train_results.json

{
    "epoch": 6.612048192771084,
    "num_input_tokens_seen": 5052448,
    "total_flos": 2.28822660837802e+17,
    "train_loss": 0.2300173058541106,
    "train_runtime": 12675.9679,
    "train_samples_per_second": 0.655,
    "train_steps_per_second": 0.04
}