eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.411515712738037,
    "eval_logits/rejected": -3.456860065460205,
    "eval_logps/chosen": -564.323974609375,
    "eval_logps/rejected": -567.8529052734375,
    "eval_loss": 0.7979298830032349,
    "eval_rewards/accuracies": 0.46875,
    "eval_rewards/chosen": 4.5177483558654785,
    "eval_rewards/diff": -0.33996284008026123,
    "eval_rewards/diff_abs": 1.2063032388687134,
    "eval_rewards/rejected": 4.610641002655029,
    "eval_rewards/student_margin": -0.09289252758026123,
    "eval_rewards/teacher_margin": 0.2470703125,
    "eval_runtime": 26.855,
    "eval_samples": 1543,
    "eval_samples_per_second": 57.457,
    "eval_steps_per_second": 0.149
}