eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 98.46549224853516,
    "eval_logits/rejected": 98.43663787841797,
    "eval_logps/chosen": -32.78712463378906,
    "eval_logps/rejected": -36.48957443237305,
    "eval_loss": 0.9517204761505127,
    "eval_rewards/accuracies": 0.5191029906272888,
    "eval_rewards/chosen": -0.10318472981452942,
    "eval_rewards/margins": 0.0537295788526535,
    "eval_rewards/rejected": -0.15691432356834412,
    "eval_runtime": 103.7591,
    "eval_samples": 343,
    "eval_samples_per_second": 3.306,
    "eval_steps_per_second": 0.414
}