all_results.json

{
    "epoch": 0.9990600241708071,
    "eval_logits/chosen": -14.553688049316406,
    "eval_logits/rejected": -14.48775863647461,
    "eval_logps/chosen": -1.8966439962387085,
    "eval_logps/rejected": -2.3994932174682617,
    "eval_loss": 2.7458858489990234,
    "eval_rewards/accuracies": 0.7725409865379333,
    "eval_rewards/chosen": -18.966440200805664,
    "eval_rewards/margins": 5.028491497039795,
    "eval_rewards/rejected": -23.994932174682617,
    "eval_runtime": 86.405,
    "eval_samples": 1941,
    "eval_samples_per_second": 22.464,
    "eval_steps_per_second": 1.412,
    "total_flos": 0.0,
    "train_loss": 3.541660060677477,
    "train_runtime": 9046.6245,
    "train_samples": 59569,
    "train_samples_per_second": 6.585,
    "train_steps_per_second": 0.051
}