trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 11.98816263510036,
  "global_step": 540,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 11.109795848344485,
      "learning_rate": 2.222222222222222e-06,
      "loss": 88.895875,
      "step": 500
    }
  ],
  "max_steps": 540,
  "num_train_epochs": 12,
  "total_flos": 16840514516156928
}