eval_results.json

{
    "epoch": 0.9981298423724285,
    "eval_logits/chosen": -1.1943395137786865,
    "eval_logits/rejected": -1.1611931324005127,
    "eval_logps/chosen": -1.2059502601623535,
    "eval_logps/rejected": -1.4740996360778809,
    "eval_loss": 1.373450517654419,
    "eval_rewards/accuracies": 0.6653226017951965,
    "eval_rewards/chosen": -3.0148754119873047,
    "eval_rewards/margins": 0.6703741550445557,
    "eval_rewards/rejected": -3.6852493286132812,
    "eval_runtime": 71.4975,
    "eval_samples": 1961,
    "eval_samples_per_second": 27.428,
    "eval_steps_per_second": 0.867
}