all_results.json

{
    "epoch": 90.0,
    "eval_kl/both": 0.013082483783364296,
    "eval_kl/chosen": 0.006705726496875286,
    "eval_kl/rejected": 0.006368208210915327,
    "eval_logits/chosen": -2.546875,
    "eval_logits/rejected": -2.5625,
    "eval_logps/chosen": -736.0,
    "eval_logps/rejected": -644.0,
    "eval_loss": 15.256285667419434,
    "eval_nll_loss": 2.453125,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.006368208210915327,
    "eval_rejected_term/min": 0.006368208210915327,
    "eval_rewards/accuracies": 0.41600000858306885,
    "eval_rewards/chosen": -41.5,
    "eval_rewards/margins": -6.59375,
    "eval_rewards/rejected": -34.75,
    "eval_runtime": 25.3758,
    "eval_samples_per_second": 39.408,
    "eval_steps_per_second": 9.852
}