llama3-8b-dpo-4xh100-pilot/all_results.json

{
    "epoch": 1.0,
    "total_flos": 0.0,
    "train_loss": 0.6806982485360877,
    "train_runtime": 636.997,
    "train_samples": 5501,
    "train_samples_per_second": 8.636,
    "train_steps_per_second": 0.135
}