all_results.json

{
    "epoch": 1.9997425003218745,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": -5.614200592041016,
    "eval_logps/chosen": -668.839111328125,
    "eval_logps/rejected": -560.14111328125,
    "eval_loss": 0.6387723684310913,
    "eval_rewards/accuracies": 0.6010000109672546,
    "eval_rewards/chosen": -0.07504798471927643,
    "eval_rewards/margins": 0.16594481468200684,
    "eval_rewards/rejected": -0.24099276959896088,
    "eval_runtime": 47.0636,
    "eval_samples_per_second": 21.248,
    "eval_steps_per_second": 2.656
}