[Perf] Eliminate full-history reasoning scans for structured outputs (#55223)

Signed-off-by: sfeng33 <4florafeng@gmail.com>
Signed-off-by: Flora Feng <4florafeng@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
Flora Feng
2026-09-08 14:07:16 -07:00
committed by GitHub
co-authored by Wentao Ye
parent 73f61d12aa
commit 29f46ac475
8 changed files with 475 additions and 42 deletions
+7
View File
@@ -65,6 +65,9 @@ class ParserEngineReasoningAdapter(ReasoningParser):
def is_reasoning_end(self, input_ids: Sequence[int]) -> bool:
return self._parser_engine.is_reasoning_end(list(input_ids))
def find_reasoning_end_offset(self, token_ids: Sequence[int]) -> int | None:
return self._parser_engine.find_reasoning_end_offset(token_ids)
def adjust_initial_state_from_prompt(self, prompt_token_ids: Sequence[int]) -> None:
self._parser_engine.adjust_initial_state_from_prompt(prompt_token_ids)
@@ -108,6 +111,10 @@ class ParserEngineReasoningAdapter(ReasoningParser):
def reasoning_end_str(self) -> str | None:
return self._parser_engine.reasoning_end_str
@property
def reasoning_end_token_ids(self) -> frozenset[int]:
return self._parser_engine.reasoning_end_token_ids
def adjust_request(
self,
request: ChatCompletionRequest | ResponsesRequest,