trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "grad_norm": 6.395843029022217,
      "learning_rate": 1.2e-05,
      "loss": 0.6016,
      "step": 500
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.6221667528152466,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.2762,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1250,
      "total_flos": 2329828408320000.0,
      "train_loss": 0.3663978805541992,
      "train_runtime": 1977.8607,
      "train_samples_per_second": 5.056,
      "train_steps_per_second": 0.632
    }
  ],
  "logging_steps": 500,
  "max_steps": 1250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 2329828408320000.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}