eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.05413578823208809,
    "eval_logits/rejected": -0.01999257504940033,
    "eval_logps/chosen": -360.3287353515625,
    "eval_logps/rejected": -371.92205810546875,
    "eval_loss": 0.5779720544815063,
    "eval_rewards/accuracies": 0.6305000185966492,
    "eval_rewards/chosen": -0.8278191685676575,
    "eval_rewards/margins": 0.4532422423362732,
    "eval_rewards/rejected": -1.2810615301132202,
    "eval_runtime": 545.1859,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.668,
    "eval_steps_per_second": 0.917
}