train_results.json

{
    "entropy": 0.07084919246179718,
    "epoch": 1.0,
    "mean_token_accuracy": 0.9790157548018864,
    "num_tokens": 17657223.0,
    "total_flos": 39414093316096.0,
    "train_loss": 0.1270361747862636,
    "train_runtime": 1974.4991,
    "train_samples": 6591,
    "train_samples_per_second": 3.338,
    "train_steps_per_second": 0.07
}