train_results.json

{
    "effective_tokens_per_sec": 3032.858200894709,
    "epoch": 3.0,
    "total_flos": 3.1865440491043553e+18,
    "train_loss": 1.1440774658359985,
    "train_runtime": 18974.7516,
    "train_samples_per_second": 1.735,
    "train_steps_per_second": 0.108
}