train_results.json

{
    "effective_tokens_per_sec": 2866.8120123509916,
    "epoch": 3.0,
    "total_flos": 4.569263266953429e+18,
    "train_loss": 1.1290170093980534,
    "train_runtime": 13562.7006,
    "train_samples_per_second": 3.09,
    "train_steps_per_second": 0.193
}