train_results.json

{
    "epoch": 20.0,
    "num_input_tokens_seen": 42678144,
    "total_flos": 1.9217775351646126e+18,
    "train_loss": 0.2895805571795578,
    "train_runtime": 8479.718,
    "train_samples_per_second": 20.01,
    "train_steps_per_second": 5.003
}