train_results.json

{
    "epoch": 20.217435897435898,
    "num_input_tokens_seen": 76105440,
    "total_flos": 1.6998510508071322e+18,
    "train_loss": 0.4908539823365289,
    "train_runtime": 36017.1103,
    "train_samples_per_second": 3.465,
    "train_steps_per_second": 0.216
}