{ "epoch": 4.0, "eval_loss": 2.3333284854888916, "eval_runtime": 352.6998, "eval_samples_per_second": 205.14, "eval_steps_per_second": 1.605, "perplexity": 10.312208509221884, "total_flos": 4255641501696000.0, "train_loss": 2.2957492570853644, "train_runtime": 58446.3006, "train_samples_per_second": 44.566, "train_steps_per_second": 0.696 }