eval_results.json

{
    "epoch": 2.8363636363636364,
    "eval_logits/chosen": -2.1622750759124756,
    "eval_logits/rejected": -2.0877833366394043,
    "eval_logps/chosen": -330.47552490234375,
    "eval_logps/rejected": -282.82672119140625,
    "eval_loss": 0.49291226267814636,
    "eval_rewards/accuracies": 0.7857142686843872,
    "eval_rewards/chosen": 4.344088554382324,
    "eval_rewards/margins": 2.9927380084991455,
    "eval_rewards/rejected": 1.3513505458831787,
    "eval_runtime": 28.655,
    "eval_samples_per_second": 13.436,
    "eval_steps_per_second": 0.244
}