eval_results.json

{
    "epoch": 2.9158878504672896,
    "eval_logits/chosen": -2.424323320388794,
    "eval_logits/rejected": -2.4079055786132812,
    "eval_logps/chosen": -228.84059143066406,
    "eval_logps/rejected": -192.46083068847656,
    "eval_loss": 0.5374945402145386,
    "eval_rewards/accuracies": 0.8333333134651184,
    "eval_rewards/chosen": 1.9424031972885132,
    "eval_rewards/margins": 2.1486706733703613,
    "eval_rewards/rejected": -0.20626752078533173,
    "eval_runtime": 49.2416,
    "eval_samples_per_second": 15.434,
    "eval_steps_per_second": 0.244
}