eval_results.json

{
    "epoch": 0.9997382884061764,
    "eval_logits/chosen": -1.0759395360946655,
    "eval_logits/rejected": -1.1037203073501587,
    "eval_logps/chosen": -1.2659075260162354,
    "eval_logps/rejected": -1.5190186500549316,
    "eval_loss": 1.376198172569275,
    "eval_rewards/accuracies": 0.7559523582458496,
    "eval_rewards/chosen": 0.06548410654067993,
    "eval_rewards/margins": 0.435606449842453,
    "eval_rewards/rejected": -0.37012240290641785,
    "eval_runtime": 262.4629,
    "eval_samples": 2000,
    "eval_samples_per_second": 7.62,
    "eval_steps_per_second": 0.24
}