eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2295007705688477,
    "eval_logits/rejected": -2.2246556282043457,
    "eval_logps/chosen": -34.054466247558594,
    "eval_logps/rejected": -37.57353973388672,
    "eval_loss": 2.790188789367676,
    "eval_rewards/accuracies": 0.560215950012207,
    "eval_rewards/chosen": -0.0059735761024057865,
    "eval_rewards/margins": 0.011101600714027882,
    "eval_rewards/rejected": -0.017075177282094955,
    "eval_runtime": 145.7539,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295
}