{ "epoch": 1.0, "eval_loss": 1.54607093334198, "eval_runtime": 30.0856, "eval_samples_per_second": 9.539, "eval_steps_per_second": 1.197, "perplexity": 4.692994821416736, "total_flos": 3.521821556736e+16, "train_loss": 1.7982163059427918, "train_runtime": 803.097, "train_samples_per_second": 4.815, "train_steps_per_second": 2.408 }