{ "epoch": 1.941952506596306, "eval_loss": 3.7669124603271484, "eval_runtime": 7.2704, "eval_samples_per_second": 92.705, "eval_steps_per_second": 11.691, "perplexity": 43.246333720958525, "total_flos": 1.8959322667863245e+17, "train_loss": 5.261244110439135, "train_runtime": 307.2859, "train_samples_per_second": 39.423, "train_steps_per_second": 0.15 }