train_results.json

{
    "epoch": 444.44444444444446,
    "num_input_tokens_seen": 11200800,
    "total_flos": 4.69048729503744e+16,
    "train_loss": 0.041804404937768734,
    "train_runtime": 14360.258,
    "train_samples_per_second": 11.142,
    "train_steps_per_second": 2.785
}