{
  "best_metric": 0.502970297029703,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-0/checkpoint-297",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 297,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.48113372921943665,
      "learning_rate": 1.1539601014308473e-05,
      "loss": 0.265,
      "step": 297
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.502970297029703,
      "eval_loss": 0.24101482331752777,
      "eval_runtime": 14.0847,
      "eval_samples_per_second": 35.854,
      "eval_steps_per_second": 1.136,
      "step": 297
    }
  ],
  "logging_steps": 500,
  "max_steps": 1485,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 777843283140.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.28450461852175,
    "learning_rate": 1.4424501267885591e-05,
    "num_train_epochs": 5,
    "temperature": 8
  }
}