{ "epoch": 1.0, "eval_loss": 1.50249183177948, "eval_runtime": 41.0302, "eval_samples_per_second": 10.846, "eval_steps_per_second": 1.365, "perplexity": 4.492870611062817, "total_flos": 4.259518857216e+16, "train_loss": 1.7606283989458218, "train_runtime": 1228.9387, "train_samples_per_second": 3.806, "train_steps_per_second": 1.903 }