eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.7827913761138916,
    "eval_logits/rejected": 0.8020623326301575,
    "eval_logps/chosen": -92.44380187988281,
    "eval_logps/rejected": -95.30902099609375,
    "eval_loss": 0.6912217736244202,
    "eval_rewards/accuracies": 0.3179999887943268,
    "eval_rewards/chosen": -0.00716440798714757,
    "eval_rewards/margins": 0.00395576748996973,
    "eval_rewards/rejected": -0.011120175942778587,
    "eval_runtime": 273.2522,
    "eval_samples": 2000,
    "eval_samples_per_second": 7.319,
    "eval_steps_per_second": 0.457
}