eval_results.json

{
    "epoch": 0.8,
    "eval_logits/chosen": -0.5457379221916199,
    "eval_logits/rejected": -0.5458806753158569,
    "eval_logps/chosen": -317.97125244140625,
    "eval_logps/ref_response": -0.5449998378753662,
    "eval_logps/rejected": -331.12335205078125,
    "eval_loss": 0.6931814551353455,
    "eval_rewards/accuracies": 0.484375,
    "eval_rewards/chosen": -0.001830391469411552,
    "eval_rewards/margins": -0.000241550849750638,
    "eval_rewards/rejected": -0.0015888408524915576,
    "eval_runtime": 54.0012,
    "eval_samples": 240,
    "eval_samples_per_second": 4.444,
    "eval_steps_per_second": 0.148
}