[Frontend] Parallelize preprocessing within the same request for pooling models online serving. (#49153)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
wang.yuqi
2026-07-22 10:56:23 +00:00
committed by GitHub
parent d6dbdb9b0d
commit 61c9ef986a
17 changed files with 538 additions and 820 deletions
@@ -16,8 +16,11 @@ from .serving import ServingClassification
router = APIRouter()
def classify(request: Request) -> ServingClassification | None:
return request.app.state.serving_classification
def classify(request: Request) -> ServingClassification:
handler = getattr(request.app.state, "serving_classification", None)
if handler is None:
raise NotImplementedError("The model does not support Classification API")
return handler
@router.post("/classify", dependencies=[Depends(validate_json_request)])
@@ -27,7 +30,4 @@ async def create_classify(
request: ClassificationRequest, raw_request: Request
) -> Response:
handler = classify(raw_request)
if handler is None:
raise NotImplementedError("The model does not support Classification API")
return await handler(request, raw_request)