llama-3-8b-base-new-dpo-hh-helpful-s_star0.6-4xh200-batch-64-20260421-214335-rerun/train_results.json

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 0.9952153347312266,
    "train_runtime": 1856.6581,
    "train_samples": 43598,
    "train_samples_per_second": 23.482,
    "train_steps_per_second": 0.367
}