trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.487562189054726,
  "global_step": 500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.49,
      "learning_rate": 8.540630182421228e-06,
      "loss": 0.3771,
      "step": 500
    }
  ],
  "max_steps": 603,
  "num_train_epochs": 3,
  "total_flos": 85085068312800.0,
  "trial_name": null,
  "trial_params": null
}