trainer_state.json

{
  "best_metric": 0.5990289449691772,
  "best_model_checkpoint": "hBERTv2_new_pretrain_mrpc/checkpoint-87",
  "epoch": 8.0,
  "global_step": 232,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 3.9200000000000004e-05,
      "loss": 0.6721,
      "step": 29
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6838235294117647,
      "eval_combined_score": 0.7480253018237863,
      "eval_f1": 0.8122270742358079,
      "eval_loss": 0.6200118660926819,
      "eval_runtime": 0.7181,
      "eval_samples_per_second": 568.192,
      "eval_steps_per_second": 5.571,
      "step": 29
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.8400000000000005e-05,
      "loss": 0.6229,
      "step": 58
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6568627450980392,
      "eval_combined_score": 0.6911764705882353,
      "eval_f1": 0.7254901960784315,
      "eval_loss": 0.6098297834396362,
      "eval_runtime": 0.7062,
      "eval_samples_per_second": 577.745,
      "eval_steps_per_second": 5.664,
      "step": 58
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.76e-05,
      "loss": 0.5689,
      "step": 87
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7034313725490197,
      "eval_combined_score": 0.7576254840972159,
      "eval_f1": 0.8118195956454122,
      "eval_loss": 0.5990289449691772,
      "eval_runtime": 0.7016,
      "eval_samples_per_second": 581.511,
      "eval_steps_per_second": 5.701,
      "step": 87
    },
    {
      "epoch": 4.0,
      "learning_rate": 3.680000000000001e-05,
      "loss": 0.4615,
      "step": 116
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.6764705882352942,
      "eval_combined_score": 0.7282352941176471,
      "eval_f1": 0.78,
      "eval_loss": 0.6688693165779114,
      "eval_runtime": 0.7052,
      "eval_samples_per_second": 578.574,
      "eval_steps_per_second": 5.672,
      "step": 116
    },
    {
      "epoch": 5.0,
      "learning_rate": 3.6e-05,
      "loss": 0.3475,
      "step": 145
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.6053921568627451,
      "eval_combined_score": 0.6413734331407914,
      "eval_f1": 0.6773547094188378,
      "eval_loss": 0.8471598029136658,
      "eval_runtime": 0.705,
      "eval_samples_per_second": 578.745,
      "eval_steps_per_second": 5.674,
      "step": 145
    },
    {
      "epoch": 6.0,
      "learning_rate": 3.52e-05,
      "loss": 0.2307,
      "step": 174
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.6102941176470589,
      "eval_combined_score": 0.6507781267846945,
      "eval_f1": 0.6912621359223301,
      "eval_loss": 0.991661548614502,
      "eval_runtime": 0.7062,
      "eval_samples_per_second": 577.727,
      "eval_steps_per_second": 5.664,
      "step": 174
    },
    {
      "epoch": 7.0,
      "learning_rate": 3.44e-05,
      "loss": 0.166,
      "step": 203
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.6544117647058824,
      "eval_combined_score": 0.7033043006306213,
      "eval_f1": 0.7521968365553603,
      "eval_loss": 1.1149414777755737,
      "eval_runtime": 0.7071,
      "eval_samples_per_second": 576.986,
      "eval_steps_per_second": 5.657,
      "step": 203
    },
    {
      "epoch": 8.0,
      "learning_rate": 3.3600000000000004e-05,
      "loss": 0.1258,
      "step": 232
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.625,
      "eval_combined_score": 0.6684322033898306,
      "eval_f1": 0.7118644067796611,
      "eval_loss": 1.3515982627868652,
      "eval_runtime": 0.7061,
      "eval_samples_per_second": 577.831,
      "eval_steps_per_second": 5.665,
      "step": 232
    },
    {
      "epoch": 8.0,
      "step": 232,
      "total_flos": 4294176753582080.0,
      "train_loss": 0.3994215819342383,
      "train_runtime": 199.797,
      "train_samples_per_second": 917.932,
      "train_steps_per_second": 7.257
    }
  ],
  "max_steps": 1450,
  "num_train_epochs": 50,
  "total_flos": 4294176753582080.0,
  "trial_name": null,
  "trial_params": null
}