eval_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": 0.07597240805625916,
    "eval_logits/rejected": 0.14311569929122925,
    "eval_logps/chosen": -1.2622054815292358,
    "eval_logps/rejected": -1.3863670825958252,
    "eval_loss": 1.262463927268982,
    "eval_rewards/accuracies": 0.5474777221679688,
    "eval_rewards/chosen": -1.2622054815292358,
    "eval_rewards/margins": 0.12416145950555801,
    "eval_rewards/rejected": -1.3863670825958252,
    "eval_runtime": 40.3519,
    "eval_samples": 1345,
    "eval_samples_per_second": 33.332,
    "eval_steps_per_second": 8.352
}