trainer_state.json

{
  "best_metric": 3.131695032119751,
  "best_model_checkpoint": "output/the-sugarcubes/checkpoint-6",
  "epoch": 1.0,
  "global_step": 6,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.83,
      "learning_rate": 9.190657300387505e-06,
      "loss": 3.5074,
      "step": 5
    },
    {
      "epoch": 1.0,
      "eval_loss": 3.131695032119751,
      "eval_runtime": 0.3523,
      "eval_samples_per_second": 22.707,
      "eval_steps_per_second": 2.838,
      "step": 6
    }
  ],
  "max_steps": 6,
  "num_train_epochs": 1,
  "total_flos": 6009716736000.0,
  "trial_name": null,
  "trial_params": null
}