{ "epoch": 4.0, "eval_loss": 2.4689557552337646, "eval_runtime": 80.8694, "eval_samples_per_second": 86.633, "eval_steps_per_second": 0.68, "perplexity": 11.810107767030958, "total_flos": 411954472550400.0, "train_loss": 2.318100369605283, "train_runtime": 14372.236, "train_samples_per_second": 17.546, "train_steps_per_second": 0.274 }