ThinkingBudgetLogitsProcessor# class tensorrt_llm.llmapi.ThinkingBudgetLogitsProcessor( thinking_token_budget: int, reasoning_start_token_ids: List[int], reasoning_end_token_ids: List[int], )[source]# Bases: LogitsProcessor Force the reasoning end token sequence once a thinking budget is spent. __init__( thinking_token_budget: int, reasoning_start_token_ids: List[int], reasoning_end_token_ids: List[int], ) → None[source]#
