{ "epoch": 4.0, "eval_loss": 1.2184005975723267, "eval_runtime": 87.9641, "eval_samples_per_second": 18.951, "eval_steps_per_second": 0.307, "perplexity": 3.3817745873907485, "total_flos": 391958715432960.0, "train_loss": 1.1227505109203395, "train_runtime": 12501.281, "train_samples_per_second": 4.798, "train_steps_per_second": 0.15 }