maxent\_grpo.training.scoring\_reference ======================================== .. automodule:: maxent_grpo.training.scoring_reference .. rubric:: Functions .. autosummary:: _coerce_int_optional _coerce_logprob_value _meta_field _sum_token_logprobs finalize_reference_stats gather_reference_logprobs reference_from_model reference_from_model_trl reference_from_vllm_meta reference_stats_from_policy_logprobs vllm_meta_has_logprobs