all_results.json

{
    "epoch": 2.999098751126561,
    "eval_kl/chosen": 1.2188414075353649e-05,
    "eval_kl/rejected": 0.00023281562607735395,
    "eval_logits/chosen": -1.40625,
    "eval_logits/rejected": -1.375,
    "eval_logps/chosen": -322.0,
    "eval_logps/rejected": -298.0,
    "eval_loss": 0.7706953287124634,
    "eval_nll_loss": 1.046875,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.00023281562607735395,
    "eval_rejected_term/min": 0.00023281562607735395,
    "eval_rewards/accuracies": 0.5320000052452087,
    "eval_rewards/chosen": -0.006683349609375,
    "eval_rewards/margins": 0.064453125,
    "eval_rewards/rejected": -0.0712890625,
    "eval_runtime": 25.2659,
    "eval_samples_per_second": 39.579,
    "eval_steps_per_second": 9.895
}