{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 930,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.61,
      "learning_rate": 2.3118279569892472e-05,
      "loss": 2.4496,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 930,
      "total_flos": 1748052427112448.0,
      "train_loss": 2.0759948074176746,
      "train_runtime": 373.6283,
      "train_samples_per_second": 24.875,
      "train_steps_per_second": 2.489
    }
  ],
  "max_steps": 930,
  "num_train_epochs": 3,
  "total_flos": 1748052427112448.0,
  "trial_name": null,
  "trial_params": null
}