eval_results.json

{
    "epoch": 0.9987943737441393,
    "eval_logits/chosen": -2.8956403732299805,
    "eval_logits/rejected": -2.887930154800415,
    "eval_logps/chosen": -1.1685596704483032,
    "eval_logps/rejected": -1.2866594791412354,
    "eval_loss": -0.07405862212181091,
    "eval_rewards/accuracies": 0.5797872543334961,
    "eval_rewards/chosen": -1.1685596704483032,
    "eval_rewards/margins": 0.11809979379177094,
    "eval_rewards/rejected": -1.2866594791412354,
    "eval_runtime": 432.6369,
    "eval_samples": 2994,
    "eval_samples_per_second": 6.92,
    "eval_steps_per_second": 0.435
}