train_results.json

{
    "epoch": 4.98793242156074,
    "num_input_tokens_seen": 5393616,
    "total_flos": 2.1382484588285133e+17,
    "train_loss": 0.5414434323177463,
    "train_runtime": 8640.8816,
    "train_samples_per_second": 11.503,
    "train_steps_per_second": 0.09
}