eval_results.json

{
    "epoch": 2.996910401647786,
    "eval_logits/chosen": -0.3054884672164917,
    "eval_logits/rejected": -0.7285566926002502,
    "eval_logps/chosen": -566.8709106445312,
    "eval_logps/rejected": -803.376953125,
    "eval_loss": 0.567832887172699,
    "eval_rewards/accuracies": 0.765625,
    "eval_rewards/chosen": -3.1023550033569336,
    "eval_rewards/margins": 2.285463333129883,
    "eval_rewards/rejected": -5.387818813323975,
    "eval_runtime": 82.2572,
    "eval_samples_per_second": 12.157,
    "eval_steps_per_second": 0.195
}