[Frontend] Add GPU-less render serving path (vllm launch render) (#36166)

This commit is contained in:
Sage
2026-03-08 16:35:09 +01:00
committed by GitHub
parent b7332b058c
commit 4497431df6
10 changed files with 712 additions and 273 deletions
+5 -7
View File
@@ -8,7 +8,7 @@ import uvloop
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.entrypoints.cli.types import CLISubcommand
from vllm.entrypoints.openai.api_server import (
build_and_serve,
build_and_serve_renderer,
setup_server,
)
from vllm.entrypoints.openai.cli_args import (
@@ -109,19 +109,17 @@ def cmd_init() -> list[CLISubcommand]:
async def run_launch_fastapi(args: argparse.Namespace) -> None:
"""Run the online serving layer with FastAPI (no GPU inference)."""
from vllm.config import VllmConfig
from vllm.v1.engine.launch import LaunchEngineClient
# 1. Socket binding
listen_address, sock = setup_server(args)
# 2. Create LaunchEngineClient (no GPU)
# 2. Build and serve the API server
engine_args = AsyncEngineArgs.from_cli_args(args)
model_config = engine_args.create_model_config()
vllm_config = VllmConfig(model_config=model_config)
engine_client = LaunchEngineClient.from_vllm_config(vllm_config)
# 3. Build app, initialize state, and start serving
shutdown_task = await build_and_serve(engine_client, listen_address, sock, args)
shutdown_task = await build_and_serve_renderer(
vllm_config, listen_address, sock, args
)
try:
await shutdown_task
finally: