[Frontend] Split ServingRender into renderer and entrypoint. (#44285)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
wang.yuqi
2026-06-23 11:19:09 +00:00
committed by GitHub
parent d32575a2d2
commit 31ca9504b1
26 changed files with 1155 additions and 939 deletions
+50 -8
View File
@@ -32,7 +32,7 @@ from vllm.entrypoints.openai.engine.protocol import GenerationError
from vllm.entrypoints.openai.models.protocol import BaseModelPath
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.serve.elastic_ep.middleware import ScalingMiddleware
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.entrypoints.serve.render.serving import ServingRender
from vllm.entrypoints.serve.sagemaker.api_router import sagemaker_standards_bootstrap
from vllm.entrypoints.serve.tokenize.serving import ServingTokenization
from vllm.entrypoints.serve.utils.api_utils import (
@@ -55,6 +55,8 @@ from vllm.entrypoints.serve.utils.server_utils import (
from vllm.exceptions import VLLMValidationError
from vllm.logger import init_logger
from vllm.reasoning import ReasoningParserManager
from vllm.renderers.online_derenderer import OnlineDerenderer
from vllm.renderers.online_renderer import OnlineRenderer
from vllm.tasks import POOLING_TASKS, SupportedTask
from vllm.tool_parsers import ToolParserManager
from vllm.tracing import instrument
@@ -360,10 +362,24 @@ async def init_app_state(
)
await state.openai_serving_models.init_static_loras()
state.openai_serving_render = OpenAIServingRender(
state.online_renderer = OnlineRenderer(
model_config=engine_client.model_config,
renderer=engine_client.renderer,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
trust_request_chat_template=args.trust_request_chat_template,
enable_auto_tools=args.enable_auto_tool_choice,
exclude_tools_when_tool_choice_none=args.exclude_tools_when_tool_choice_none,
tool_parser=args.tool_call_parser,
reasoning_parser=args.structured_outputs_config.reasoning_parser,
default_chat_template_kwargs=args.default_chat_template_kwargs,
log_error_stack=args.log_error_stack,
)
state.online_derenderer = OnlineDerenderer(
model_config=engine_client.model_config,
renderer=engine_client.renderer,
model_registry=state.openai_serving_models.registry,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
@@ -378,13 +394,19 @@ async def init_app_state(
state.serving_tokenization = ServingTokenization(
state.openai_serving_models,
state.openai_serving_render,
state.online_renderer,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
default_chat_template_kwargs=args.default_chat_template_kwargs,
trust_request_chat_template=args.trust_request_chat_template,
)
state.serving_render = ServingRender(
state.openai_serving_models,
state.online_renderer,
state.online_derenderer,
request_logger=request_logger,
)
if "generate" in supported_tasks:
from vllm.entrypoints.generate.api_router import init_generate_state
@@ -423,8 +445,8 @@ async def init_render_app_state(
"""
from vllm.entrypoints.chat_utils import load_chat_template
from vllm.entrypoints.openai.models.serving import OpenAIModelRegistry
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
from vllm.renderers import renderer_from_config
from vllm.renderers.online_renderer import OnlineRenderer
served_model_names = args.served_model_name or [args.model]
model_registry = OpenAIModelRegistry(
@@ -443,10 +465,24 @@ async def init_render_app_state(
renderer = renderer_from_config(vllm_config)
resolved_chat_template = load_chat_template(args.chat_template)
state.openai_serving_render = OpenAIServingRender(
state.online_renderer = OnlineRenderer(
model_config=vllm_config.model_config,
renderer=renderer,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
trust_request_chat_template=args.trust_request_chat_template,
enable_auto_tools=args.enable_auto_tool_choice,
exclude_tools_when_tool_choice_none=args.exclude_tools_when_tool_choice_none,
tool_parser=args.tool_call_parser,
reasoning_parser=args.reasoning_parser,
default_chat_template_kwargs=args.default_chat_template_kwargs,
log_error_stack=args.log_error_stack,
)
state.online_derenderer = OnlineDerenderer(
model_config=vllm_config.model_config,
renderer=renderer,
model_registry=model_registry,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
@@ -462,13 +498,19 @@ async def init_render_app_state(
state.openai_serving_models = model_registry
state.serving_tokenization = ServingTokenization(
model_registry,
state.openai_serving_render,
state.online_renderer,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
default_chat_template_kwargs=args.default_chat_template_kwargs,
trust_request_chat_template=args.trust_request_chat_template,
)
state.serving_render = ServingRender(
model_registry,
state.online_renderer,
state.online_derenderer,
request_logger=request_logger,
)
state.vllm_config = vllm_config
# Disable stats logging — there is no engine to poll.