train_results.json

{
    "all_params": 153894144,
    "epoch": 3.0,
    "memory_footprint": 403239936,
    "total_flos": 2506179136462848.0,
    "train_loss": 1.8787952170227513,
    "train_runtime": 633.4063,
    "train_samples_per_second": 4.163,
    "train_steps_per_second": 0.521,
    "trainable_params": 1769472,
    "trainable_params_percent": 1.1497981365684713
}