eval_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 99.54110717773438,
    "eval_logits/rejected": 99.75216674804688,
    "eval_logps/chosen": -11.42675495147705,
    "eval_logps/rejected": -15.375484466552734,
    "eval_loss": 61.015228271484375,
    "eval_rewards/accuracies": 0.8020833134651184,
    "eval_rewards/chosen": -0.49881264567375183,
    "eval_rewards/margins": 0.1921073943376541,
    "eval_rewards/rejected": -0.6909201145172119,
    "eval_runtime": 120.2204,
    "eval_samples": 750,
    "eval_samples_per_second": 6.239,
    "eval_steps_per_second": 0.2
}