{ "epoch": 1.0, "eval_loss": 1.9178457260131836, "eval_runtime": 80.5533, "eval_samples_per_second": 21.774, "eval_steps_per_second": 2.731, "perplexity": 6.806280072216437, "total_flos": 1.58472345255936e+17, "train_loss": 2.0172890398896968, "train_runtime": 2382.4555, "train_samples_per_second": 7.283, "train_steps_per_second": 3.642 }