eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.7731637954711914,
    "eval_logits/rejected": -2.7412285804748535,
    "eval_logps/chosen": -303.5331115722656,
    "eval_logps/rejected": -303.9143371582031,
    "eval_loss": 0.6104059815406799,
    "eval_rewards/accuracies": 0.6869999766349792,
    "eval_rewards/chosen": -0.1893969476222992,
    "eval_rewards/margins": 0.26395705342292786,
    "eval_rewards/margins_max": 1.0323494672775269,
    "eval_rewards/margins_min": -0.46421656012535095,
    "eval_rewards/margins_std": 0.5045729875564575,
    "eval_rewards/rejected": -0.45335400104522705,
    "eval_runtime": 428.0701,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.672,
    "eval_steps_per_second": 0.292
}