eval_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -1.4189956188201904,
    "eval_logits/rejected": -1.3920025825500488,
    "eval_logps/chosen": -244.38746643066406,
    "eval_logps/rejected": -295.19293212890625,
    "eval_loss": 0.6342188119888306,
    "eval_rewards/accuracies": 0.6579925417900085,
    "eval_rewards/chosen": -1.8568360805511475,
    "eval_rewards/margins": 0.4635196328163147,
    "eval_rewards/rejected": -2.3203558921813965,
    "eval_runtime": 354.8785,
    "eval_samples": 4304,
    "eval_samples_per_second": 12.128,
    "eval_steps_per_second": 1.516
}