llama-3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.5-s_star-1.0/train_results.json

{
    "epoch": 0.999244142101285,
    "total_flos": 0.0,
    "train_loss": 1.1173522615757363,
    "train_runtime": 1752.852,
    "train_samples": 42336,
    "train_samples_per_second": 24.153,
    "train_steps_per_second": 0.377
}