eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2319114208221436,
    "eval_logits/rejected": -2.2271130084991455,
    "eval_logps/chosen": -33.75971984863281,
    "eval_logps/rejected": -37.29259490966797,
    "eval_loss": 0.49754369258880615,
    "eval_rewards/accuracies": 0.5394518375396729,
    "eval_rewards/chosen": 0.05496572330594063,
    "eval_rewards/margins": 0.010160263627767563,
    "eval_rewards/rejected": 0.044805459678173065,
    "eval_runtime": 145.6941,
    "eval_samples": 343,
    "eval_samples_per_second": 2.354,
    "eval_steps_per_second": 0.295
}