{ "epoch": 400.0, "eval_loss": 4.94140625, "eval_runtime": 3.7873, "eval_samples": 258, "eval_samples_per_second": 68.123, "eval_steps_per_second": 8.713, "perplexity": 139.96693973881514, "train_loss": 0.4328666927939967, "train_runtime": 22235.2418, "train_samples": 603, "train_samples_per_second": 10.848, "train_steps_per_second": 0.684 }