{
  "best_metric": 0.6418591737747192,
  "best_model_checkpoint": "BERT-WMM/run-10/checkpoint-533",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 533,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.94,
      "grad_norm": 11.003177642822266,
      "learning_rate": 3.5116101047118276e-06,
      "loss": 0.7883,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7370892018779343,
      "eval_loss": 0.6418591737747192,
      "eval_runtime": 2.0865,
      "eval_samples_per_second": 1020.838,
      "eval_steps_per_second": 64.222,
      "step": 533
    }
  ],
  "logging_steps": 500,
  "max_steps": 3198,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 500,
  "total_flos": 359491040412336.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 4.162390331678438e-06,
    "num_train_epochs": 6,
    "per_device_train_batch_size": 16,
    "weight_decay": 0.000561727996440909
  }
}