{
  "final_eval_loss": 2.395210620244344,
  "final_eval_ppl": 10.970508426932215,
  "num_samples": 6000
}