eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.1651915311813354,
    "eval_logits/rejected": -1.0289729833602905,
    "eval_logps/chosen": -526.7527465820312,
    "eval_logps/rejected": -636.4378662109375,
    "eval_loss": 0.47883233428001404,
    "eval_rewards/accuracies": 0.7465000152587891,
    "eval_rewards/chosen": -2.6215085983276367,
    "eval_rewards/margins": 1.2971659898757935,
    "eval_rewards/rejected": -3.9186742305755615,
    "eval_runtime": 1597.3624,
    "eval_samples": 2000,
    "eval_samples_per_second": 1.252,
    "eval_steps_per_second": 0.157
}