{ "epoch": 4.0, "eval_loss": 2.2968251705169678, "eval_runtime": 40.9962, "eval_samples_per_second": 203.629, "eval_steps_per_second": 1.61, "perplexity": 9.942566340729925, "total_flos": 490890602741760.0, "train_loss": 2.2402087043862937, "train_runtime": 7138.091, "train_samples_per_second": 42.101, "train_steps_per_second": 0.658 }