eval_results.json

{
    "epoch": 1.0,
    "eval_error": 0.2004745602607727,
    "eval_logits/chosen": -3.2435498237609863,
    "eval_logits/rejected": -3.2166450023651123,
    "eval_logps/chosen": -487.0957336425781,
    "eval_logps/rejected": -494.91912841796875,
    "eval_loss": 0.7033457159996033,
    "eval_preference/chosen": 0.5359722971916199,
    "eval_preference/rejected": 0.4640277028083801,
    "eval_rewards/accuracies": 0.5490000247955322,
    "eval_rewards/chosen": -3.492784023284912,
    "eval_rewards/margins": 0.6411816477775574,
    "eval_rewards/rejected": -4.133965492248535,
    "eval_runtime": 42.0461,
    "eval_samples_per_second": 23.783,
    "eval_steps_per_second": 2.973
}