2052
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.170770765146723e-05,
      "loss": 1.5859,
      "step": 2054
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.1693315109356637e-05,
      "loss": 1.4639,
      "step": 2056
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.1678918955172112e-05,
      "loss": 1.5708,
      "step": 2058
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.1664519219622628e-05,
      "loss": 1.6653,
      "step": 2060
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.1650115933424804e-05,
      "loss": 1.633,
      "step": 2062
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.163570912730283e-05,
      "loss": 1.588,
      "step": 2064
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.1621298831988398e-05,
      "loss": 1.5914,
      "step": 2066
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.1606885078220658e-05,
      "loss": 1.5181,
      "step": 2068
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.1592467896746122e-05,
      "loss": 1.4927,
      "step": 2070
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.1578047318318624e-05,
      "loss": 1.5709,
      "step": 2072
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.156362337369924e-05,
      "loss": 1.5502,
      "step": 2074
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.1549196093656223e-05,
      "loss": 1.5939,
      "step": 2076
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.1534765508964952e-05,
      "loss": 1.5504,
      "step": 2078
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.1520331650407839e-05,
      "loss": 1.5904,
      "step": 2080
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.1505894548774294e-05,
      "loss": 1.5527,
      "step": 2082
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.1491454234860645e-05,
      "loss": 1.5581,
      "step": 2084
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.1477010739470056e-05,
      "loss": 1.5685,
      "step": 2086
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.1462564093412493e-05,
      "loss": 1.608,
      "step": 2088
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.144811432750464e-05,
      "loss": 1.6009,
      "step": 2090
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1433661472569832e-05,
      "loss": 1.5341,
      "step": 2092
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1419205559437998e-05,
      "loss": 1.5827,
      "step": 2094
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1404746618945588e-05,
      "loss": 1.576,
      "step": 2096
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1390284681935509e-05,
      "loss": 1.5971,
      "step": 2098
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1375819779257058e-05,
      "loss": 1.5065,
      "step": 2100
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1361351941765866e-05,
      "loss": 1.6612,
      "step": 2102
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1346881200323817e-05,
      "loss": 1.5207,
      "step": 2104
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1332407585798992e-05,
      "loss": 1.6724,
      "step": 2106
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1317931129065602e-05,
      "loss": 1.4631,
      "step": 2108
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1303451861003918e-05,
      "loss": 1.5509,
      "step": 2110
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1288969812500209e-05,
      "loss": 1.5472,
      "step": 2112
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1274485014446675e-05,
      "loss": 1.6121,
      "step": 2114
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1259997497741385e-05,
      "loss": 1.5248,
      "step": 2116
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1245507293288204e-05,
      "loss": 1.5002,
      "step": 2118
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1231014431996727e-05,
      "loss": 1.6081,
      "step": 2120
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1216518944782219e-05,
      "loss": 1.5393,
      "step": 2122
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1202020862565555e-05,
      "loss": 1.576,
      "step": 2124
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1187520216273126e-05,
      "loss": 1.5501,
      "step": 2126
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.117301703683681e-05,
      "loss": 1.5741,
      "step": 2128
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1158511355193888e-05,
      "loss": 1.5463,
      "step": 2130
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1144003202286964e-05,
      "loss": 1.6043,
      "step": 2132
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1129492609063927e-05,
      "loss": 1.5227,
      "step": 2134
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1114979606477867e-05,
      "loss": 1.572,
      "step": 2136
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1100464225487013e-05,
      "loss": 1.5958,
      "step": 2138
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1085946497054674e-05,
      "loss": 1.5177,
      "step": 2140
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1071426452149152e-05,
      "loss": 1.5828,
      "step": 2142
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1056904121743702e-05,
      "loss": 1.5872,
      "step": 2144
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1042379536816457e-05,
      "loss": 1.6537,
      "step": 2146
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1027852728350343e-05,
      "loss": 1.5949,
      "step": 2148
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.101332372733305e-05,
      "loss": 1.6051,
      "step": 2150
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.0998792564756927e-05,
      "loss": 1.5928,
      "step": 2152
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.0984259271618947e-05,
      "loss": 1.5785,
      "step": 2154
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.096972387892062e-05,
      "loss": 1.5114,
      "step": 2156
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.0955186417667936e-05,
      "loss": 1.5796,
      "step": 2158
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.09406469188713e-05,
      "loss": 1.6061,
      "step": 2160
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.0926105413545463e-05,
      "loss": 1.5535,
      "step": 2162
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.0911561932709453e-05,
      "loss": 1.5657,
      "step": 2164
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.0897016507386513e-05,
      "loss": 1.5686,
      "step": 2166
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.0882469168604042e-05,
      "loss": 1.5352,
      "step": 2168
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0867919947393504e-05,
      "loss": 1.5629,
      "step": 2170
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0853368874790392e-05,
      "loss": 1.5478,
      "step": 2172
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.083881598183415e-05,
      "loss": 1.6021,
      "step": 2174
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.082426129956809e-05,
      "loss": 1.5736,
      "step": 2176
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0809704859039357e-05,
      "loss": 1.5961,
      "step": 2178
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0795146691298835e-05,
      "loss": 1.5399,
      "step": 2180
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.07805868274011e-05,
      "loss": 1.5734,
      "step": 2182
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0766025298404346e-05,
      "loss": 1.5648,
      "step": 2184
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0751462135370313e-05,
      "loss": 1.5733,
      "step": 2186
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0736897369364232e-05,
      "loss": 1.6152,
      "step": 2188
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0722331031454749e-05,
      "loss": 1.6522,
      "step": 2190
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0707763152713869e-05,
      "loss": 1.5796,
      "step": 2192
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0693193764216875e-05,
      "loss": 1.4848,
      "step": 2194
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0678622897042279e-05,
      "loss": 1.4827,
      "step": 2196
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0664050582271744e-05,
      "loss": 1.5397,
      "step": 2198
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0649476850990019e-05,
      "loss": 1.6373,
      "step": 2200
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.063490173428488e-05,
      "loss": 1.5799,
      "step": 2202
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.062032526324705e-05,
      "loss": 1.5911,
      "step": 2204
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.0605747468970148e-05,
      "loss": 1.6362,
      "step": 2206
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.0591168382550617e-05,
      "loss": 1.5298,
      "step": 2208
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.0576588035087647e-05,
      "loss": 1.5764,
      "step": 2210
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.0562006457683128e-05,
      "loss": 1.5423,
      "step": 2212
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.0547423681441567e-05,
      "loss": 1.5802,
      "step": 2214
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.053283973747003e-05,
      "loss": 1.5937,
      "step": 2216
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.0518254656878075e-05,
      "loss": 1.5666,
      "step": 2218
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.050366847077769e-05,
      "loss": 1.5285,
      "step": 2220
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.0489081210283206e-05,
      "loss": 1.4526,
      "step": 2222
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0474492906511258e-05,
      "loss": 1.5371,
      "step": 2224
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0459903590580706e-05,
      "loss": 1.5546,
      "step": 2226
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0445313293612567e-05,
      "loss": 1.5165,
      "step": 2228
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.043072204672995e-05,
      "loss": 1.5631,
      "step": 2230
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0416129881057987e-05,
      "loss": 1.5341,
      "step": 2232
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0401536827723778e-05,
      "loss": 1.5122,
      "step": 2234
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0386942917856312e-05,
      "loss": 1.5344,
      "step": 2236
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.03723481825864e-05,
      "loss": 1.5843,
      "step": 2238
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0357752653046625e-05,
      "loss": 1.5024,
      "step": 2240
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0343156360371256e-05,
      "loss": 1.4637,
      "step": 2242
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0328559335696188e-05,
      "loss": 1.5855,
      "step": 2244
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0313961610158887e-05,
      "loss": 1.5466,
      "step": 2246
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0299363214898305e-05,
      "loss": 1.5501,
      "step": 2248
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.028476418105483e-05,
      "loss": 1.5245,
      "step": 2250
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0270164539770206e-05,
      "loss": 1.4956,
      "step": 2252
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0255564322187476e-05,
      "loss": 1.524,
      "step": 2254
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0240963559450909e-05,
      "loss": 1.5229,
      "step": 2256
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0226362282705942e-05,
      "loss": 1.4596,
      "step": 2258
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0211760523099107e-05,
      "loss": 1.5043,
      "step": 2260
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0197158311777957e-05,
      "loss": 1.5721,
      "step": 2262
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0182555679891026e-05,
      "loss": 1.4634,
      "step": 2264
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0167952658587729e-05,
      "loss": 1.5929,
      "step": 2266
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.015334927901832e-05,
      "loss": 1.5438,
      "step": 2268
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0138745572333816e-05,
      "loss": 1.5338,
      "step": 2270
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0124141569685933e-05,
      "loss": 1.5582,
      "step": 2272
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0109537302227012e-05,
      "loss": 1.558,
      "step": 2274
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0094932801109968e-05,
      "loss": 1.505,
      "step": 2276
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0080328097488206e-05,
      "loss": 1.5063,
      "step": 2278
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0065723222515566e-05,
      "loss": 1.5605,
      "step": 2280
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0051118207346263e-05,
      "loss": 1.5184,
      "step": 2282
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.003651308313479e-05,
      "loss": 1.5838,
      "step": 2284
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0021907881035891e-05,
      "loss": 1.5031,
      "step": 2286
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0007302632204472e-05,
      "loss": 1.5279,
      "step": 2288
    },
    {
      "epoch": 2.06,
      "learning_rate": 9.992697367795531e-06,
      "loss": 1.5222,
      "step": 2290
    },
    {
      "epoch": 2.07,
      "learning_rate": 9.97809211896411e-06,
      "loss": 1.4966,
      "step": 2292
    },
    {
      "epoch": 2.07,
      "learning_rate": 9.963486916865212e-06,
      "loss": 1.4892,
      "step": 2294
    },
    {
      "epoch": 2.07,
      "learning_rate": 9.948881792653744e-06,
      "loss": 1.48,
      "step": 2296
    },
    {
      "epoch": 2.07,
      "learning_rate": 9.934276777484436e-06,
      "loss": 1.4611,
      "step": 2298
    },
    {
      "epoch": 2.07,
      "learning_rate": 9.919671902511798e-06,
      "loss": 1.5598,
      "step": 2300
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.905067198890035e-06,
      "loss": 1.4727,
      "step": 2302
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.89046269777299e-06,
      "loss": 1.5649,
      "step": 2304
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.875858430314072e-06,
      "loss": 1.4784,
      "step": 2306
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.861254427666187e-06,
      "loss": 1.5344,
      "step": 2308
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.846650720981682e-06,
      "loss": 1.5672,
      "step": 2310
    },
    {
      "epoch": 2.08,
      "learning_rate": 9.832047341412273e-06,
      "loss": 1.4945,
      "step": 2312
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.817444320108976e-06,
      "loss": 1.5149,
      "step": 2314
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.802841688222043e-06,
      "loss": 1.4418,
      "step": 2316
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.788239476900899e-06,
      "loss": 1.5727,
      "step": 2318
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.773637717294061e-06,
      "loss": 1.4015,
      "step": 2320
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.759036440549093e-06,
      "loss": 1.5113,
      "step": 2322
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.744435677812526e-06,
      "loss": 1.5221,
      "step": 2324
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.729835460229796e-06,
      "loss": 1.5461,
      "step": 2326
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.715235818945171e-06,
      "loss": 1.4714,
      "step": 2328
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.700636785101696e-06,
      "loss": 1.5814,
      "step": 2330
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.686038389841116e-06,
      "loss": 1.5544,
      "step": 2332
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.671440664303813e-06,
      "loss": 1.5014,
      "step": 2334
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.656843639628748e-06,
      "loss": 1.5115,
      "step": 2336
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.642247346953376e-06,
      "loss": 1.5332,
      "step": 2338
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.627651817413605e-06,
      "loss": 1.4812,
      "step": 2340
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.613057082143694e-06,
      "loss": 1.4951,
      "step": 2342
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.598463172276224e-06,
      "loss": 1.4513,
      "step": 2344
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.583870118942014e-06,
      "loss": 1.5211,
      "step": 2346
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.576573923202763e-06,
      "loss": 1.5283,
      "step": 2348
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.561982213034707e-06,
      "loss": 1.5975,
      "step": 2350
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.547391437218815e-06,
      "loss": 1.4903,
      "step": 2352
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.532801626879209e-06,
      "loss": 1.537,
      "step": 2354
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.518212813137966e-06,
      "loss": 1.4725,
      "step": 2356
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.50362502711503e-06,
      "loss": 1.4511,
      "step": 2358
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.489038299928157e-06,
      "loss": 1.5214,
      "step": 2360
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.47445266269284e-06,
      "loss": 1.5027,
      "step": 2362
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.45986814652225e-06,
      "loss": 1.5368,
      "step": 2364
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.445284782527157e-06,
      "loss": 1.5637,
      "step": 2366
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.430702601815892e-06,
      "loss": 1.4993,
      "step": 2368
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.416121635494245e-06,
      "loss": 1.5066,
      "step": 2370
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.401541914665425e-06,
      "loss": 1.4701,
      "step": 2372
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.38696347042998e-06,
      "loss": 1.546,
      "step": 2374
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.372386333885738e-06,
      "loss": 1.5266,
      "step": 2376
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.35781053612773e-06,
      "loss": 1.5479,
      "step": 2378
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.34323610824814e-06,
      "loss": 1.596,
      "step": 2380
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.328663081336225e-06,
      "loss": 1.5529,
      "step": 2382
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.314091486478255e-06,
      "loss": 1.4637,
      "step": 2384
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.29952135475745e-06,
      "loss": 1.5238,
      "step": 2386
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.284952717253898e-06,
      "loss": 1.5229,
      "step": 2388
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.270385605044509e-06,
      "loss": 1.5199,
      "step": 2390
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.255820049202933e-06,
      "loss": 1.5131,
      "step": 2392
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.241256080799507e-06,
      "loss": 1.5487,
      "step": 2394
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.226693730901174e-06,
      "loss": 1.6053,
      "step": 2396
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.212133030571438e-06,
      "loss": 1.5352,
      "step": 2398
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.19757401087026e-06,
      "loss": 1.5733,
      "step": 2400
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.18301670285404e-06,
      "loss": 1.4881,
      "step": 2402
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.168461137575515e-06,
      "loss": 1.4606,
      "step": 2404
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.153907346083706e-06,
      "loss": 1.4393,
      "step": 2406
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.139355359423855e-06,
      "loss": 1.5519,
      "step": 2408
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.124805208637349e-06,
      "loss": 1.4962,
      "step": 2410
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.110256924761655e-06,
      "loss": 1.4699,
      "step": 2412
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.102983492613489e-06,
      "loss": 1.4995,
      "step": 2414
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.08843806729055e-06,
      "loss": 1.5511,
      "step": 2416
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.073894586454538e-06,
      "loss": 1.6146,
      "step": 2418
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.059353081128702e-06,
      "loss": 1.443,
      "step": 2420
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.044813582332067e-06,
      "loss": 1.4808,
      "step": 2422
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.030276121079384e-06,
      "loss": 1.5527,
      "step": 2424
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.015740728381055e-06,
      "loss": 1.5851,
      "step": 2426
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.001207435243074e-06,
      "loss": 1.4776,
      "step": 2428
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.986676272666952e-06,
      "loss": 1.4741,
      "step": 2430
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.972147271649662e-06,
      "loss": 1.5235,
      "step": 2432
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.95762046318355e-06,
      "loss": 1.5066,
      "step": 2434
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.9430958782563e-06,
      "loss": 1.4489,
      "step": 2436
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.928573547850852e-06,
      "loss": 1.4724,
      "step": 2438
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.91405350294533e-06,
      "loss": 1.5894,
      "step": 2440
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.899535774512986e-06,
      "loss": 1.4172,
      "step": 2442
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.885020393522136e-06,
      "loss": 1.4497,
      "step": 2444
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.870507390936076e-06,
      "loss": 1.5584,
      "step": 2446
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.85599679771304e-06,
      "loss": 1.5338,
      "step": 2448
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.841488644806115e-06,
      "loss": 1.5138,
      "step": 2450
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.826982963163188e-06,
      "loss": 1.529,
      "step": 2452
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.812479783726874e-06,
      "loss": 1.4863,
      "step": 2454
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.797979137434452e-06,
      "loss": 1.5418,
      "step": 2456
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.783481055217783e-06,
      "loss": 1.5622,
      "step": 2458
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.768985568003277e-06,
      "loss": 1.5281,
      "step": 2460
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.754492706711798e-06,
      "loss": 1.5676,
      "step": 2462
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.740002502258616e-06,
      "loss": 1.4468,
      "step": 2464
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.725514985553327e-06,
      "loss": 1.5271,
      "step": 2466
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.711030187499795e-06,
      "loss": 1.5094,
      "step": 2468
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.696548138996086e-06,
      "loss": 1.5152,
      "step": 2470
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.682068870934401e-06,
      "loss": 1.4206,
      "step": 2472
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.66759241420101e-06,
      "loss": 1.4556,
      "step": 2474
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.653118799676183e-06,
      "loss": 1.4849,
      "step": 2476
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.638648058234139e-06,
      "loss": 1.4729,
      "step": 2478
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.624180220742945e-06,
      "loss": 1.4827,
      "step": 2480
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.609715318064496e-06,
      "loss": 1.5816,
      "step": 2482
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.595253381054414e-06,
      "loss": 1.4841,
      "step": 2484
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.580794440562003e-06,
      "loss": 1.4788,
      "step": 2486
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.566338527430168e-06,
      "loss": 1.4628,
      "step": 2488
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.551885672495363e-06,
      "loss": 1.5261,
      "step": 2490
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.53743590658751e-06,
      "loss": 1.5421,
      "step": 2492
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.522989260529948e-06,
      "loss": 1.5478,
      "step": 2494
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.508545765139359e-06,
      "loss": 1.5603,
      "step": 2496
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.494105451225706e-06,
      "loss": 1.4666,
      "step": 2498
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.479668349592165e-06,
      "loss": 1.5813,
      "step": 2500
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.465234491035053e-06,
      "loss": 1.5637,
      "step": 2502
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.45080390634378e-06,
      "loss": 1.5104,
      "step": 2504
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.436376626300764e-06,
      "loss": 1.553,
      "step": 2506
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.421952681681378e-06,
      "loss": 1.5615,
      "step": 2508
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.407532103253878e-06,
      "loss": 1.5202,
      "step": 2510
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.393114921779345e-06,
      "loss": 1.5754,
      "step": 2512
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.378701168011603e-06,
      "loss": 1.5279,
      "step": 2514
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.364290872697175e-06,
      "loss": 1.491,
      "step": 2516
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.349884066575198e-06,
      "loss": 1.5686,
      "step": 2518
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.335480780377374e-06,
      "loss": 1.4939,
      "step": 2520
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.321081044827894e-06,
      "loss": 1.4625,
      "step": 2522
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.306684890643368e-06,
      "loss": 1.4761,
      "step": 2524
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.292292348532774e-06,
      "loss": 1.5043,
      "step": 2526
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.277903449197383e-06,
      "loss": 1.4946,
      "step": 2528
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.263518223330698e-06,
      "loss": 1.5154,
      "step": 2530
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.24913670161838e-06,
      "loss": 1.4999,
      "step": 2532
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.2347589147382e-06,
      "loss": 1.5577,
      "step": 2534
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.220384893359944e-06,
      "loss": 1.5162,
      "step": 2536
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.20601466814538e-06,
      "loss": 1.5193,
      "step": 2538
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.191648269748173e-06,
      "loss": 1.5696,
      "step": 2540
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.177285728813829e-06,
      "loss": 1.5841,
      "step": 2542
    },
    {
      "epoch": 2.29,
      "learning_rate": 8.16292707597962e-06,
      "loss": 1.5085,
      "step": 2544
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.14857234187453e-06,
      "loss": 1.5908,
      "step": 2546
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.134221557119175e-06,
      "loss": 1.5277,
      "step": 2548
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.119874752325753e-06,
      "loss": 1.452,
      "step": 2550
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.105531958097973e-06,
      "loss": 1.5027,
      "step": 2552
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.091193205030987e-06,
      "loss": 1.5185,
      "step": 2554
    },
    {
      "epoch": 2.3,
      "learning_rate": 8.076858523711327e-06,
      "loss": 1.4928,
      "step": 2556
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.062527944716837e-06,
      "loss": 1.5301,
      "step": 2558
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.048201498616613e-06,
      "loss": 1.577,
      "step": 2560
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.033879215970935e-06,
      "loss": 1.4957,
      "step": 2562
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.019561127331202e-06,
      "loss": 1.4609,
      "step": 2564
    },
    {
      "epoch": 2.31,
      "learning_rate": 8.005247263239862e-06,
      "loss": 1.6149,
      "step": 2566
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.990937654230362e-06,
      "loss": 1.5086,
      "step": 2568
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.976632330827056e-06,
      "loss": 1.4795,
      "step": 2570
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.962331323545171e-06,
      "loss": 1.5367,
      "step": 2572
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.94803466289072e-06,
      "loss": 1.4965,
      "step": 2574
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.933742379360446e-06,
      "loss": 1.5235,
      "step": 2576
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.919454503441757e-06,
      "loss": 1.4606,
      "step": 2578
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.905171065612655e-06,
      "loss": 1.569,
      "step": 2580
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.890892096341677e-06,
      "loss": 1.5586,
      "step": 2582
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.876617626087825e-06,
      "loss": 1.516,
      "step": 2584
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.862347685300511e-06,
      "loss": 1.4995,
      "step": 2586
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.848082304419478e-06,
      "loss": 1.5585,
      "step": 2588
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.833821513874747e-06,
      "loss": 1.534,
      "step": 2590
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.819565344086541e-06,
      "loss": 1.5516,
      "step": 2592
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.805313825465232e-06,
      "loss": 1.5351,
      "step": 2594
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.791066988411268e-06,
      "loss": 1.565,
      "step": 2596
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.776824863315111e-06,
      "loss": 1.5207,
      "step": 2598
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.762587480557175e-06,
      "loss": 1.4396,
      "step": 2600
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.748354870507753e-06,
      "loss": 1.449,
      "step": 2602
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.734127063526958e-06,
      "loss": 1.4943,
      "step": 2604
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.719904089964658e-06,
      "loss": 1.5601,
      "step": 2606
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.705685980160413e-06,
      "loss": 1.5115,
      "step": 2608
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.691472764443404e-06,
      "loss": 1.5607,
      "step": 2610
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.67726447313238e-06,
      "loss": 1.4841,
      "step": 2612
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.66306113653557e-06,
      "loss": 1.4853,
      "step": 2614
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.64886278495065e-06,
      "loss": 1.5695,
      "step": 2616
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.634669448664655e-06,
      "loss": 1.5634,
      "step": 2618
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.62048115795392e-06,
      "loss": 1.5444,
      "step": 2620
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.606297943084024e-06,
      "loss": 1.5306,
      "step": 2622
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.5921198343097145e-06,
      "loss": 1.5062,
      "step": 2624
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.577946861874843e-06,
      "loss": 1.562,
      "step": 2626
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.563779056012309e-06,
      "loss": 1.4988,
      "step": 2628
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.549616446943992e-06,
      "loss": 1.437,
      "step": 2630
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.535459064880684e-06,
      "loss": 1.4982,
      "step": 2632
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.521306940022025e-06,
      "loss": 1.5545,
      "step": 2634
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.507160102556451e-06,
      "loss": 1.5476,
      "step": 2636
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.4930185826611e-06,
      "loss": 1.4544,
      "step": 2638
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.478882410501784e-06,
      "loss": 1.5477,
      "step": 2640
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.464751616232902e-06,
      "loss": 1.5119,
      "step": 2642
    },
    {
      "epoch": 2.38,
      "learning_rate": 7.4506262299973775e-06,
      "loss": 1.5679,
      "step": 2644
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.4365062819266076e-06,
      "loss": 1.5516,
      "step": 2646
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.422391802140376e-06,
      "loss": 1.5458,
      "step": 2648
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.408282820746813e-06,
      "loss": 1.6268,
      "step": 2650
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.394179367842314e-06,
      "loss": 1.5683,
      "step": 2652
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.380081473511484e-06,
      "loss": 1.5416,
      "step": 2654
    },
    {
      "epoch": 2.39,
      "learning_rate": 7.365989167827068e-06,
      "loss": 1.5423,
      "step": 2656
    },
    {
      "epoch": 2.4,
      "learning_rate": 7.351902480849899e-06,
      "loss": 1.5289,
      "step": 2658
    },
    {
      "epoch": 2.4,
      "learning_rate": 7.337821442628805e-06,
      "loss": 1.4968,
      "step": 2660
    },
    {
      "epoch": 2.4,
      "learning_rate": 7.323746083200583e-06,
      "loss": 1.5061,
      "step": 2662
    },
    {
      "epoch": 2.4,
      "learning_rate": 7.309676432589908e-06,
      "loss": 1.5273,
      "step": 2664
    },
    {
      "epoch": 2.4,
      "learning_rate": 7.295612520809281e-06,
      "loss": 1.5042,
      "step": 2666
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.281554377858959e-06,
      "loss": 1.4732,
      "step": 2668
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.267502033726895e-06,
      "loss": 1.5093,
      "step": 2670
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.253455518388668e-06,
      "loss": 1.5265,
      "step": 2672
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.239414861807427e-06,
      "loss": 1.5804,
      "step": 2674
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.2253800939338225e-06,
      "loss": 1.5111,
      "step": 2676
    },
    {
      "epoch": 2.41,
      "learning_rate": 7.211351244705947e-06,
      "loss": 1.5801,
      "step": 2678
    },
    {
      "epoch": 2.42,
      "learning_rate": 7.197328344049266e-06,
      "loss": 1.4756,
      "step": 2680
    },
    {
      "epoch": 2.42,
      "learning_rate": 7.183311421876546e-06,
      "loss": 1.5308,
      "step": 2682
    },
    {
      "epoch": 2.42,
      "learning_rate": 7.169300508087815e-06,
      "loss": 1.5217,
      "step": 2684
    },
    {
      "epoch": 2.42,
      "learning_rate": 7.155295632570279e-06,
      "loss": 1.5401,
      "step": 2686
    },
    {
      "epoch": 2.42,
      "learning_rate": 7.14129682519826e-06,
      "loss": 1.4986,
      "step": 2688
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.127304115833141e-06,
      "loss": 1.5343,
      "step": 2690
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.113317534323296e-06,
      "loss": 1.5726,
      "step": 2692
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.099337110504023e-06,
      "loss": 1.4744,
      "step": 2694
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.08536287419749e-06,
      "loss": 1.4813,
      "step": 2696
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.071394855212663e-06,
      "loss": 1.5123,
      "step": 2698
    },
    {
      "epoch": 2.43,
      "learning_rate": 7.0574330833452475e-06,
      "loss": 1.5364,
      "step": 2700
    },
    {
      "epoch": 2.44,
      "learning_rate": 7.043477588377623e-06,
      "loss": 1.5404,
      "step": 2702
    },
    {
      "epoch": 2.44,
      "learning_rate": 7.029528400078784e-06,
      "loss": 1.5873,
      "step": 2704
    },
    {
      "epoch": 2.44,
      "learning_rate": 7.015585548204256e-06,
      "loss": 1.5007,
      "step": 2706
    },
    {
      "epoch": 2.44,
      "learning_rate": 7.001649062496065e-06,
      "loss": 1.585,
      "step": 2708
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.987718972682651e-06,
      "loss": 1.5347,
      "step": 2710
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.97379530847881e-06,
      "loss": 1.5023,
      "step": 2712
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.959878099585634e-06,
      "loss": 1.5362,
      "step": 2714
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.94596737569044e-06,
      "loss": 1.5489,
      "step": 2716
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.932063166466716e-06,
      "loss": 1.4816,
      "step": 2718
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.918165501574051e-06,
      "loss": 1.513,
      "step": 2720
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.904274410658076e-06,
      "loss": 1.5079,
      "step": 2722
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.890389923350398e-06,
      "loss": 1.5264,
      "step": 2724
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.876512069268541e-06,
      "loss": 1.4754,
      "step": 2726
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.862640878015867e-06,
      "loss": 1.4453,
      "step": 2728
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.84877637918154e-06,
      "loss": 1.5165,
      "step": 2730
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.834918602340439e-06,
      "loss": 1.6026,
      "step": 2732
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.82106757705311e-06,
      "loss": 1.5246,
      "step": 2734
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.807223332865696e-06,
      "loss": 1.5018,
      "step": 2736
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.793385899309866e-06,
      "loss": 1.4946,
      "step": 2738
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.779555305902774e-06,
      "loss": 1.562,
      "step": 2740
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.765731582146977e-06,
      "loss": 1.526,
      "step": 2742
    },
    {
      "epoch": 2.47,
      "learning_rate": 6.751914757530375e-06,
      "loss": 1.5335,
      "step": 2744
    },
    {
      "epoch": 2.48,
      "learning_rate": 6.738104861526158e-06,
      "loss": 1.5006,
      "step": 2746
    },
    {
      "epoch": 2.48,
      "learning_rate": 6.7243019235927355e-06,
      "loss": 1.5107,
      "step": 2748
    },
    {
      "epoch": 2.48,
      "learning_rate": 6.7105059731736645e-06,
      "loss": 1.5225,
      "step": 2750
    },
    {
      "epoch": 2.48,
      "learning_rate": 6.6967170396976064e-06,
      "loss": 1.5346,
      "step": 2752
    },
    {
      "epoch": 2.48,
      "learning_rate": 6.682935152578254e-06,
      "loss": 1.6253,
      "step": 2754
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.669160341214265e-06,
      "loss": 1.4834,
      "step": 2756
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.6553926349892074e-06,
      "loss": 1.5052,
      "step": 2758
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.6416320632714936e-06,
      "loss": 1.6093,
      "step": 2760
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.627878655414311e-06,
      "loss": 1.4979,
      "step": 2762
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.614132440755569e-06,
      "loss": 1.5197,
      "step": 2764
    },
    {
      "epoch": 2.49,
      "learning_rate": 6.600393448617838e-06,
      "loss": 1.5165,
      "step": 2766
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.586661708308273e-06,
      "loss": 1.5064,
      "step": 2768
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.5729372491185695e-06,
      "loss": 1.5066,
      "step": 2770
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.559220100324879e-06,
      "loss": 1.4739,
      "step": 2772
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.5455102911877665e-06,
      "loss": 1.4729,
      "step": 2774
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.531807850952143e-06,
      "loss": 1.5633,
      "step": 2776
    },
    {
      "epoch": 2.5,
      "learning_rate": 6.518112808847197e-06,
      "loss": 1.5471,
      "step": 2778
    },
    {
      "epoch": 2.51,
      "learning_rate": 6.504425194086334e-06,
      "loss": 1.4803,
      "step": 2780
    },
    {
      "epoch": 2.51,
      "learning_rate": 6.490745035867123e-06,
      "loss": 1.5166,
      "step": 2782
    },
    {
      "epoch": 2.51,
      "learning_rate": 6.477072363371215e-06,
      "loss": 1.5522,
      "step": 2784
    },
    {
      "epoch": 2.51,
      "learning_rate": 6.4634072057643045e-06,
      "loss": 1.5213,
      "step": 2786
    },
    {
      "epoch": 2.51,
      "learning_rate": 6.449749592196053e-06,
      "loss": 1.5297,
      "step": 2788
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.436099551800023e-06,
      "loss": 1.4882,
      "step": 2790
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.422457113693633e-06,
      "loss": 1.4607,
      "step": 2792
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.40882230697808e-06,
      "loss": 1.5366,
      "step": 2794
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.395195160738273e-06,
      "loss": 1.5072,
      "step": 2796
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.381575704042792e-06,
      "loss": 1.5576,
      "step": 2798
    },
    {
      "epoch": 2.52,
      "learning_rate": 6.367963965943813e-06,
      "loss": 1.5283,
      "step": 2800
    },
    {
      "epoch": 2.53,
      "learning_rate": 6.35435997547704e-06,
      "loss": 1.4778,
      "step": 2802
    },
    {
      "epoch": 2.53,
      "learning_rate": 6.340763761661665e-06,
      "loss": 1.511,
      "step": 2804
    },
    {
      "epoch": 2.53,
      "learning_rate": 6.327175353500272e-06,
      "loss": 1.5468,
      "step": 2806
    },
    {
      "epoch": 2.53,
      "learning_rate": 6.3135947799788065e-06,
      "loss": 1.5025,
      "step": 2808
    },
    {
      "epoch": 2.53,
      "learning_rate": 6.3000220700664985e-06,
      "loss": 1.5675,
      "step": 2810
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.286457252715805e-06,
      "loss": 1.5266,
      "step": 2812
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.272900356862346e-06,
      "loss": 1.4049,
      "step": 2814
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.259351411424849e-06,
      "loss": 1.5185,
      "step": 2816
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.245810445305068e-06,
      "loss": 1.5236,
      "step": 2818
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.232277487387751e-06,
      "loss": 1.513,
      "step": 2820
    },
    {
      "epoch": 2.54,
      "learning_rate": 6.218752566540555e-06,
      "loss": 1.5418,
      "step": 2822
    },
    {
      "epoch": 2.55,
      "learning_rate": 6.205235711614e-06,
      "loss": 1.4801,
      "step": 2824
    },
    {
      "epoch": 2.55,
      "learning_rate": 6.191726951441397e-06,
      "loss": 1.5086,
      "step": 2826
    },
    {
      "epoch": 2.55,
      "learning_rate": 6.17822631483878e-06,
      "loss": 1.503,
      "step": 2828
    },
    {
      "epoch": 2.55,
      "learning_rate": 6.16473383060487e-06,
      "loss": 1.477,
      "step": 2830
    },
    {
      "epoch": 2.55,
      "learning_rate": 6.151249527520988e-06,
      "loss": 1.5222,
      "step": 2832
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.137773434351009e-06,
      "loss": 1.4894,
      "step": 2834
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.12430557984129e-06,
      "loss": 1.5005,
      "step": 2836
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.110845992720622e-06,
      "loss": 1.5306,
      "step": 2838
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.097394701700146e-06,
      "loss": 1.4587,
      "step": 2840
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.083951735473316e-06,
      "loss": 1.4603,
      "step": 2842
    },
    {
      "epoch": 2.56,
      "learning_rate": 6.070517122715833e-06,
      "loss": 1.5366,
      "step": 2844
    },
    {
      "epoch": 2.57,
      "learning_rate": 6.057090892085571e-06,
      "loss": 1.5082,
      "step": 2846
    },
    {
      "epoch": 2.57,
      "learning_rate": 6.043673072222525e-06,
      "loss": 1.5352,
      "step": 2848
    },
    {
      "epoch": 2.57,
      "learning_rate": 6.030263691748754e-06,
      "loss": 1.5126,
      "step": 2850
    },
    {
      "epoch": 2.57,
      "learning_rate": 6.016862779268301e-06,
      "loss": 1.4523,
      "step": 2852
    },
    {
      "epoch": 2.57,
      "learning_rate": 6.003470363367161e-06,
      "loss": 1.4748,
      "step": 2854
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.990086472613195e-06,
      "loss": 1.5156,
      "step": 2856
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.976711135556086e-06,
      "loss": 1.4555,
      "step": 2858
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.963344380727267e-06,
      "loss": 1.546,
      "step": 2860
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.949986236639857e-06,
      "loss": 1.522,
      "step": 2862
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.936636731788621e-06,
      "loss": 1.5375,
      "step": 2864
    },
    {
      "epoch": 2.58,
      "learning_rate": 5.9232958946498854e-06,
      "loss": 1.5262,
      "step": 2866
    },
    {
      "epoch": 2.59,
      "learning_rate": 5.909963753681492e-06,
      "loss": 1.54,
      "step": 2868
    },
    {
      "epoch": 2.59,
      "learning_rate": 5.896640337322725e-06,
      "loss": 1.6126,
      "step": 2870
    },
    {
      "epoch": 2.59,
      "learning_rate": 5.883325673994274e-06,
      "loss": 1.515,
      "step": 2872
    },
    {
      "epoch": 2.59,
      "learning_rate": 5.870019792098134e-06,
      "loss": 1.4953,
      "step": 2874
    },
    {
      "epoch": 2.59,
      "learning_rate": 5.8567227200175865e-06,
      "loss": 1.5431,
      "step": 2876
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.843434486117115e-06,
      "loss": 1.5919,
      "step": 2878
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.830155118742347e-06,
      "loss": 1.4651,
      "step": 2880
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.816884646219997e-06,
      "loss": 1.5157,
      "step": 2882
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.803623096857819e-06,
      "loss": 1.5199,
      "step": 2884
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.790370498944504e-06,
      "loss": 1.4584,
      "step": 2886
    },
    {
      "epoch": 2.6,
      "learning_rate": 5.7771268807496794e-06,
      "loss": 1.4451,
      "step": 2888
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.763892270523792e-06,
      "loss": 1.501,
      "step": 2890
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.750666696498097e-06,
      "loss": 1.4653,
      "step": 2892
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.737450186884555e-06,
      "loss": 1.5312,
      "step": 2894
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.724242769875794e-06,
      "loss": 1.5635,
      "step": 2896
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.711044473645061e-06,
      "loss": 1.5362,
      "step": 2898
    },
    {
      "epoch": 2.61,
      "learning_rate": 5.6978553263461265e-06,
      "loss": 1.519,
      "step": 2900
    },
    {
      "epoch": 2.62,
      "learning_rate": 5.684675356113263e-06,
      "loss": 1.5324,
      "step": 2902
    },
    {
      "epoch": 2.62,
      "learning_rate": 5.6715045910611525e-06,
      "loss": 1.5318,
      "step": 2904
    },
    {
      "epoch": 2.62,
      "learning_rate": 5.6583430592848565e-06,
      "loss": 1.556,
      "step": 2906
    },
    {
      "epoch": 2.62,
      "learning_rate": 5.645190788859719e-06,
      "loss": 1.4563,
      "step": 2908
    },
    {
      "epoch": 2.62,
      "learning_rate": 5.632047807841352e-06,
      "loss": 1.5101,
      "step": 2910
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.618914144265532e-06,
      "loss": 1.4905,
      "step": 2912
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.605789826148178e-06,
      "loss": 1.553,
      "step": 2914
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.592674881485258e-06,
      "loss": 1.5348,
      "step": 2916
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.579569338252758e-06,
      "loss": 1.5323,
      "step": 2918
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.5664732244066015e-06,
      "loss": 1.5046,
      "step": 2920
    },
    {
      "epoch": 2.63,
      "learning_rate": 5.553386567882592e-06,
      "loss": 1.4862,
      "step": 2922
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.5403093965963806e-06,
      "loss": 1.5642,
      "step": 2924
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.5272417384433605e-06,
      "loss": 1.5198,
      "step": 2926
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.514183621298655e-06,
      "loss": 1.4895,
      "step": 2928
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.501135073017008e-06,
      "loss": 1.5053,
      "step": 2930
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.488096121432778e-06,
      "loss": 1.4725,
      "step": 2932
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.475066794359833e-06,
      "loss": 1.5248,
      "step": 2934
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.4620471195915304e-06,
      "loss": 1.5257,
      "step": 2936
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.449037124900615e-06,
      "loss": 1.524,
      "step": 2938
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.4360368380392025e-06,
      "loss": 1.4771,
      "step": 2940
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.42304628673869e-06,
      "loss": 1.5273,
      "step": 2942
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.4100654987097044e-06,
      "loss": 1.488,
      "step": 2944
    },
    {
      "epoch": 2.66,
      "learning_rate": 5.397094501642059e-06,
      "loss": 1.5327,
      "step": 2946
    },
    {
      "epoch": 2.66,
      "learning_rate": 5.384133323204666e-06,
      "loss": 1.5387,
      "step": 2948
    },
    {
      "epoch": 2.66,
      "learning_rate": 5.371181991045512e-06,
      "loss": 1.4628,
      "step": 2950
    },
    {
      "epoch": 2.66,
      "learning_rate": 5.358240532791555e-06,
      "loss": 1.4749,
      "step": 2952
    },
    {
      "epoch": 2.66,
      "learning_rate": 5.345308976048715e-06,
      "loss": 1.5011,
      "step": 2954
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.332387348401775e-06,
      "loss": 1.5534,
      "step": 2956
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.319475677414348e-06,
      "loss": 1.4919,
      "step": 2958
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.306573990628796e-06,
      "loss": 1.464,
      "step": 2960
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.293682315566202e-06,
      "loss": 1.4961,
      "step": 2962
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.280800679726272e-06,
      "loss": 1.509,
      "step": 2964
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.267929110587308e-06,
      "loss": 1.4824,
      "step": 2966
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.255067635606139e-06,
      "loss": 1.4869,
      "step": 2968
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.242216282218057e-06,
      "loss": 1.5432,
      "step": 2970
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.229375077836769e-06,
      "loss": 1.5251,
      "step": 2972
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.2165440498543276e-06,
      "loss": 1.5558,
      "step": 2974
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.20372322564108e-06,
      "loss": 1.553,
      "step": 2976
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.190912632545599e-06,
      "loss": 1.5385,
      "step": 2978
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.178112297894655e-06,
      "loss": 1.4673,
      "step": 2980
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.16532224899311e-06,
      "loss": 1.5011,
      "step": 2982
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.152542513123906e-06,
      "loss": 1.5183,
      "step": 2984
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.13977311754797e-06,
      "loss": 1.4702,
      "step": 2986
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.1270140895041785e-06,
      "loss": 1.5591,
      "step": 2988
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.114265456209297e-06,
      "loss": 1.5238,
      "step": 2990
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.101527244857907e-06,
      "loss": 1.5689,
      "step": 2992
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.088799482622372e-06,
      "loss": 1.5431,
      "step": 2994
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.076082196652754e-06,
      "loss": 1.6206,
      "step": 2996
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.063375414076775e-06,
      "loss": 1.4451,
      "step": 2998
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.050679161999741e-06,
      "loss": 1.5286,
      "step": 3000
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.037993467504515e-06,
      "loss": 1.5318,
      "step": 3002
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.025318357651419e-06,
      "loss": 1.5027,
      "step": 3004
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.012653859478209e-06,
      "loss": 1.5066,
      "step": 3006
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.000000000000003e-06,
      "loss": 1.5122,
      "step": 3008
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.987356806209214e-06,
      "loss": 1.5269,
      "step": 3010
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.974724305075523e-06,
      "loss": 1.5193,
      "step": 3012
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.962102523545782e-06,
      "loss": 1.5,
      "step": 3014
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.949491488543996e-06,
      "loss": 1.5606,
      "step": 3016
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.936891226971232e-06,
      "loss": 1.4576,
      "step": 3018
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.92430176570558e-06,
      "loss": 1.53,
      "step": 3020
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.911723131602089e-06,
      "loss": 1.5139,
      "step": 3022
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.899155351492721e-06,
      "loss": 1.5269,
      "step": 3024
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.88659845218627e-06,
      "loss": 1.4767,
      "step": 3026
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.874052460468339e-06,
      "loss": 1.5009,
      "step": 3028
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.861517403101249e-06,
      "loss": 1.5074,
      "step": 3030
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.848993306823997e-06,
      "loss": 1.5336,
      "step": 3032
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.836480198352208e-06,
      "loss": 1.4969,
      "step": 3034
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.823978104378056e-06,
      "loss": 1.4998,
      "step": 3036
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.811487051570235e-06,
      "loss": 1.4481,
      "step": 3038
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.7990070665738734e-06,
      "loss": 1.572,
      "step": 3040
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.786538176010494e-06,
      "loss": 1.516,
      "step": 3042
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.774080406477954e-06,
      "loss": 1.5844,
      "step": 3044
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.761633784550393e-06,
      "loss": 1.5226,
      "step": 3046
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.749198336778159e-06,
      "loss": 1.5655,
      "step": 3048
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.736774089687781e-06,
      "loss": 1.4839,
      "step": 3050
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.724361069781882e-06,
      "loss": 1.498,
      "step": 3052
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.711959303539137e-06,
      "loss": 1.485,
      "step": 3054
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.699568817414224e-06,
      "loss": 1.484,
      "step": 3056
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.687189637837748e-06,
      "loss": 1.4352,
      "step": 3058
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.674821791216207e-06,
      "loss": 1.4599,
      "step": 3060
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.662465303931912e-06,
      "loss": 1.5327,
      "step": 3062
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.650120202342959e-06,
      "loss": 1.4812,
      "step": 3064
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.637786512783134e-06,
      "loss": 1.4903,
      "step": 3066
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.625464261561902e-06,
      "loss": 1.4594,
      "step": 3068
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.613153474964311e-06,
      "loss": 1.4989,
      "step": 3070
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.600854179250971e-06,
      "loss": 1.5395,
      "step": 3072
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.588566400657965e-06,
      "loss": 1.4893,
      "step": 3074
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.576290165396811e-06,
      "loss": 1.5686,
      "step": 3076
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.564025499654414e-06,
      "loss": 1.5599,
      "step": 3078
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.551772429592987e-06,
      "loss": 1.5061,
      "step": 3080
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.539530981350017e-06,
      "loss": 1.5466,
      "step": 3082
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.527301181038189e-06,
      "loss": 1.5349,
      "step": 3084
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.515083054745363e-06,
      "loss": 1.5065,
      "step": 3086
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.502876628534465e-06,
      "loss": 1.5608,
      "step": 3088
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.49068192844349e-06,
      "loss": 1.5091,
      "step": 3090
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.4784989804854055e-06,
      "loss": 1.5016,
      "step": 3092
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.46632781064812e-06,
      "loss": 1.482,
      "step": 3094
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.454168444894404e-06,
      "loss": 1.574,
      "step": 3096
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.4420209091618675e-06,
      "loss": 1.5043,
      "step": 3098
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.429885229362866e-06,
      "loss": 1.5039,
      "step": 3100
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.417761431384471e-06,
      "loss": 1.5083,
      "step": 3102
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.405649541088419e-06,
      "loss": 1.4593,
      "step": 3104
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.393549584311029e-06,
      "loss": 1.4436,
      "step": 3106
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.381461586863183e-06,
      "loss": 1.4931,
      "step": 3108
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.369385574530227e-06,
      "loss": 1.5188,
      "step": 3110
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.3573215730719665e-06,
      "loss": 1.4873,
      "step": 3112
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.3452696082225685e-06,
      "loss": 1.4547,
      "step": 3114
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.3332297056905385e-06,
      "loss": 1.5253,
      "step": 3116
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.321201891158635e-06,
      "loss": 1.4432,
      "step": 3118
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.309186190283849e-06,
      "loss": 1.5252,
      "step": 3120
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.2971826286973175e-06,
      "loss": 1.4943,
      "step": 3122
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.285191232004285e-06,
      "loss": 1.4916,
      "step": 3124
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.273212025784056e-06,
      "loss": 1.4142,
      "step": 3126
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.261245035589917e-06,
      "loss": 1.5139,
      "step": 3128
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.24929028694911e-06,
      "loss": 1.4081,
      "step": 3130
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.237347805362757e-06,
      "loss": 1.4579,
      "step": 3132
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.225417616305809e-06,
      "loss": 1.536,
      "step": 3134
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.213499745226999e-06,
      "loss": 1.5318,
      "step": 3136
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.201594217548789e-06,
      "loss": 1.4739,
      "step": 3138
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.189701058667301e-06,
      "loss": 1.5138,
      "step": 3140
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.177820293952285e-06,
      "loss": 1.4591,
      "step": 3142
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.165951948747043e-06,
      "loss": 1.4755,
      "step": 3144
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.15409604836838e-06,
      "loss": 1.5083,
      "step": 3146
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.142252618106573e-06,
      "loss": 1.5349,
      "step": 3148
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.1304216832252755e-06,
      "loss": 1.5078,
      "step": 3150
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.118603268961509e-06,
      "loss": 1.4798,
      "step": 3152
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.1067974005255705e-06,
      "loss": 1.5029,
      "step": 3154
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.095004103101001e-06,
      "loss": 1.5056,
      "step": 3156
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.083223401844523e-06,
      "loss": 1.5195,
      "step": 3158
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.071455321885996e-06,
      "loss": 1.5299,
      "step": 3160
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.059699888328348e-06,
      "loss": 1.5126,
      "step": 3162
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.047957126247542e-06,
      "loss": 1.4871,
      "step": 3164
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.036227060692498e-06,
      "loss": 1.487,
      "step": 3166
    },
    {
      "epoch": 2.86,
      "learning_rate": 4.024509716685058e-06,
      "loss": 1.5214,
      "step": 3168
    },
    {
      "epoch": 2.86,
      "learning_rate": 4.0128051192199325e-06,
      "loss": 1.4821,
      "step": 3170
    },
    {
      "epoch": 2.86,
      "learning_rate": 4.001113293264631e-06,
      "loss": 1.46,
      "step": 3172
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.989434263759433e-06,
      "loss": 1.4978,
      "step": 3174
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.97776805561731e-06,
      "loss": 1.4738,
      "step": 3176
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.9661146937238895e-06,
      "loss": 1.5618,
      "step": 3178
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.954474202937389e-06,
      "loss": 1.5181,
      "step": 3180
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.942846608088583e-06,
      "loss": 1.482,
      "step": 3182
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.931231933980724e-06,
      "loss": 1.4962,
      "step": 3184
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.919630205389516e-06,
      "loss": 1.5599,
      "step": 3186
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.908041447063034e-06,
      "loss": 1.497,
      "step": 3188
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.896465683721695e-06,
      "loss": 1.4755,
      "step": 3190
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.884902940058196e-06,
      "loss": 1.5611,
      "step": 3192
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.8733532407374555e-06,
      "loss": 1.543,
      "step": 3194
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.861816610396574e-06,
      "loss": 1.506,
      "step": 3196
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.850293073644767e-06,
      "loss": 1.5262,
      "step": 3198
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.838782655063326e-06,
      "loss": 1.5193,
      "step": 3200
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.827285379205549e-06,
      "loss": 1.5002,
      "step": 3202
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.815801270596715e-06,
      "loss": 1.482,
      "step": 3204
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.804330353733998e-06,
      "loss": 1.4588,
      "step": 3206
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.7928726530864502e-06,
      "loss": 1.4336,
      "step": 3208
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.7814281930949173e-06,
      "loss": 1.515,
      "step": 3210
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.769996998172002e-06,
      "loss": 1.541,
      "step": 3212
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.758579092702023e-06,
      "loss": 1.5338,
      "step": 3214
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.7471745010409333e-06,
      "loss": 1.5087,
      "step": 3216
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.735783247516305e-06,
      "loss": 1.479,
      "step": 3218
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.7244053564272377e-06,
      "loss": 1.5071,
      "step": 3220
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.7130408520443505e-06,
      "loss": 1.4724,
      "step": 3222
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.7016897586096778e-06,
      "loss": 1.5396,
      "step": 3224
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.690352100336675e-06,
      "loss": 1.5369,
      "step": 3226
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.679027901410117e-06,
      "loss": 1.5212,
      "step": 3228
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.667717185986085e-06,
      "loss": 1.4958,
      "step": 3230
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.6564199781918865e-06,
      "loss": 1.4261,
      "step": 3232
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6451363021260146e-06,
      "loss": 1.4705,
      "step": 3234
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.63386618185811e-06,
      "loss": 1.4583,
      "step": 3236
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6226096414288816e-06,
      "loss": 1.4864,
      "step": 3238
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6113667048500833e-06,
      "loss": 1.5096,
      "step": 3240
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6001373961044385e-06,
      "loss": 1.5321,
      "step": 3242
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5889217391456167e-06,
      "loss": 1.4713,
      "step": 3244
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5777197578981425e-06,
      "loss": 1.5169,
      "step": 3246
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5665314762573933e-06,
      "loss": 1.4016,
      "step": 3248
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5553569180895044e-06,
      "loss": 1.4573,
      "step": 3250
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5441961072313504e-06,
      "loss": 1.5512,
      "step": 3252
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5330490674904737e-06,
      "loss": 1.4868,
      "step": 3254
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.521915822645039e-06,
      "loss": 1.5873,
      "step": 3256
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.510796396443793e-06,
      "loss": 1.5238,
      "step": 3258
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.499690812605997e-06,
      "loss": 1.5263,
      "step": 3260
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.4885990948213922e-06,
      "loss": 1.5268,
      "step": 3262
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.477521266750131e-06,
      "loss": 1.5712,
      "step": 3264
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.4664573520227564e-06,
      "loss": 1.5208,
      "step": 3266
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.4554073742401028e-06,
      "loss": 1.4913,
      "step": 3268
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.444371356973305e-06,
      "loss": 1.5158,
      "step": 3270
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.433349323763696e-06,
      "loss": 1.4721,
      "step": 3272
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.4223412981227964e-06,
      "loss": 1.4383,
      "step": 3274
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.411347303532231e-06,
      "loss": 1.5057,
      "step": 3276
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.4003673634437084e-06,
      "loss": 1.4861,
      "step": 3278
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.389401501278946e-06,
      "loss": 1.469,
      "step": 3280
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.3784497404296347e-06,
      "loss": 1.5809,
      "step": 3282
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.367512104257391e-06,
      "loss": 1.4898,
      "step": 3284
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.35658861609369e-06,
      "loss": 1.4907,
      "step": 3286
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.3456792992398433e-06,
      "loss": 1.5118,
      "step": 3288
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.334784176966912e-06,
      "loss": 1.4053,
      "step": 3290
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.3239032725156983e-06,
      "loss": 1.5658,
      "step": 3292
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.3130366090966604e-06,
      "loss": 1.5098,
      "step": 3294
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.3021842098898938e-06,
      "loss": 1.5287,
      "step": 3296
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.2913460980450506e-06,
      "loss": 1.5354,
      "step": 3298
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.2805222966813187e-06,
      "loss": 1.509,
      "step": 3300
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.2697128288873536e-06,
      "loss": 1.6128,
      "step": 3302
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.2589177177212304e-06,
      "loss": 1.5342,
      "step": 3304
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.248136986210414e-06,
      "loss": 1.5372,
      "step": 3306
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.2373706573516795e-06,
      "loss": 1.4921,
      "step": 3308
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.2266187541110937e-06,
      "loss": 1.5076,
      "step": 3310
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.2158812994239396e-06,
      "loss": 1.5531,
      "step": 3312
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.2051583161946865e-06,
      "loss": 1.4592,
      "step": 3314
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.1944498272969273e-06,
      "loss": 1.4771,
      "step": 3316
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.183755855573346e-06,
      "loss": 1.5302,
      "step": 3318
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.1730764238356517e-06,
      "loss": 1.4898,
      "step": 3320
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.162411554864544e-06,
      "loss": 1.5458,
      "step": 3322
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.1517612714096534e-06,
      "loss": 1.5402,
      "step": 3324
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.141125596189494e-06,
      "loss": 1.5306,
      "step": 3326
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.130504551891431e-06,
      "loss": 1.467,
      "step": 3328
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.119898161171605e-06,
      "loss": 1.5479,
      "step": 3330
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.109306446654912e-06,
      "loss": 1.444,
      "step": 3332
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.0987294309349316e-06,
      "loss": 1.5628,
      "step": 3334
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.088167136573894e-06,
      "loss": 1.473,
      "step": 3336
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.0776195861026202e-06,
      "loss": 1.5232,
      "step": 3338
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.0670868020204936e-06,
      "loss": 1.4349,
      "step": 3340
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.0565688067953836e-06,
      "loss": 1.4101,
      "step": 3342
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.0460656228636254e-06,
      "loss": 1.5501,
      "step": 3344
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.0355772726299536e-06,
      "loss": 1.5632,
      "step": 3346
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.0251037784674555e-06,
      "loss": 1.4991,
      "step": 3348
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.014645162717542e-06,
      "loss": 1.543,
      "step": 3350
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.0042014476898717e-06,
      "loss": 1.4972,
      "step": 3352
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.9937726556623303e-06,
      "loss": 1.3964,
      "step": 3354
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.9833588088809627e-06,
      "loss": 1.4989,
      "step": 3356
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.9729599295599366e-06,
      "loss": 1.5028,
      "step": 3358
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.962576039881485e-06,
      "loss": 1.4963,
      "step": 3360
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.952207161995879e-06,
      "loss": 1.4059,
      "step": 3362
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.9418533180213613e-06,
      "loss": 1.4411,
      "step": 3364
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.9315145300441006e-06,
      "loss": 1.4897,
      "step": 3366
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.9211908201181592e-06,
      "loss": 1.4588,
      "step": 3368
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.910882210265421e-06,
      "loss": 1.5836,
      "step": 3370
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.9005887224755745e-06,
      "loss": 1.4342,
      "step": 3372
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.8903103787060395e-06,
      "loss": 1.4968,
      "step": 3374
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.8800472008819426e-06,
      "loss": 1.5249,
      "step": 3376
    },
    {
      "epoch": 3.05,
      "learning_rate": 2.8697992108960494e-06,
      "loss": 1.447,
      "step": 3378
    },
    {
      "epoch": 3.05,
      "learning_rate": 2.8595664306087313e-06,
      "loss": 1.5357,
      "step": 3380
    },
    {
      "epoch": 3.05,
      "learning_rate": 2.849348881847913e-06,
      "loss": 1.5141,
      "step": 3382
    },
    {
      "epoch": 3.05,
      "learning_rate": 2.8391465864090338e-06,
      "loss": 1.4493,
      "step": 3384
    },
    {
      "epoch": 3.05,
      "learning_rate": 2.828959566054996e-06,
      "loss": 1.5233,
      "step": 3386
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.8187878425161088e-06,
      "loss": 1.48,
      "step": 3388
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.8086314374900637e-06,
      "loss": 1.5077,
      "step": 3390
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.79849037264186e-06,
      "loss": 1.5296,
      "step": 3392
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.788364669603789e-06,
      "loss": 1.4566,
      "step": 3394
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.7782543499753644e-06,
      "loss": 1.5193,
      "step": 3396
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.7681594353232934e-06,
      "loss": 1.5448,
      "step": 3398
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.758079947181409e-06,
      "loss": 1.4608,
      "step": 3400
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.7480159070506517e-06,
      "loss": 1.4614,
      "step": 3402
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.737967336399002e-06,
      "loss": 1.5236,
      "step": 3404
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.7279342566614374e-06,
      "loss": 1.4661,
      "step": 3406
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.717916689239903e-06,
      "loss": 1.5057,
      "step": 3408
    },
    {
      "epoch": 3.07,
      "learning_rate": 2.707914655503242e-06,
      "loss": 1.4687,
      "step": 3410
    },
    {
      "epoch": 3.08,
      "learning_rate": 2.697928176787176e-06,
      "loss": 1.4575,
      "step": 3412
    },
    {
      "epoch": 3.08,
      "learning_rate": 2.687957274394224e-06,
      "loss": 1.5014,
      "step": 3414
    },
    {
      "epoch": 3.08,
      "learning_rate": 2.678001969593701e-06,
      "loss": 1.4898,
      "step": 3416
    },
    {
      "epoch": 3.08,
      "learning_rate": 2.6680622836216375e-06,
      "loss": 1.4814,
      "step": 3418
    },
    {
      "epoch": 3.08,
      "learning_rate": 2.658138237680753e-06,
      "loss": 1.4391,
      "step": 3420
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.6482298529403973e-06,
      "loss": 1.4201,
      "step": 3422
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.638337150536523e-06,
      "loss": 1.5524,
      "step": 3424
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.6284601515716213e-06,
      "loss": 1.522,
      "step": 3426
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.6185988771146864e-06,
      "loss": 1.5003,
      "step": 3428
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.608753348201177e-06,
      "loss": 1.4368,
      "step": 3430
    },
    {
      "epoch": 3.09,
      "learning_rate": 2.5989235858329575e-06,
      "loss": 1.5304,
      "step": 3432
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.5891096109782644e-06,
      "loss": 1.4369,
      "step": 3434
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.579311444571655e-06,
      "loss": 1.4968,
      "step": 3436
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.569529107513964e-06,
      "loss": 1.4706,
      "step": 3438
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.5597626206722583e-06,
      "loss": 1.4667,
      "step": 3440
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.5500120048798026e-06,
      "loss": 1.3906,
      "step": 3442
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.540277280935998e-06,
      "loss": 1.4251,
      "step": 3444
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.530558469606351e-06,
      "loss": 1.4692,
      "step": 3446
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.520855591622421e-06,
      "loss": 1.4652,
      "step": 3448
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.511168667681779e-06,
      "loss": 1.4424,
      "step": 3450
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.5014977184479696e-06,
      "loss": 1.4772,
      "step": 3452
    },
    {
      "epoch": 3.11,
      "learning_rate": 2.491842764550453e-06,
      "loss": 1.4563,
      "step": 3454
    },
    {
      "epoch": 3.12,
      "learning_rate": 2.482203826584577e-06,
      "loss": 1.4652,
      "step": 3456
    },
    {
      "epoch": 3.12,
      "learning_rate": 2.4725809251115208e-06,
      "loss": 1.5157,
      "step": 3458
    },
    {
      "epoch": 3.12,
      "learning_rate": 2.4629740806582534e-06,
      "loss": 1.4707,
      "step": 3460
    },
    {
      "epoch": 3.12,
      "learning_rate": 2.453383313717496e-06,
      "loss": 1.4329,
      "step": 3462
    },
    {
      "epoch": 3.12,
      "learning_rate": 2.443808644747675e-06,
      "loss": 1.4575,
      "step": 3464
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.434250094172872e-06,
      "loss": 1.4706,
      "step": 3466
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.424707682382794e-06,
      "loss": 1.4185,
      "step": 3468
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.4151814297327157e-06,
      "loss": 1.4808,
      "step": 3470
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.405671356543441e-06,
      "loss": 1.4116,
      "step": 3472
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.3961774831012672e-06,
      "loss": 1.5432,
      "step": 3474
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.386699829657928e-06,
      "loss": 1.4426,
      "step": 3476
    },
    {
      "epoch": 3.14,
      "learning_rate": 2.3772384164305662e-06,
      "loss": 1.4822,
      "step": 3478
    },
    {
      "epoch": 3.14,
      "learning_rate": 2.367793263601673e-06,
      "loss": 1.555,
      "step": 3480
    },
    {
      "epoch": 3.14,
      "learning_rate": 2.35836439131906e-06,
      "loss": 1.5162,
      "step": 3482
    },
    {
      "epoch": 3.14,
      "learning_rate": 2.348951819695805e-06,
      "loss": 1.5948,
      "step": 3484
    },
    {
      "epoch": 3.14,
      "learning_rate": 2.339555568810221e-06,
      "loss": 1.4229,
      "step": 3486
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.330175658705799e-06,
      "loss": 1.5137,
      "step": 3488
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.3254918377005877e-06,
      "loss": 1.4265,
      "step": 3490
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.3161364762732095e-06,
      "loss": 1.4405,
      "step": 3492
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.3067975055831337e-06,
      "loss": 1.4645,
      "step": 3494
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.2974749455516786e-06,
      "loss": 1.5029,
      "step": 3496
    },
    {
      "epoch": 3.15,
      "learning_rate": 2.2881688160651406e-06,
      "loss": 1.4532,
      "step": 3498
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.2788791369747774e-06,
      "loss": 1.5009,
      "step": 3500
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.2696059280967474e-06,
      "loss": 1.4555,
      "step": 3502
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.260349209212086e-06,
      "loss": 1.4426,
      "step": 3504
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.2511090000666424e-06,
      "loss": 1.5772,
      "step": 3506
    },
    {
      "epoch": 3.16,
      "learning_rate": 2.2418853203710523e-06,
      "loss": 1.4803,
      "step": 3508
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.232678189800698e-06,
      "loss": 1.4887,
      "step": 3510
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.2234876279956484e-06,
      "loss": 1.4796,
      "step": 3512
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.214313654560644e-06,
      "loss": 1.5214,
      "step": 3514
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.20515628906502e-06,
      "loss": 1.5006,
      "step": 3516
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.196015551042706e-06,
      "loss": 1.5007,
      "step": 3518
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.186891459992145e-06,
      "loss": 1.4674,
      "step": 3520
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.177784035376286e-06,
      "loss": 1.4773,
      "step": 3522
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.1686932966225094e-06,
      "loss": 1.459,
      "step": 3524
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.1596192631226175e-06,
      "loss": 1.4893,
      "step": 3526
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.150561954232768e-06,
      "loss": 1.481,
      "step": 3528
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.1415213892734452e-06,
      "loss": 1.4857,
      "step": 3530
    },
    {
      "epoch": 3.18,
      "learning_rate": 2.1324975875294207e-06,
      "loss": 1.4183,
      "step": 3532
    },
    {
      "epoch": 3.19,
      "learning_rate": 2.1234905682496987e-06,
      "loss": 1.5281,
      "step": 3534
    },
    {
      "epoch": 3.19,
      "learning_rate": 2.1145003506474957e-06,
      "loss": 1.4928,
      "step": 3536
    },
    {
      "epoch": 3.19,
      "learning_rate": 2.10552695390018e-06,
      "loss": 1.4607,
      "step": 3538
    },
    {
      "epoch": 3.19,
      "learning_rate": 2.096570397149239e-06,
      "loss": 1.4789,
      "step": 3540
    },
    {
      "epoch": 3.19,
      "learning_rate": 2.0876306995002384e-06,
      "loss": 1.4629,
      "step": 3542
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.0787078800227868e-06,
      "loss": 1.4693,
      "step": 3544
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.069801957750479e-06,
      "loss": 1.5246,
      "step": 3546
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.060912951680878e-06,
      "loss": 1.5408,
      "step": 3548
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.0520408807754532e-06,
      "loss": 1.4941,
      "step": 3550
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.043185763959549e-06,
      "loss": 1.4488,
      "step": 3552
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.034347620122351e-06,
      "loss": 1.4593,
      "step": 3554
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.0255264681168306e-06,
      "loss": 1.5067,
      "step": 3556
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.0167223267597246e-06,
      "loss": 1.452,
      "step": 3558
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.0079352148314724e-06,
      "loss": 1.4799,
      "step": 3560
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.999165151076192e-06,
      "loss": 1.4979,
      "step": 3562
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.990412154201633e-06,
      "loss": 1.4793,
      "step": 3564
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.9816762428791457e-06,
      "loss": 1.508,
      "step": 3566
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.9729574357436266e-06,
      "loss": 1.435,
      "step": 3568
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.9642557513934935e-06,
      "loss": 1.5827,
      "step": 3570
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.955571208390633e-06,
      "loss": 1.4942,
      "step": 3572
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.946903825260368e-06,
      "loss": 1.5088,
      "step": 3574
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.9382536204914214e-06,
      "loss": 1.486,
      "step": 3576
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.9296206125358664e-06,
      "loss": 1.4394,
      "step": 3578
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.921004819809099e-06,
      "loss": 1.4363,
      "step": 3580
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.9124062606897884e-06,
      "loss": 1.3946,
      "step": 3582
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.9038249535198428e-06,
      "loss": 1.4707,
      "step": 3584
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.895260916604368e-06,
      "loss": 1.46,
      "step": 3586
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.8867141682116373e-06,
      "loss": 1.5116,
      "step": 3588
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.8781847265730347e-06,
      "loss": 1.4524,
      "step": 3590
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.8696726098830364e-06,
      "loss": 1.4589,
      "step": 3592
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.861177836299155e-06,
      "loss": 1.3879,
      "step": 3594
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.85270042394191e-06,
      "loss": 1.4768,
      "step": 3596
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.8442403908947904e-06,
      "loss": 1.394,
      "step": 3598
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.835797755204205e-06,
      "loss": 1.4304,
      "step": 3600
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.8273725348794624e-06,
      "loss": 1.5056,
      "step": 3602
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.818964747892713e-06,
      "loss": 1.559,
      "step": 3604
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.8105744121789226e-06,
      "loss": 1.5817,
      "step": 3606
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.8022015456358277e-06,
      "loss": 1.4484,
      "step": 3608
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.7938461661239093e-06,
      "loss": 1.4658,
      "step": 3610
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.7855082914663346e-06,
      "loss": 1.4426,
      "step": 3612
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.7771879394489432e-06,
      "loss": 1.5372,
      "step": 3614
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.768885127820187e-06,
      "loss": 1.4458,
      "step": 3616
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.7605998742911001e-06,
      "loss": 1.4432,
      "step": 3618
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.7523321965352757e-06,
      "loss": 1.4313,
      "step": 3620
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.7440821121887997e-06,
      "loss": 1.4853,
      "step": 3622
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.7358496388502422e-06,
      "loss": 1.4187,
      "step": 3624
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.727634794080596e-06,
      "loss": 1.5036,
      "step": 3626
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.7194375954032604e-06,
      "loss": 1.5692,
      "step": 3628
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.7112580603039785e-06,
      "loss": 1.4624,
      "step": 3630
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.703096206230831e-06,
      "loss": 1.4422,
      "step": 3632
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.6949520505941686e-06,
      "loss": 1.5489,
      "step": 3634
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.6868256107666015e-06,
      "loss": 1.4635,
      "step": 3636
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.6787169040829387e-06,
      "loss": 1.4007,
      "step": 3638
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.6706259478401653e-06,
      "loss": 1.5285,
      "step": 3640
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.6625527592974077e-06,
      "loss": 1.468,
      "step": 3642
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.6544973556758826e-06,
      "loss": 1.4775,
      "step": 3644
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.646459754158878e-06,
      "loss": 1.4737,
      "step": 3646
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.6384399718916977e-06,
      "loss": 1.5256,
      "step": 3648
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.6304380259816477e-06,
      "loss": 1.473,
      "step": 3650
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.6224539334979683e-06,
      "loss": 1.5123,
      "step": 3652
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.614487711471835e-06,
      "loss": 1.4314,
      "step": 3654
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.6065393768962889e-06,
      "loss": 1.4433,
      "step": 3656
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.5986089467262256e-06,
      "loss": 1.4855,
      "step": 3658
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.5906964378783373e-06,
      "loss": 1.4353,
      "step": 3660
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.5828018672310996e-06,
      "loss": 1.4591,
      "step": 3662
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.5749252516247139e-06,
      "loss": 1.4398,
      "step": 3664
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.5670666078610809e-06,
      "loss": 1.449,
      "step": 3666
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.5592259527037744e-06,
      "loss": 1.4773,
      "step": 3668
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.551403302877984e-06,
      "loss": 1.5057,
      "step": 3670
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.5435986750705046e-06,
      "loss": 1.4788,
      "step": 3672
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.5358120859296721e-06,
      "loss": 1.4658,
      "step": 3674
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.5280435520653569e-06,
      "loss": 1.4303,
      "step": 3676
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.5202930900489056e-06,
      "loss": 1.4822,
      "step": 3678
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.512560716413125e-06,
      "loss": 1.4529,
      "step": 3680
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.5048464476522251e-06,
      "loss": 1.4679,
      "step": 3682
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.497150300221808e-06,
      "loss": 1.5103,
      "step": 3684
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.4894722905388103e-06,
      "loss": 1.5214,
      "step": 3686
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.4818124349814812e-06,
      "loss": 1.4546,
      "step": 3688
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.4741707498893487e-06,
      "loss": 1.4666,
      "step": 3690
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.466547251563175e-06,
      "loss": 1.54,
      "step": 3692
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.4589419562649342e-06,
      "loss": 1.5121,
      "step": 3694
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.4513548802177635e-06,
      "loss": 1.5541,
      "step": 3696
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.44378603960594e-06,
      "loss": 1.4956,
      "step": 3698
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.43623545057484e-06,
      "loss": 1.4957,
      "step": 3700
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.4287031292309105e-06,
      "loss": 1.5291,
      "step": 3702
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.4211890916416249e-06,
      "loss": 1.461,
      "step": 3704
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.4136933538354636e-06,
      "loss": 1.4236,
      "step": 3706
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.406215931801862e-06,
      "loss": 1.4785,
      "step": 3708
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.398756841491189e-06,
      "loss": 1.4242,
      "step": 3710
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.3913160988147134e-06,
      "loss": 1.4035,
      "step": 3712
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.383893719644558e-06,
      "loss": 1.409,
      "step": 3714
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.3764897198136818e-06,
      "loss": 1.4705,
      "step": 3716
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.3691041151158336e-06,
      "loss": 1.4163,
      "step": 3718
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.361736921305522e-06,
      "loss": 1.5217,
      "step": 3720
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.354388154097983e-06,
      "loss": 1.4778,
      "step": 3722
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.3470578291691506e-06,
      "loss": 1.4679,
      "step": 3724
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.339745962155613e-06,
      "loss": 1.4595,
      "step": 3726
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.332452568654592e-06,
      "loss": 1.4545,
      "step": 3728
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.3251776642238957e-06,
      "loss": 1.5006,
      "step": 3730
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3179212643818928e-06,
      "loss": 1.4342,
      "step": 3732
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3106833846074872e-06,
      "loss": 1.4721,
      "step": 3734
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3034640403400667e-06,
      "loss": 1.4343,
      "step": 3736
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.2962632469794901e-06,
      "loss": 1.4435,
      "step": 3738
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.289081019886036e-06,
      "loss": 1.503,
      "step": 3740
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.2819173743803848e-06,
      "loss": 1.4574,
      "step": 3742
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.2747723257435729e-06,
      "loss": 1.4816,
      "step": 3744
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.2676458892169752e-06,
      "loss": 1.465,
      "step": 3746
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.2605380800022594e-06,
      "loss": 1.4731,
      "step": 3748
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.2534489132613603e-06,
      "loss": 1.5092,
      "step": 3750
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.2463784041164461e-06,
      "loss": 1.4065,
      "step": 3752
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.2393265676498812e-06,
      "loss": 1.5526,
      "step": 3754
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.232293418904207e-06,
      "loss": 1.4752,
      "step": 3756
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.2252789728820913e-06,
      "loss": 1.4876,
      "step": 3758
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.2182832445463177e-06,
      "loss": 1.4785,
      "step": 3760
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.2113062488197347e-06,
      "loss": 1.4643,
      "step": 3762
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.2043480005852315e-06,
      "loss": 1.5868,
      "step": 3764
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1974085146857085e-06,
      "loss": 1.5001,
      "step": 3766
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1904878059240443e-06,
      "loss": 1.4765,
      "step": 3768
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1835858890630613e-06,
      "loss": 1.4759,
      "step": 3770
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1767027788254993e-06,
      "loss": 1.4678,
      "step": 3772
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1698384898939774e-06,
      "loss": 1.5006,
      "step": 3774
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.1629930369109643e-06,
      "loss": 1.4419,
      "step": 3776
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.156166434478758e-06,
      "loss": 1.4797,
      "step": 3778
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.149358697159435e-06,
      "loss": 1.48,
      "step": 3780
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.1425698394748408e-06,
      "loss": 1.4535,
      "step": 3782
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.1357998759065358e-06,
      "loss": 1.4366,
      "step": 3784
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.1290488208957894e-06,
      "loss": 1.4915,
      "step": 3786
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.1223166888435243e-06,
      "loss": 1.4562,
      "step": 3788
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.1156034941103067e-06,
      "loss": 1.517,
      "step": 3790
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.1089092510163025e-06,
      "loss": 1.4764,
      "step": 3792
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.1022339738412546e-06,
      "loss": 1.4986,
      "step": 3794
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.0955776768244465e-06,
      "loss": 1.5427,
      "step": 3796
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.088940374164672e-06,
      "loss": 1.401,
      "step": 3798
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.082322080020215e-06,
      "loss": 1.532,
      "step": 3800
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.0757228085088011e-06,
      "loss": 1.4071,
      "step": 3802
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.06914257370759e-06,
      "loss": 1.4352,
      "step": 3804
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.0625813896531212e-06,
      "loss": 1.4857,
      "step": 3806
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.0560392703413103e-06,
      "loss": 1.4942,
      "step": 3808
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.0495162297273876e-06,
      "loss": 1.4106,
      "step": 3810
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.0430122817259015e-06,
      "loss": 1.4163,
      "step": 3812
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.036527440210665e-06,
      "loss": 1.4498,
      "step": 3814
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.0300617190147388e-06,
      "loss": 1.4431,
      "step": 3816
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.0236151319303945e-06,
      "loss": 1.4551,
      "step": 3818
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.0171876927090852e-06,
      "loss": 1.4693,
      "step": 3820
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.010779415061428e-06,
      "loss": 1.4752,
      "step": 3822
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.0043903126571553e-06,
      "loss": 1.4794,
      "step": 3824
    },
    {
      "epoch": 3.45,
      "learning_rate": 9.980203991251047e-07,
      "loss": 1.4714,
      "step": 3826
    },
    {
      "epoch": 3.45,
      "learning_rate": 9.916696880531739e-07,
      "loss": 1.4768,
      "step": 3828
    },
    {
      "epoch": 3.45,
      "learning_rate": 9.85338192988309e-07,
      "loss": 1.4577,
      "step": 3830
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.79025927436451e-07,
      "loss": 1.4632,
      "step": 3832
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.727329048625355e-07,
      "loss": 1.4518,
      "step": 3834
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.664591386904432e-07,
      "loss": 1.456,
      "step": 3836
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.60204642302982e-07,
      "loss": 1.5072,
      "step": 3838
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.539694290418488e-07,
      "loss": 1.4777,
      "step": 3840
    },
    {
      "epoch": 3.46,
      "learning_rate": 9.477535122076142e-07,
      "loss": 1.5157,
      "step": 3842
    },
    {
      "epoch": 3.47,
      "learning_rate": 9.415569050596829e-07,
      "loss": 1.517,
      "step": 3844
    },
    {
      "epoch": 3.47,
      "learning_rate": 9.353796208162669e-07,
      "loss": 1.4609,
      "step": 3846
    },
    {
      "epoch": 3.47,
      "learning_rate": 9.292216726543668e-07,
      "loss": 1.388,
      "step": 3848
    },
    {
      "epoch": 3.47,
      "learning_rate": 9.230830737097297e-07,
      "loss": 1.5012,
      "step": 3850
    },
    {
      "epoch": 3.47,
      "learning_rate": 9.169638370768341e-07,
      "loss": 1.4105,
      "step": 3852
    },
    {
      "epoch": 3.48,
      "learning_rate": 9.108639758088489e-07,
      "loss": 1.4909,
      "step": 3854
    },
    {
      "epoch": 3.48,
      "learning_rate": 9.04783502917621e-07,
      "loss": 1.4654,
      "step": 3856
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.987224313736309e-07,
      "loss": 1.4628,
      "step": 3858
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.926807741059818e-07,
      "loss": 1.4521,
      "step": 3860
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.866585440023545e-07,
      "loss": 1.5005,
      "step": 3862
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.806557539089988e-07,
      "loss": 1.4683,
      "step": 3864
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.746724166306897e-07,
      "loss": 1.5085,
      "step": 3866
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.687085449307053e-07,
      "loss": 1.4309,
      "step": 3868
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.627641515308072e-07,
      "loss": 1.521,
      "step": 3870
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.56839249111201e-07,
      "loss": 1.5306,
      "step": 3872
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.509338503105224e-07,
      "loss": 1.4273,
      "step": 3874
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.450479677257962e-07,
      "loss": 1.4743,
      "step": 3876
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.391816139124187e-07,
      "loss": 1.4535,
      "step": 3878
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.33334801384128e-07,
      "loss": 1.5456,
      "step": 3880
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.275075426129831e-07,
      "loss": 1.4466,
      "step": 3882
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.216998500293238e-07,
      "loss": 1.5223,
      "step": 3884
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.159117360217606e-07,
      "loss": 1.4852,
      "step": 3886
    },
    {
      "epoch": 3.51,
      "learning_rate": 8.101432129371345e-07,
      "loss": 1.5151,
      "step": 3888
    },
    {
      "epoch": 3.51,
      "learning_rate": 8.04394293080496e-07,
      "loss": 1.4739,
      "step": 3890
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.986649887150877e-07,
      "loss": 1.4808,
      "step": 3892
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.929553120622968e-07,
      "loss": 1.5439,
      "step": 3894
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.87265275301654e-07,
      "loss": 1.4876,
      "step": 3896
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.815948905707871e-07,
      "loss": 1.5406,
      "step": 3898
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.759441699654068e-07,
      "loss": 1.4724,
      "step": 3900
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.703131255392759e-07,
      "loss": 1.5031,
      "step": 3902
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.647017693041881e-07,
      "loss": 1.4945,
      "step": 3904
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.591101132299383e-07,
      "loss": 1.4353,
      "step": 3906
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.535381692442945e-07,
      "loss": 1.4728,
      "step": 3908
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.479859492329856e-07,
      "loss": 1.5694,
      "step": 3910
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.42453465039652e-07,
      "loss": 1.4905,
      "step": 3912
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.36940728465848e-07,
      "loss": 1.4811,
      "step": 3914
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.314477512709972e-07,
      "loss": 1.4571,
      "step": 3916
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.259745451723765e-07,
      "loss": 1.4766,
      "step": 3918
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.20521121845087e-07,
      "loss": 1.4502,
      "step": 3920
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.150874929220298e-07,
      "loss": 1.4518,
      "step": 3922
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.0967366999388e-07,
      "loss": 1.4984,
      "step": 3924
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.042796646090688e-07,
      "loss": 1.472,
      "step": 3926
    },
    {
      "epoch": 3.54,
      "learning_rate": 6.989054882737523e-07,
      "loss": 1.4784,
      "step": 3928
    },
    {
      "epoch": 3.54,
      "learning_rate": 6.935511524517835e-07,
      "loss": 1.4565,
      "step": 3930
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.88216668564704e-07,
      "loss": 1.4471,
      "step": 3932
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.829020479916936e-07,
      "loss": 1.4408,
      "step": 3934
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.776073020695728e-07,
      "loss": 1.4734,
      "step": 3936
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.723324420927601e-07,
      "loss": 1.4693,
      "step": 3938
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.670774793132606e-07,
      "loss": 1.4931,
      "step": 3940
    },
    {
      "epoch": 3.55,
      "learning_rate": 6.618424249406297e-07,
      "loss": 1.4914,
      "step": 3942
    },
    {
      "epoch": 3.56,
      "learning_rate": 6.566272901419579e-07,
      "loss": 1.4823,
      "step": 3944
    },
    {
      "epoch": 3.56,
      "learning_rate": 6.514320860418477e-07,
      "loss": 1.5082,
      "step": 3946
    },
    {
      "epoch": 3.56,
      "learning_rate": 6.462568237223787e-07,
      "loss": 1.5022,
      "step": 3948
    },
    {
      "epoch": 3.56,
      "learning_rate": 6.411015142231025e-07,
      "loss": 1.4829,
      "step": 3950
    },
    {
      "epoch": 3.56,
      "learning_rate": 6.35966168540998e-07,
      "loss": 1.5037,
      "step": 3952
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.308507976304701e-07,
      "loss": 1.5403,
      "step": 3954
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.257554124033005e-07,
      "loss": 1.4079,
      "step": 3956
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.206800237286514e-07,
      "loss": 1.5042,
      "step": 3958
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.156246424330215e-07,
      "loss": 1.458,
      "step": 3960
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.105892793002377e-07,
      "loss": 1.4147,
      "step": 3962
    },
    {
      "epoch": 3.57,
      "learning_rate": 6.055739450714182e-07,
      "loss": 1.4517,
      "step": 3964
    },
    {
      "epoch": 3.58,
      "learning_rate": 6.005786504449651e-07,
      "loss": 1.4733,
      "step": 3966
    },
    {
      "epoch": 3.58,
      "learning_rate": 5.956034060765248e-07,
      "loss": 1.395,
      "step": 3968
    },
    {
      "epoch": 3.58,
      "learning_rate": 5.906482225789767e-07,
      "loss": 1.4927,
      "step": 3970
    },
    {
      "epoch": 3.58,
      "learning_rate": 5.857131105224123e-07,
      "loss": 1.4974,
      "step": 3972
    },
    {
      "epoch": 3.58,
      "learning_rate": 5.807980804341007e-07,
      "loss": 1.5244,
      "step": 3974
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.759031427984818e-07,
      "loss": 1.4225,
      "step": 3976
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.710283080571233e-07,
      "loss": 1.4604,
      "step": 3978
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.661735866087237e-07,
      "loss": 1.4633,
      "step": 3980
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.61338988809067e-07,
      "loss": 1.4687,
      "step": 3982
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.565245249710194e-07,
      "loss": 1.4547,
      "step": 3984
    },
    {
      "epoch": 3.59,
      "learning_rate": 5.517302053644902e-07,
      "loss": 1.5225,
      "step": 3986
    },
    {
      "epoch": 3.6,
      "learning_rate": 5.469560402164254e-07,
      "loss": 1.4871,
      "step": 3988
    },
    {
      "epoch": 3.6,
      "learning_rate": 5.422020397107753e-07,
      "loss": 1.4344,
      "step": 3990
    },
    {
      "epoch": 3.6,
      "learning_rate": 5.374682139884735e-07,
      "loss": 1.5054,
      "step": 3992
    },
    {
      "epoch": 3.6,
      "learning_rate": 5.327545731474237e-07,
      "loss": 1.4971,
      "step": 3994
    },
    {
      "epoch": 3.6,
      "learning_rate": 5.280611272424696e-07,
      "loss": 1.5025,
      "step": 3996
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.23387886285377e-07,
      "loss": 1.4235,
      "step": 3998
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.187348602448106e-07,
      "loss": 1.6073,
      "step": 4000
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.141020590463142e-07,
      "loss": 1.466,
      "step": 4002
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.094894925722893e-07,
      "loss": 1.4511,
      "step": 4004
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.048971706619777e-07,
      "loss": 1.5318,
      "step": 4006
    },
    {
      "epoch": 3.61,
      "learning_rate": 5.003251031114287e-07,
      "loss": 1.4607,
      "step": 4008
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.957732996734943e-07,
      "loss": 1.4208,
      "step": 4010
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.912417700577965e-07,
      "loss": 1.4769,
      "step": 4012
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.867305239307096e-07,
      "loss": 1.4936,
      "step": 4014
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.822395709153426e-07,
      "loss": 1.5077,
      "step": 4016
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.77768920591516e-07,
      "loss": 1.5571,
      "step": 4018
    },
    {
      "epoch": 3.62,
      "learning_rate": 4.73318582495742e-07,
      "loss": 1.4608,
      "step": 4020
    },
    {
      "epoch": 3.63,
      "learning_rate": 4.688885661212039e-07,
      "loss": 1.5269,
      "step": 4022
    },
    {
      "epoch": 3.63,
      "learning_rate": 4.644788809177336e-07,
      "loss": 1.4387,
      "step": 4024
    },
    {
      "epoch": 3.63,
      "learning_rate": 4.6008953629179676e-07,
      "loss": 1.5016,
      "step": 4026
    },
    {
      "epoch": 3.63,
      "learning_rate": 4.5572054160647025e-07,
      "loss": 1.4119,
      "step": 4028
    },
    {
      "epoch": 3.63,
      "learning_rate": 4.5137190618141813e-07,
      "loss": 1.479,
      "step": 4030
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.470436392928812e-07,
      "loss": 1.474,
      "step": 4032
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.42735750173644e-07,
      "loss": 1.491,
      "step": 4034
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.384482480130281e-07,
      "loss": 1.466,
      "step": 4036
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.341811419568653e-07,
      "loss": 1.4711,
      "step": 4038
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.2993444110747795e-07,
      "loss": 1.4969,
      "step": 4040
    },
    {
      "epoch": 3.64,
      "learning_rate": 4.257081545236641e-07,
      "loss": 1.4628,
      "step": 4042
    },
    {
      "epoch": 3.65,
      "learning_rate": 4.2150229122067565e-07,
      "loss": 1.4757,
      "step": 4044
    },
    {
      "epoch": 3.65,
      "learning_rate": 4.1731686017019377e-07,
      "loss": 1.4377,
      "step": 4046
    },
    {
      "epoch": 3.65,
      "learning_rate": 4.131518703003201e-07,
      "loss": 1.3858,
      "step": 4048
    },
    {
      "epoch": 3.65,
      "learning_rate": 4.090073304955511e-07,
      "loss": 1.4229,
      "step": 4050
    },
    {
      "epoch": 3.65,
      "learning_rate": 4.0488324959675805e-07,
      "loss": 1.4659,
      "step": 4052
    },
    {
      "epoch": 3.66,
      "learning_rate": 4.007796364011762e-07,
      "loss": 1.3815,
      "step": 4054
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.9669649966237347e-07,
      "loss": 1.4247,
      "step": 4056
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.9263384809024143e-07,
      "loss": 1.4442,
      "step": 4058
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.885916903509779e-07,
      "loss": 1.4681,
      "step": 4060
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.845700350670567e-07,
      "loss": 1.5604,
      "step": 4062
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.8056889081722556e-07,
      "loss": 1.4922,
      "step": 4064
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.765882661364739e-07,
      "loss": 1.4546,
      "step": 4066
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.7262816951602057e-07,
      "loss": 1.4963,
      "step": 4068
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.68688609403296e-07,
      "loss": 1.4392,
      "step": 4070
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.6476959420192473e-07,
      "loss": 1.5181,
      "step": 4072
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.6087113227170287e-07,
      "loss": 1.5524,
      "step": 4074
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.569932319285885e-07,
      "loss": 1.4297,
      "step": 4076
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.5313590144467446e-07,
      "loss": 1.49,
      "step": 4078
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.4929914904817677e-07,
      "loss": 1.4437,
      "step": 4080
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.4548298292341744e-07,
      "loss": 1.501,
      "step": 4082
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.4168741121080153e-07,
      "loss": 1.5746,
      "step": 4084
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.379124420068081e-07,
      "loss": 1.4168,
      "step": 4086
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.3415808336396127e-07,
      "loss": 1.4926,
      "step": 4088
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.304243432908294e-07,
      "loss": 1.4883,
      "step": 4090
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.267112297519881e-07,
      "loss": 1.4488,
      "step": 4092
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.230187506680227e-07,
      "loss": 1.4376,
      "step": 4094
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.193469139154959e-07,
      "loss": 1.4768,
      "step": 4096
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.156957273269434e-07,
      "loss": 1.4695,
      "step": 4098
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.120651986908474e-07,
      "loss": 1.4708,
      "step": 4100
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.084553357516218e-07,
      "loss": 1.4751,
      "step": 4102
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.0486614620960476e-07,
      "loss": 1.5633,
      "step": 4104
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.012976377210275e-07,
      "loss": 1.4757,
      "step": 4106
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.9774981789801205e-07,
      "loss": 1.4903,
      "step": 4108
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.942226943085424e-07,
      "loss": 1.4601,
      "step": 4110
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.9071627447646353e-07,
      "loss": 1.4546,
      "step": 4112
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.872305658814445e-07,
      "loss": 1.4475,
      "step": 4114
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.8376557595898635e-07,
      "loss": 1.5515,
      "step": 4116
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.803213121003856e-07,
      "loss": 1.4752,
      "step": 4118
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.7689778165273184e-07,
      "loss": 1.4019,
      "step": 4120
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.7349499191888675e-07,
      "loss": 1.4234,
      "step": 4122
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.701129501574673e-07,
      "loss": 1.4638,
      "step": 4124
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.667516635828349e-07,
      "loss": 1.4352,
      "step": 4126
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.634111393650751e-07,
      "loss": 1.3558,
      "step": 4128
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.600913846299857e-07,
      "loss": 1.5526,
      "step": 4130
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.567924064590599e-07,
      "loss": 1.5054,
      "step": 4132
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.5351421188947287e-07,
      "loss": 1.4943,
      "step": 4134
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.50256807914061e-07,
      "loss": 1.4276,
      "step": 4136
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.4702020148131836e-07,
      "loss": 1.4076,
      "step": 4138
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.438043994953687e-07,
      "loss": 1.4831,
      "step": 4140
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.406094088159616e-07,
      "loss": 1.5224,
      "step": 4142
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.374352362584509e-07,
      "loss": 1.4096,
      "step": 4144
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.3428188859378253e-07,
      "loss": 1.4154,
      "step": 4146
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.3114937254848258e-07,
      "loss": 1.5115,
      "step": 4148
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.280376948046359e-07,
      "loss": 1.3861,
      "step": 4150
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.2494686199988069e-07,
      "loss": 1.4862,
      "step": 4152
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.2187688072738632e-07,
      "loss": 1.4352,
      "step": 4154
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.1882775753584884e-07,
      "loss": 1.5093,
      "step": 4156
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.1579949892946206e-07,
      "loss": 1.4904,
      "step": 4158
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.127921113679221e-07,
      "loss": 1.5107,
      "step": 4160
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.1129624623544843e-07,
      "loss": 1.3679,
      "step": 4162
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.083201772557053e-07,
      "loss": 1.4715,
      "step": 4164
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.0536499527522792e-07,
      "loss": 1.4586,
      "step": 4166
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.0243070659782705e-07,
      "loss": 1.4954,
      "step": 4168
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.9951731748274473e-07,
      "loss": 1.449,
      "step": 4170
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.966248341446375e-07,
      "loss": 1.4139,
      "step": 4172
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.937532627535721e-07,
      "loss": 1.4097,
      "step": 4174
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.9090260943500416e-07,
      "loss": 1.4908,
      "step": 4176
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.8807288026977066e-07,
      "loss": 1.5407,
      "step": 4178
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.8526408129407093e-07,
      "loss": 1.5268,
      "step": 4180
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.8247621849946107e-07,
      "loss": 1.4683,
      "step": 4182
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.7970929783283408e-07,
      "loss": 1.4642,
      "step": 4184
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.7696332519641313e-07,
      "loss": 1.4373,
      "step": 4186
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.742383064477371e-07,
      "loss": 1.48,
      "step": 4188
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.7153424739964174e-07,
      "loss": 1.4326,
      "step": 4190
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.6885115382026084e-07,
      "loss": 1.4323,
      "step": 4192
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.6618903143299836e-07,
      "loss": 1.4751,
      "step": 4194
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.635478859165296e-07,
      "loss": 1.4597,
      "step": 4196
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.609277229047801e-07,
      "loss": 1.5044,
      "step": 4198
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.5832854798691788e-07,
      "loss": 1.4625,
      "step": 4200
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.5575036670734123e-07,
      "loss": 1.4762,
      "step": 4202
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.5319318456566424e-07,
      "loss": 1.5121,
      "step": 4204
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.5065700701670793e-07,
      "loss": 1.5383,
      "step": 4206
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.481418394704881e-07,
      "loss": 1.5195,
      "step": 4208
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.4564768729220414e-07,
      "loss": 1.5122,
      "step": 4210
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.431745558022235e-07,
      "loss": 1.4437,
      "step": 4212
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.4072245027607734e-07,
      "loss": 1.5162,
      "step": 4214
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.3829137594444154e-07,
      "loss": 1.4974,
      "step": 4216
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.358813379931345e-07,
      "loss": 1.4813,
      "step": 4218
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.3349234156309732e-07,
      "loss": 1.4017,
      "step": 4220
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.3112439175038794e-07,
      "loss": 1.3822,
      "step": 4222
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.287774936061692e-07,
      "loss": 1.4827,
      "step": 4224
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2645165213669651e-07,
      "loss": 1.4623,
      "step": 4226
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2414687230331124e-07,
      "loss": 1.5188,
      "step": 4228
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2186315902242285e-07,
      "loss": 1.4864,
      "step": 4230
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.1960051716550903e-07,
      "loss": 1.5138,
      "step": 4232
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.1735895155909338e-07,
      "loss": 1.427,
      "step": 4234
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.1513846698474551e-07,
      "loss": 1.405,
      "step": 4236
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.1293906817906209e-07,
      "loss": 1.5173,
      "step": 4238
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.1076075983366574e-07,
      "loss": 1.448,
      "step": 4240
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.0860354659518512e-07,
      "loss": 1.4635,
      "step": 4242
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.0646743306525376e-07,
      "loss": 1.4324,
      "step": 4244
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.043524238004956e-07,
      "loss": 1.3653,
      "step": 4246
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.0225852331251729e-07,
      "loss": 1.4727,
      "step": 4248
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.0018573606789483e-07,
      "loss": 1.446,
      "step": 4250
    },
    {
      "epoch": 3.83,
      "learning_rate": 9.813406648816804e-08,
      "loss": 1.4742,
      "step": 4252
    },
    {
      "epoch": 3.84,
      "learning_rate": 9.610351894983162e-08,
      "loss": 1.4897,
      "step": 4254
    },
    {
      "epoch": 3.84,
      "learning_rate": 9.409409778432189e-08,
      "loss": 1.5121,
      "step": 4256
    },
    {
      "epoch": 3.84,
      "learning_rate": 9.21058072780101e-08,
      "loss": 1.4643,
      "step": 4258
    },
    {
      "epoch": 3.84,
      "learning_rate": 9.013865167219471e-08,
      "loss": 1.4174,
      "step": 4260
    },
    {
      "epoch": 3.84,
      "learning_rate": 8.819263516308796e-08,
      "loss": 1.5006,
      "step": 4262
    },
    {
      "epoch": 3.84,
      "learning_rate": 8.62677619018104e-08,
      "loss": 1.4582,
      "step": 4264
    },
    {
      "epoch": 3.85,
      "learning_rate": 8.436403599438203e-08,
      "loss": 1.5086,
      "step": 4266
    },
    {
      "epoch": 3.85,
      "learning_rate": 8.24814615017111e-08,
      "loss": 1.4035,
      "step": 4268
    },
    {
      "epoch": 3.85,
      "learning_rate": 8.062004243958866e-08,
      "loss": 1.4,
      "step": 4270
    },
    {
      "epoch": 3.85,
      "learning_rate": 7.877978277867737e-08,
      "loss": 1.4812,
      "step": 4272
    },
    {
      "epoch": 3.85,
      "learning_rate": 7.696068644450383e-08,
      "loss": 1.4961,
      "step": 4274
    },
    {
      "epoch": 3.86,
      "learning_rate": 7.51627573174507e-08,
      "loss": 1.4979,
      "step": 4276
    },
    {
      "epoch": 3.86,
      "learning_rate": 7.338599923274791e-08,
      "loss": 1.437,
      "step": 4278
    },
    {
      "epoch": 3.86,
      "learning_rate": 7.163041598046483e-08,
      "loss": 1.4799,
      "step": 4280
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.98960113055025e-08,
      "loss": 1.5305,
      "step": 4282
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.818278890758367e-08,
      "loss": 1.5565,
      "step": 4284
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.649075244124614e-08,
      "loss": 1.4055,
      "step": 4286
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.481990551583939e-08,
      "loss": 1.4872,
      "step": 4288
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.317025169550684e-08,
      "loss": 1.514,
      "step": 4290
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.154179449918918e-08,
      "loss": 1.4633,
      "step": 4292
    },
    {
      "epoch": 3.87,
      "learning_rate": 5.993453740060773e-08,
      "loss": 1.4542,
      "step": 4294
    },
    {
      "epoch": 3.87,
      "learning_rate": 5.8348483828263305e-08,
      "loss": 1.4561,
      "step": 4296
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.678363716542623e-08,
      "loss": 1.4949,
      "step": 4298
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.5240000750129695e-08,
      "loss": 1.3946,
      "step": 4300
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.3717577875161964e-08,
      "loss": 1.5279,
      "step": 4302
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.2216371788061936e-08,
      "loss": 1.4138,
      "step": 4304
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.0736385691106946e-08,
      "loss": 1.4473,
      "step": 4306
    },
    {
      "epoch": 3.88,
      "learning_rate": 4.9277622741310535e-08,
      "loss": 1.5867,
      "step": 4308
    },
    {
      "epoch": 3.89,
      "learning_rate": 4.78400860504169e-08,
      "loss": 1.4765,
      "step": 4310
    },
    {
      "epoch": 3.89,
      "learning_rate": 4.642377868488646e-08,
      "loss": 1.5305,
      "step": 4312
    },
    {
      "epoch": 3.89,
      "learning_rate": 4.5028703665900286e-08,
      "loss": 1.5238,
      "step": 4314
    },
    {
      "epoch": 3.89,
      "learning_rate": 4.365486396934349e-08,
      "loss": 1.5024,
      "step": 4316
    },
    {
      "epoch": 3.89,
      "learning_rate": 4.230226252580516e-08,
      "loss": 1.4614,
      "step": 4318
    },
    {
      "epoch": 3.9,
      "learning_rate": 4.0970902220570654e-08,
      "loss": 1.4628,
      "step": 4320
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.9660785893614883e-08,
      "loss": 1.4929,
      "step": 4322
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.837191633959458e-08,
      "loss": 1.4142,
      "step": 4324
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.710429630784718e-08,
      "loss": 1.4489,
      "step": 4326
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.585792850238079e-08,
      "loss": 1.4397,
      "step": 4328
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.4632815581866484e-08,
      "loss": 1.5186,
      "step": 4330
    },
    {
      "epoch": 3.91,
      "learning_rate": 3.3428960159642697e-08,
      "loss": 1.3846,
      "step": 4332
    },
    {
      "epoch": 3.91,
      "learning_rate": 3.2246364803695254e-08,
      "loss": 1.4404,
      "step": 4334
    },
    {
      "epoch": 3.91,
      "learning_rate": 3.108503203666402e-08,
      "loss": 1.4731,
      "step": 4336
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.994496433583072e-08,
      "loss": 1.5085,
      "step": 4338
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.8826164133117784e-08,
      "loss": 1.4772,
      "step": 4340
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.7728633815079508e-08,
      "loss": 1.47,
      "step": 4342
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.6652375722897582e-08,
      "loss": 1.5466,
      "step": 4344
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5597392152379996e-08,
      "loss": 1.4395,
      "step": 4346
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.456368535394993e-08,
      "loss": 1.4466,
      "step": 4348
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.3551257532649085e-08,
      "loss": 1.5694,
      "step": 4350
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.256011084812326e-08,
      "loss": 1.4094,
      "step": 4352
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.1590247414624566e-08,
      "loss": 1.3878,
      "step": 4354
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.0641669301004752e-08,
      "loss": 1.4059,
      "step": 4356
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.9714378530711896e-08,
      "loss": 1.4624,
      "step": 4358
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.8808377081785954e-08,
      "loss": 1.5119,
      "step": 4360
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.7923666886849877e-08,
      "loss": 1.5316,
      "step": 4362
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.706024983311405e-08,
      "loss": 1.4541,
      "step": 4364
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.62181277623652e-08,
      "loss": 1.4447,
      "step": 4366
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.539730247096638e-08,
      "loss": 1.5013,
      "step": 4368
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.4597775709849215e-08,
      "loss": 1.4457,
      "step": 4370
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.3819549184516113e-08,
      "loss": 1.4869,
      "step": 4372
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.3062624555029157e-08,
      "loss": 1.4717,
      "step": 4374
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.2327003436013452e-08,
      "loss": 1.4999,
      "step": 4376
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.1612687396650445e-08,
      "loss": 1.3948,
      "step": 4378
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.0919677960674612e-08,
      "loss": 1.5142,
      "step": 4380
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.0247976606369004e-08,
      "loss": 1.4844,
      "step": 4382
    },
    {
      "epoch": 3.95,
      "learning_rate": 9.59758476656636e-09,
      "loss": 1.4479,
      "step": 4384
    },
    {
      "epoch": 3.95,
      "learning_rate": 8.96850382864134e-09,
      "loss": 1.4565,
      "step": 4386
    },
    {
      "epoch": 3.96,
      "learning_rate": 8.360735134510522e-09,
      "loss": 1.5029,
      "step": 4388
    },
    {
      "epoch": 3.96,
      "learning_rate": 7.774279980626853e-09,
      "loss": 1.4039,
      "step": 4390
    },
    {
      "epoch": 3.96,
      "learning_rate": 7.2091396179807534e-09,
      "loss": 1.4211,
      "step": 4392
    },
    {
      "epoch": 3.96,
      "learning_rate": 6.665315252094573e-09,
      "loss": 1.484,
      "step": 4394
    },
    {
      "epoch": 3.96,
      "learning_rate": 6.142808043020365e-09,
      "loss": 1.5201,
      "step": 4396
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.64161910533767e-09,
      "loss": 1.502,
      "step": 4398
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.161749508150182e-09,
      "loss": 1.5364,
      "step": 4400
    },
    {
      "epoch": 3.97,
      "learning_rate": 4.703200275087971e-09,
      "loss": 1.4239,
      "step": 4402
    },
    {
      "epoch": 3.97,
      "learning_rate": 4.26597238429749e-09,
      "loss": 1.3793,
      "step": 4404
    },
    {
      "epoch": 3.97,
      "learning_rate": 3.850066768448235e-09,
      "loss": 1.564,
      "step": 4406
    },
    {
      "epoch": 3.97,
      "learning_rate": 3.4554843147216467e-09,
      "loss": 1.5356,
      "step": 4408
    },
    {
      "epoch": 3.98,
      "learning_rate": 3.0822258648177673e-09,
      "loss": 1.5149,
      "step": 4410
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.730292214947472e-09,
      "loss": 1.4139,
      "step": 4412
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.3996841158346885e-09,
      "loss": 1.5084,
      "step": 4414
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.090402272710845e-09,
      "loss": 1.4369,
      "step": 4416
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.8024473453170933e-09,
      "loss": 1.5157,
      "step": 4418
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.535819947900974e-09,
      "loss": 1.5208,
      "step": 4420
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.29052064921531e-09,
      "loss": 1.4772,
      "step": 4422
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.0665499725170948e-09,
      "loss": 1.4908,
      "step": 4424
    },
    {
      "epoch": 3.99,
      "learning_rate": 8.639083955663819e-10,
      "loss": 1.4492,
      "step": 4426
    },
    {
      "epoch": 3.99,
      "learning_rate": 6.825963506262856e-10,
      "loss": 1.4351,
      "step": 4428
    },
    {
      "epoch": 3.99,
      "learning_rate": 5.226142244585397e-10,
      "loss": 1.4981,
      "step": 4430
    },
    {
      "epoch": 4.0,
      "learning_rate": 3.839623583301588e-10,
      "loss": 1.4821,
      "step": 4432
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.6664104800122604e-10,
      "loss": 1.4517,
      "step": 4434
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.7065054373710533e-10,
      "loss": 1.4852,
      "step": 4436
    },
    {
      "epoch": 4.0,
      "step": 4436,
      "total_flos": 1.1913558786415002e+17,
      "train_loss": 1.5918720781641893,
      "train_runtime": 35147.4688,
      "train_samples_per_second": 8.077,
      "train_steps_per_second": 0.126
    }
  ],
  "logging_steps": 2,
  "max_steps": 4436,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 200,
  "total_flos": 1.1913558786415002e+17,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}