qwen3-8b-base-beta-dpo-hh-helpful-4xh200-batch-64/train_results.json

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 1.1809129529850073,
    "train_runtime": 2340.6187,
    "train_samples": 43598,
    "train_samples_per_second": 18.627,
    "train_steps_per_second": 0.291
}