eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2434139251708984,
    "eval_logits/rejected": -2.2385354042053223,
    "eval_logps/chosen": -34.0224494934082,
    "eval_logps/rejected": -37.51002883911133,
    "eval_loss": 0.9995055198669434,
    "eval_rewards/accuracies": 0.5141196250915527,
    "eval_rewards/chosen": 0.0012102578766644,
    "eval_rewards/margins": 0.0005511684576049447,
    "eval_rewards/rejected": 0.0006590897683054209,
    "eval_runtime": 145.6584,
    "eval_samples": 343,
    "eval_samples_per_second": 2.355,
    "eval_steps_per_second": 0.295
}