{
  "best_metric": 0.6666666666666666,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-0/checkpoint-214",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 214,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.3824027478694916,
      "learning_rate": 5.531898354978819e-05,
      "loss": 0.0835,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6666666666666666,
      "eval_f1": 0.0,
      "eval_loss": 0.06421080976724625,
      "eval_mcc": 0.0,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 3.1927,
      "eval_samples_per_second": 534.663,
      "eval_steps_per_second": 16.914,
      "step": 214
    }
  ],
  "logging_steps": 500,
  "max_steps": 2996,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 14,
  "save_steps": 500,
  "total_flos": 524775664440.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.08651897585698409,
    "learning_rate": 5.9574289976694975e-05,
    "num_train_epochs": 14,
    "temperature": 43
  }
}