{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.061835151351194675,
  "global_step": 400,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 9.999226963512678e-06,
      "loss": 1.1361,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.998453927025357e-06,
      "loss": 0.7353,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.997680890538034e-06,
      "loss": 0.6909,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.996907854050712e-06,
      "loss": 0.6498,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.99613481756339e-06,
      "loss": 0.6414,
      "step": 25
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.995361781076068e-06,
      "loss": 0.6415,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.994588744588745e-06,
      "loss": 0.6317,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.993815708101423e-06,
      "loss": 0.6378,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.9930426716141e-06,
      "loss": 0.6347,
      "step": 45
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.992269635126779e-06,
      "loss": 0.5924,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.991496598639456e-06,
      "loss": 0.6046,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.990723562152135e-06,
      "loss": 0.6045,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.989950525664813e-06,
      "loss": 0.6,
      "step": 65
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.98917748917749e-06,
      "loss": 0.5504,
      "step": 70
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.988404452690169e-06,
      "loss": 0.5747,
      "step": 75
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.987631416202846e-06,
      "loss": 0.5526,
      "step": 80
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.986858379715523e-06,
      "loss": 0.5958,
      "step": 85
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.9860853432282e-06,
      "loss": 0.608,
      "step": 90
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.985312306740878e-06,
      "loss": 0.5988,
      "step": 95
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.984539270253557e-06,
      "loss": 0.5861,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.983766233766234e-06,
      "loss": 0.5749,
      "step": 105
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.982993197278913e-06,
      "loss": 0.5498,
      "step": 110
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.98222016079159e-06,
      "loss": 0.5841,
      "step": 115
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.981447124304268e-06,
      "loss": 0.5973,
      "step": 120
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.980674087816947e-06,
      "loss": 0.5954,
      "step": 125
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.979901051329624e-06,
      "loss": 0.527,
      "step": 130
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.979128014842301e-06,
      "loss": 0.5321,
      "step": 135
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.978354978354979e-06,
      "loss": 0.5781,
      "step": 140
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.977581941867656e-06,
      "loss": 0.5119,
      "step": 145
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.976808905380335e-06,
      "loss": 0.5271,
      "step": 150
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.976035868893012e-06,
      "loss": 0.5814,
      "step": 155
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.97526283240569e-06,
      "loss": 0.518,
      "step": 160
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.974489795918369e-06,
      "loss": 0.5335,
      "step": 165
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.973716759431046e-06,
      "loss": 0.5062,
      "step": 170
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.972943722943725e-06,
      "loss": 0.5253,
      "step": 175
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.972170686456402e-06,
      "loss": 0.5856,
      "step": 180
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.97139764996908e-06,
      "loss": 0.5196,
      "step": 185
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.970624613481757e-06,
      "loss": 0.4764,
      "step": 190
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.969851576994434e-06,
      "loss": 0.5254,
      "step": 195
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.969078540507111e-06,
      "loss": 0.5442,
      "step": 200
    },
    {
      "epoch": 0.03,
      "eval_accuracy": 0.5804400673190799,
      "eval_accuracy_sklearn": 0.5804400673190799,
      "eval_f1": 0.5294915349019279,
      "eval_loss": 0.7918509840965271,
      "eval_precision": 0.6370946036872561,
      "eval_recall": 0.45298409281186464,
      "eval_runtime": 4914.2737,
      "eval_samples_per_second": 16.323,
      "eval_steps_per_second": 2.04,
      "step": 200
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.96830550401979e-06,
      "loss": 0.5163,
      "step": 205
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.967532467532468e-06,
      "loss": 0.5044,
      "step": 210
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.966759431045147e-06,
      "loss": 0.5078,
      "step": 215
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.965986394557824e-06,
      "loss": 0.4623,
      "step": 220
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.965213358070501e-06,
      "loss": 0.5359,
      "step": 225
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.96444032158318e-06,
      "loss": 0.5068,
      "step": 230
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.963667285095858e-06,
      "loss": 0.5029,
      "step": 235
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.962894248608535e-06,
      "loss": 0.5084,
      "step": 240
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.962121212121212e-06,
      "loss": 0.4783,
      "step": 245
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.96134817563389e-06,
      "loss": 0.5216,
      "step": 250
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.960575139146569e-06,
      "loss": 0.54,
      "step": 255
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.959802102659246e-06,
      "loss": 0.5494,
      "step": 260
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.959029066171925e-06,
      "loss": 0.5401,
      "step": 265
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.958256029684602e-06,
      "loss": 0.5073,
      "step": 270
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.95748299319728e-06,
      "loss": 0.4598,
      "step": 275
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.956709956709958e-06,
      "loss": 0.4913,
      "step": 280
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.955936920222636e-06,
      "loss": 0.4947,
      "step": 285
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.955163883735313e-06,
      "loss": 0.4806,
      "step": 290
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.95439084724799e-06,
      "loss": 0.4659,
      "step": 295
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.953617810760668e-06,
      "loss": 0.4555,
      "step": 300
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.952844774273347e-06,
      "loss": 0.4606,
      "step": 305
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.952071737786024e-06,
      "loss": 0.4905,
      "step": 310
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.951298701298701e-06,
      "loss": 0.4423,
      "step": 315
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.95052566481138e-06,
      "loss": 0.4855,
      "step": 320
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.949752628324058e-06,
      "loss": 0.486,
      "step": 325
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.948979591836737e-06,
      "loss": 0.4774,
      "step": 330
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.948206555349414e-06,
      "loss": 0.4909,
      "step": 335
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.947433518862091e-06,
      "loss": 0.47,
      "step": 340
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.946660482374768e-06,
      "loss": 0.4496,
      "step": 345
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.945887445887446e-06,
      "loss": 0.5146,
      "step": 350
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.945114409400125e-06,
      "loss": 0.4876,
      "step": 355
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.944341372912802e-06,
      "loss": 0.4747,
      "step": 360
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.94356833642548e-06,
      "loss": 0.4614,
      "step": 365
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.942795299938158e-06,
      "loss": 0.4755,
      "step": 370
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.942022263450836e-06,
      "loss": 0.4785,
      "step": 375
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.941249226963513e-06,
      "loss": 0.4581,
      "step": 380
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.940476190476192e-06,
      "loss": 0.4671,
      "step": 385
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.93970315398887e-06,
      "loss": 0.4327,
      "step": 390
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.938930117501547e-06,
      "loss": 0.4906,
      "step": 395
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.938157081014226e-06,
      "loss": 0.5006,
      "step": 400
    },
    {
      "epoch": 0.06,
      "eval_accuracy": 0.5519790562862308,
      "eval_accuracy_sklearn": 0.5519790562862308,
      "eval_f1": 0.39240549130993435,
      "eval_loss": 0.969095766544342,
      "eval_precision": 0.6691460531626593,
      "eval_recall": 0.27759837340031096,
      "eval_runtime": 4903.2377,
      "eval_samples_per_second": 16.36,
      "eval_steps_per_second": 2.045,
      "step": 400
    }
  ],
  "max_steps": 64680,
  "num_train_epochs": 10,
  "total_flos": 4.75071167594496e+16,
  "trial_name": null,
  "trial_params": null
}