eval_results.json

{
    "epoch": 0.9987943737441393,
    "eval_logits/chosen": -3.4056549072265625,
    "eval_logits/rejected": -3.3998796939849854,
    "eval_logps/chosen": -0.5611971616744995,
    "eval_logps/rejected": -0.6183008551597595,
    "eval_loss": 1.5195728540420532,
    "eval_rewards/accuracies": 0.5691489577293396,
    "eval_rewards/chosen": -0.5611971616744995,
    "eval_rewards/margins": 0.057103704661130905,
    "eval_rewards/rejected": -0.6183008551597595,
    "eval_runtime": 434.9992,
    "eval_samples": 2994,
    "eval_samples_per_second": 6.883,
    "eval_steps_per_second": 0.432
}