aup-fullft-kto_w1-w1lam9.68e-4-seed5/train_results.json

{
    "epoch": 3.0,
    "total_flos": 9.97585917447635e+17,
    "train_loss": 0.5860847269726952,
    "train_runtime": 11055.2237,
    "train_samples_per_second": 13.407,
    "train_steps_per_second": 0.209
}