{ "epoch": 0.4, "eval_loss": 0.7879278063774109, "eval_runtime": 2660.4642, "eval_samples": 5000, "eval_samples_per_second": 9.427, "eval_steps_per_second": 4.714, "perplexity": 2.198835289644634, "train_loss": 0.8401528125, "train_runtime": 72520.074, "train_samples": 100000, "train_samples_per_second": 2.758, "train_steps_per_second": 0.345 }