[Frontend][4/n] Improve pooling entrypoints | pooling. (#39153)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
wang.yuqi
2026-04-09 10:09:45 +00:00
committed by GitHub
parent b6c9be509e
commit 66c079ae83
43 changed files with 554 additions and 733 deletions
+2 -4
View File
@@ -67,20 +67,18 @@ def init_pooling_state(
from vllm.entrypoints.chat_utils import load_chat_template
from vllm.entrypoints.pooling.classify.serving import ServingClassification
from vllm.entrypoints.pooling.embed.serving import ServingEmbedding
from vllm.entrypoints.pooling.pooling.serving import OpenAIServingPooling
from vllm.entrypoints.pooling.pooling.serving import ServingPooling
from vllm.entrypoints.pooling.scoring.serving import ServingScores
from vllm.tasks import POOLING_TASKS
model_config = engine_client.model_config
resolved_chat_template = load_chat_template(args.chat_template)
state.serving_pooling = (
(
OpenAIServingPooling(
ServingPooling(
engine_client,
state.openai_serving_models,
state.openai_serving_render,
supported_tasks=supported_tasks,
request_logger=request_logger,
chat_template=resolved_chat_template,