eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.805530071258545,
    "eval_logits/rejected": -2.802974224090576,
    "eval_logps/chosen": -31.315858840942383,
    "eval_logps/rejected": -34.838993072509766,
    "eval_loss": 6.074533939361572,
    "eval_rewards/accuracies": 0.5660299062728882,
    "eval_rewards/chosen": -0.006680913269519806,
    "eval_rewards/margins": 0.021486198529601097,
    "eval_rewards/rejected": -0.0281671155244112,
    "eval_runtime": 112.8101,
    "eval_samples": 343,
    "eval_samples_per_second": 3.041,
    "eval_steps_per_second": 0.381
}