all_results.json

{
    "epoch": 2.999098751126561,
    "eval_kl/chosen": 0.0003600713098421693,
    "eval_kl/rejected": 0.0003134958096779883,
    "eval_logits/chosen": -1.40625,
    "eval_logits/rejected": -1.3671875,
    "eval_logps/chosen": -322.0,
    "eval_logps/rejected": -298.0,
    "eval_loss": 0.6936718821525574,
    "eval_nll_loss": 1.046875,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.0003134958096779883,
    "eval_rejected_term/min": 0.0003134958096779883,
    "eval_rewards/accuracies": 0.5640000104904175,
    "eval_rewards/chosen": 0.000202178955078125,
    "eval_rewards/margins": 0.103515625,
    "eval_rewards/rejected": -0.10302734375,
    "eval_runtime": 31.7822,
    "eval_samples_per_second": 31.464,
    "eval_steps_per_second": 7.866
}