trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 744,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.69,
      "learning_rate": 1.639784946236559e-05,
      "loss": 2.0688,
      "step": 500
    },
    {
      "epoch": 4.0,
      "step": 744,
      "total_flos": 341344235320704.0,
      "train_loss": 1.9904336006410661,
      "train_runtime": 3213.5193,
      "train_samples_per_second": 1.852,
      "train_steps_per_second": 0.232
    }
  ],
  "logging_steps": 500,
  "max_steps": 744,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 341344235320704.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}