trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 9.561304836895388,
  "global_step": 8500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 5.62,
      "learning_rate": 4.380630630630631e-05,
      "loss": 0.5361,
      "step": 5000
    }
  ],
  "max_steps": 8890,
  "num_train_epochs": 10,
  "total_flos": 1.2465519921620582e+17,
  "trial_name": null,
  "trial_params": null
}