run-22/checkpoint-288/trainer_state.json

{
  "best_metric": 0.8072407045009785,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-22/checkpoint-192",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 288,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.634097933769226,
      "learning_rate": 0.0001003330154271938,
      "loss": 0.5664,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.773972602739726,
      "eval_f1": 0.8027327070879591,
      "eval_loss": 0.48113834857940674,
      "eval_precision": 0.7121212121212122,
      "eval_recall": 0.9197651663405088,
      "eval_runtime": 28.856,
      "eval_samples_per_second": 35.417,
      "eval_steps_per_second": 1.109,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.8999040126800537,
      "learning_rate": 5.01665077135969e-05,
      "loss": 0.4776,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8072407045009785,
      "eval_f1": 0.8226822682268229,
      "eval_loss": 0.4426460564136505,
      "eval_precision": 0.7616666666666667,
      "eval_recall": 0.8943248532289628,
      "eval_runtime": 28.4561,
      "eval_samples_per_second": 35.915,
      "eval_steps_per_second": 1.125,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.798825979232788,
      "learning_rate": 0.0,
      "loss": 0.4529,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8072407045009785,
      "eval_f1": 0.8261253309796999,
      "eval_loss": 0.43465864658355713,
      "eval_precision": 0.752411575562701,
      "eval_recall": 0.9158512720156555,
      "eval_runtime": 37.2794,
      "eval_samples_per_second": 27.415,
      "eval_steps_per_second": 0.858,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 288,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.8605867890458055,
    "learning_rate": 0.0001504995231407907,
    "num_train_epochs": 3,
    "temperature": 22
  }
}