[Frontend] Delegate tokenization serving preprocessing to OpenAIServingRender (#37266)

Signed-off-by: Sage Ahrac <sagiahrak@gmail.com>
This commit is contained in:
Sage
2026-03-17 11:22:54 +00:00
committed by GitHub
parent 4af9ed21cb
commit 00f8e0d211
5 changed files with 33 additions and 30 deletions
+6 -3
View File
@@ -11,6 +11,7 @@ from vllm.entrypoints.logger import RequestLogger
from vllm.entrypoints.openai.engine.protocol import ErrorResponse
from vllm.entrypoints.openai.engine.serving import OpenAIServing
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.entrypoints.serve.tokenize.protocol import (
DetokenizeRequest,
DetokenizeResponse,
@@ -31,6 +32,7 @@ class OpenAIServingTokenization(OpenAIServing):
self,
engine_client: EngineClient,
models: OpenAIServingModels,
openai_serving_render: OpenAIServingRender,
*,
request_logger: RequestLogger | None,
chat_template: str | None,
@@ -44,6 +46,7 @@ class OpenAIServingTokenization(OpenAIServing):
request_logger=request_logger,
)
self.openai_serving_render = openai_serving_render
self.chat_template = chat_template
self.chat_template_content_format: Final = chat_template_content_format
self.default_chat_template_kwargs = default_chat_template_kwargs or {}
@@ -68,7 +71,7 @@ class OpenAIServingTokenization(OpenAIServing):
if request.tools is None
else [tool.model_dump() for tool in request.tools]
)
error_check_ret = self._validate_chat_template(
error_check_ret = self.openai_serving_render.validate_chat_template(
request_chat_template=request.chat_template,
chat_template_kwargs=request.chat_template_kwargs,
trust_request_chat_template=self.trust_request_chat_template,
@@ -76,7 +79,7 @@ class OpenAIServingTokenization(OpenAIServing):
if error_check_ret is not None:
return error_check_ret
_, engine_prompts = await self._preprocess_chat(
_, engine_prompts = await self.openai_serving_render.preprocess_chat(
request,
request.messages,
default_template=self.chat_template,
@@ -85,7 +88,7 @@ class OpenAIServingTokenization(OpenAIServing):
tool_dicts=tool_dicts,
)
else:
engine_prompts = await self._preprocess_completion(
engine_prompts = await self.openai_serving_render.preprocess_completion(
request,
prompt_input=request.prompt,
prompt_embeds=None,