eval_results.json

{
    "epoch": 2.9930715935334873,
    "eval_logits/chosen": -0.35042497515678406,
    "eval_logits/rejected": -0.3381310701370239,
    "eval_logps/chosen": -150.01190185546875,
    "eval_logps/rejected": -158.89979553222656,
    "eval_loss": 1.6688746213912964,
    "eval_nll_loss": 0.416054904460907,
    "eval_rewards/accuracies": 0.6336206793785095,
    "eval_rewards/chosen": -15.001189231872559,
    "eval_rewards/margins": 0.8887884616851807,
    "eval_rewards/rejected": -15.88998031616211,
    "eval_runtime": 41.8137,
    "eval_samples": 1826,
    "eval_samples_per_second": 43.67,
    "eval_steps_per_second": 0.694
}