trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 96,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 4.0,
      "step": 96,
      "total_flos": 25084035072000.0,
      "train_loss": 3.8089685440063477,
      "train_runtime": 55.2428,
      "train_samples_per_second": 1.738,
      "train_steps_per_second": 1.738
    }
  ],
  "max_steps": 96,
  "num_train_epochs": 4,
  "total_flos": 25084035072000.0,
  "trial_name": null,
  "trial_params": null
}