eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2436861991882324,
    "eval_logits/rejected": -2.2388129234313965,
    "eval_logps/chosen": -34.01886749267578,
    "eval_logps/rejected": -37.50727081298828,
    "eval_loss": 0.34474730491638184,
    "eval_rewards/accuracies": 0.5161960124969482,
    "eval_rewards/chosen": 0.01411630492657423,
    "eval_rewards/margins": 0.005696756765246391,
    "eval_rewards/rejected": 0.008419547230005264,
    "eval_runtime": 145.8183,
    "eval_samples": 343,
    "eval_samples_per_second": 2.352,
    "eval_steps_per_second": 0.295
}