{
  "best_metric": 0.5,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-5/checkpoint-67",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 67,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.22246789932250977,
      "learning_rate": 1.5679734473172616e-05,
      "loss": 0.2458,
      "step": 67
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5,
      "eval_f1": 0.0,
      "eval_loss": 0.1889893114566803,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 28.3905,
      "eval_samples_per_second": 35.998,
      "eval_steps_per_second": 1.127,
      "step": 67
    }
  ],
  "logging_steps": 500,
  "max_steps": 670,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 235695197280.0,
  "train_batch_size": 46,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.19075569878013487,
    "learning_rate": 1.7421927192414017e-05,
    "num_train_epochs": 10,
    "per_device_train_batch_size": 46,
    "temperature": 24
  }
}