eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.3996024131774902,
    "eval_logits/rejected": -3.2167699337005615,
    "eval_logps/chosen": -208.22796630859375,
    "eval_logps/rejected": -1466.67333984375,
    "eval_loss": 0.0005253218114376068,
    "eval_rewards/accuracies": 1.0,
    "eval_rewards/chosen": -1.7119580507278442,
    "eval_rewards/margins": 12.642820358276367,
    "eval_rewards/rejected": -14.354778289794922,
    "eval_runtime": 3.4665,
    "eval_samples": 100,
    "eval_samples_per_second": 28.847,
    "eval_steps_per_second": 1.154
}