trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 806,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.2406947890818858,
      "grad_norm": 4.517418384552002,
      "learning_rate": 1.5037220843672458e-05,
      "loss": 0.4706,
      "step": 500
    }
  ],
  "logging_steps": 500,
  "max_steps": 2015,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 2541938728670208.0,
  "train_batch_size": 48,
  "trial_name": null,
  "trial_params": null
}