eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2436792850494385,
    "eval_logits/rejected": -2.23879337310791,
    "eval_logps/chosen": -34.02314376831055,
    "eval_logps/rejected": -37.50123596191406,
    "eval_loss": 1.0000426769256592,
    "eval_rewards/accuracies": 0.5020764470100403,
    "eval_rewards/chosen": 0.0001141041939263232,
    "eval_rewards/margins": -3.9732771256240085e-05,
    "eval_rewards/rejected": 0.00015383693971671164,
    "eval_runtime": 145.4546,
    "eval_samples": 343,
    "eval_samples_per_second": 2.358,
    "eval_steps_per_second": 0.296
}