{
  "eval_loss": 6.64558858620493,
  "perplexity": 769.3827624683506,
  "eval_samples": 76
}