{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.3404825737265416,
  "eval_steps": 500,
  "global_step": 500,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7680187821388245,
      "eval_loss": 0.4622006416320801,
      "eval_runtime": 9.5056,
      "eval_samples_per_second": 313.816,
      "eval_steps_per_second": 19.673,
      "step": 373
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 21.834245681762695,
      "learning_rate": 3.516708684536568e-06,
      "loss": 0.5575,
      "step": 500
    }
  ],
  "logging_steps": 500,
  "max_steps": 3730,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 2813316543119808.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 4.0610908338456346e-06,
    "per_device_train_batch_size": 32
  }
}