all_results.json

{
    "epoch": 2.999214865218529,
    "eval_kl": 0.0,
    "eval_logits/chosen": -1429059328.0,
    "eval_logits/rejected": -1154032768.0,
    "eval_logps/chosen": -548.4809500247402,
    "eval_logps/rejected": -840.9109895574342,
    "eval_loss": 0.4489765763282776,
    "eval_rewards/chosen": -2.1658832261256804,
    "eval_rewards/margins": 3.285384799732102,
    "eval_rewards/rejected": -5.451268025857782,
    "eval_runtime": 93.4316,
    "eval_samples": 4000,
    "eval_samples_per_second": 42.812,
    "eval_steps_per_second": 0.674,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 0.0431,
    "train_samples": 122269,
    "train_samples_per_second": 8512981.493,
    "train_steps_per_second": 132983.787
}