all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -0.7336508631706238,
    "eval_logits/rejected": -0.7156788110733032,
    "eval_logps/chosen": -311.4344787597656,
    "eval_logps/rejected": -277.3082580566406,
    "eval_loss": 0.6741109490394592,
    "eval_rewards/accuracies": 0.6119999885559082,
    "eval_rewards/chosen": 0.05623208358883858,
    "eval_rewards/margins": 0.04554646834731102,
    "eval_rewards/rejected": 0.010685612447559834,
    "eval_runtime": 23.4224,
    "eval_samples_per_second": 42.694,
    "eval_steps_per_second": 5.337
}