eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.6475114822387695,
    "eval_logits/rejected": -3.464193344116211,
    "eval_logps/chosen": -984.8101806640625,
    "eval_logps/rejected": -1480.8428955078125,
    "eval_loss": 0.4465055763721466,
    "eval_rewards/accuracies": 0.7845849990844727,
    "eval_rewards/chosen": -0.5929392576217651,
    "eval_rewards/margins": 1.0842887163162231,
    "eval_rewards/rejected": -1.6772277355194092,
    "eval_runtime": 1610.6329,
    "eval_samples": 16161,
    "eval_samples_per_second": 10.034,
    "eval_steps_per_second": 0.314
}