llama-3-8b-base-new-dpo-hh-helpful-4xh200-batch-64-q_t-0.45-s_star-0.4-eta-1/all_results.json

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 1.1280472663705745,
    "train_runtime": 1739.8346,
    "train_samples": 43598,
    "train_samples_per_second": 25.059,
    "train_steps_per_second": 0.391
}