biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

53
vllm/entrypoints/launcher.py
View File

@@ -12,8 +12,10 @@ from fastapi import FastAPI, Request, Response
from vllm import envs
from vllm.engine.protocol import EngineClient
from vllm.entrypoints.constants import (H11_MAX_HEADER_COUNT_DEFAULT,
H11_MAX_INCOMPLETE_EVENT_SIZE_DEFAULT)
from vllm.entrypoints.constants import (
H11_MAX_HEADER_COUNT_DEFAULT,
H11_MAX_INCOMPLETE_EVENT_SIZE_DEFAULT,
)
from vllm.entrypoints.ssl import SSLCertRefresher
from vllm.logger import init_logger
from vllm.utils import find_process_using_port
@@ -22,10 +24,12 @@ from vllm.v1.engine.exceptions import EngineDeadError, EngineGenerateError
logger = init_logger(__name__)
async def serve_http(app: FastAPI,
sock: Optional[socket.socket],
enable_ssl_refresh: bool = False,
**uvicorn_kwargs: Any):
async def serve_http(
app: FastAPI,
sock: Optional[socket.socket],
enable_ssl_refresh: bool = False,
**uvicorn_kwargs: Any,
):
"""
Start a FastAPI app using Uvicorn, with support for custom Uvicorn config
options. Supports http header limits via h11_max_incomplete_event_size and
@@ -39,11 +43,12 @@ async def serve_http(app: FastAPI,
if methods is None or path is None:
continue
logger.info("Route: %s, Methods: %s", path, ', '.join(methods))
logger.info("Route: %s, Methods: %s", path, ", ".join(methods))
# Extract header limit options if present
h11_max_incomplete_event_size = uvicorn_kwargs.pop(
"h11_max_incomplete_event_size", None)
"h11_max_incomplete_event_size", None
)
h11_max_header_count = uvicorn_kwargs.pop("h11_max_header_count", None)
# Set safe defaults if not provided
@@ -62,16 +67,19 @@ async def serve_http(app: FastAPI,
loop = asyncio.get_running_loop()
watchdog_task = loop.create_task(
watchdog_loop(server, app.state.engine_client))
server_task = loop.create_task(
server.serve(sockets=[sock] if sock else None))
watchdog_task = loop.create_task(watchdog_loop(server, app.state.engine_client))
server_task = loop.create_task(server.serve(sockets=[sock] if sock else None))
ssl_cert_refresher = None if not enable_ssl_refresh else SSLCertRefresher(
ssl_context=config.ssl,
key_path=config.ssl_keyfile,
cert_path=config.ssl_certfile,
ca_path=config.ssl_ca_certs)
ssl_cert_refresher = (
None
if not enable_ssl_refresh
else SSLCertRefresher(
ssl_context=config.ssl,
key_path=config.ssl_keyfile,
cert_path=config.ssl_certfile,
ca_path=config.ssl_ca_certs,
)
)
def signal_handler() -> None:
# prevents the uvicorn signal handler to exit early
@@ -95,7 +103,10 @@ async def serve_http(app: FastAPI,
if process is not None:
logger.warning(
"port %s is used by process %s launched with command:\n%s",
port, process, " ".join(process.cmdline()))
port,
process,
" ".join(process.cmdline()),
)
logger.info("Shutting down FastAPI HTTP server.")
return server.shutdown()
finally:
@@ -131,14 +142,14 @@ def _add_shutdown_handlers(app: FastAPI, server: uvicorn.Server) -> None:
"""
VLLM V1 AsyncLLM catches exceptions and returns
only two types: EngineGenerateError and EngineDeadError.
EngineGenerateError is raised by the per request generate()
method. This error could be request specific (and therefore
recoverable - e.g. if there is an error in input processing).
EngineDeadError is raised by the background output_handler
method. This error is global and therefore not recoverable.
We register these @app.exception_handlers to return nice
responses to the end user if they occur and shut down if needed.
See https://fastapi.tiangolo.com/tutorial/handling-errors/