eval_results.json

{
    "epoch": 1.9978142076502732,
    "eval_logits/chosen": -1.4140465259552002,
    "eval_logits/rejected": -1.3327398300170898,
    "eval_logps/chosen": -0.8071117401123047,
    "eval_logps/rejected": -1.9723843336105347,
    "eval_loss": 1.2685352563858032,
    "eval_rewards/accuracies": 0.8674699068069458,
    "eval_rewards/chosen": -8.071117401123047,
    "eval_rewards/margins": 11.652724266052246,
    "eval_rewards/rejected": -19.723840713500977,
    "eval_runtime": 33.7378,
    "eval_samples": 1318,
    "eval_samples_per_second": 39.066,
    "eval_steps_per_second": 2.46
}