eval_results.json

{
    "epoch": 2.994871794871795,
    "eval_logits/chosen": -2.296416759490967,
    "eval_logits/rejected": -2.2604804039001465,
    "eval_logps/chosen": -191.28500366210938,
    "eval_logps/rejected": -219.8085479736328,
    "eval_loss": 0.4512675702571869,
    "eval_rewards/accuracies": 0.8211538195610046,
    "eval_rewards/chosen": -3.2807652950286865,
    "eval_rewards/margins": 2.812060594558716,
    "eval_rewards/rejected": -6.092825889587402,
    "eval_runtime": 509.575,
    "eval_samples_per_second": 16.316,
    "eval_steps_per_second": 0.255
}