checkpoint-500/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 8.403361344537815,
  "global_step": 500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 8.4,
      "learning_rate": 4.96e-05,
      "loss": 0.0899,
      "step": 500
    }
  ],
  "max_steps": 590,
  "num_train_epochs": 10,
  "total_flos": 1.4888535460528128e+16,
  "trial_name": null,
  "trial_params": null
}