eval_results.json

{
    "epoch": 0.9988571428571429,
    "eval_logits/chosen": 0.6892262101173401,
    "eval_logits/rejected": 2.4422545433044434,
    "eval_logps/chosen": -432.7330322265625,
    "eval_logps/rejected": -518.172607421875,
    "eval_loss": 0.541784942150116,
    "eval_rewards/accuracies": 0.767241358757019,
    "eval_rewards/chosen": -1.5712907314300537,
    "eval_rewards/margins": 1.419674038887024,
    "eval_rewards/rejected": -2.990964651107788,
    "eval_runtime": 91.4769,
    "eval_samples": 1831,
    "eval_samples_per_second": 20.016,
    "eval_steps_per_second": 0.317
}