eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.4300496578216553,
    "eval_logits/rejected": -2.387667179107666,
    "eval_logps/chosen": -452.2376708984375,
    "eval_logps/rejected": -458.4551696777344,
    "eval_loss": 0.5859559178352356,
    "eval_rewards/accuracies": 0.6890000104904175,
    "eval_rewards/chosen": -1.1601862907409668,
    "eval_rewards/margins": 0.45331665873527527,
    "eval_rewards/rejected": -1.613502860069275,
    "eval_runtime": 304.9269,
    "eval_samples": 2000,
    "eval_samples_per_second": 6.559,
    "eval_steps_per_second": 0.82
}