Add pydocstyle to the ruff rules (#52136)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Signed-off-by: Harry Mellor <hej.mellor@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
+4
-6
@@ -23,8 +23,7 @@ async def reset_prefix_cache(
|
||||
reset_running_requests: bool = Query(default=False),
|
||||
reset_external: bool = Query(default=False),
|
||||
):
|
||||
"""
|
||||
Reset the local prefix cache.
|
||||
"""Reset the local prefix cache.
|
||||
|
||||
Optionally, if the query parameter `reset_external=true`
|
||||
also resets the external (connector-managed) prefix cache.
|
||||
@@ -35,6 +34,7 @@ async def reset_prefix_cache(
|
||||
|
||||
Example:
|
||||
POST /reset_prefix_cache?reset_external=true
|
||||
|
||||
"""
|
||||
logger.info("Resetting prefix cache...")
|
||||
|
||||
@@ -46,8 +46,7 @@ async def reset_prefix_cache(
|
||||
|
||||
@router.post("/reset_mm_cache")
|
||||
async def reset_mm_cache(raw_request: Request):
|
||||
"""
|
||||
Reset the multi-modal cache. Note that we currently do not check if the
|
||||
"""Reset the multi-modal cache. Note that we currently do not check if the
|
||||
multi-modal cache is successfully reset in the API server.
|
||||
"""
|
||||
logger.info("Resetting multi-modal cache...")
|
||||
@@ -57,8 +56,7 @@ async def reset_mm_cache(raw_request: Request):
|
||||
|
||||
@router.post("/reset_encoder_cache")
|
||||
async def reset_encoder_cache(raw_request: Request):
|
||||
"""
|
||||
Reset the encoder cache. Note that we currently do not check if the
|
||||
"""Reset the encoder cache. Note that we currently do not check if the
|
||||
encoder cache is successfully reset in the API server.
|
||||
"""
|
||||
logger.info("Resetting encoder cache...")
|
||||
|
||||
@@ -36,6 +36,8 @@ async def pause_generation(
|
||||
"""Pause generation requests to allow weight updates.
|
||||
|
||||
Args:
|
||||
raw_request: The incoming FastAPI request, used to reach the engine
|
||||
client on the app state.
|
||||
mode: How to handle in-flight requests:
|
||||
- ``"abort"``: Abort all in-flight requests immediately (default).
|
||||
- ``"wait"``: Wait for in-flight requests to complete.
|
||||
@@ -43,8 +45,8 @@ async def pause_generation(
|
||||
wait_for_inflight_requests: DEPRECATED. Use ``mode="wait"`` instead.
|
||||
clear_cache: DEPRECATED. Whether to clear KV/prefix caches after
|
||||
draining. Ignored when mode="keep".
|
||||
"""
|
||||
|
||||
"""
|
||||
engine = engine_client(raw_request)
|
||||
|
||||
try:
|
||||
@@ -74,7 +76,6 @@ async def pause_generation(
|
||||
@router.post("/resume")
|
||||
async def resume_generation(raw_request: Request) -> JSONResponse:
|
||||
"""Resume generation after a pause."""
|
||||
|
||||
engine = engine_client(raw_request)
|
||||
|
||||
try:
|
||||
@@ -97,7 +98,6 @@ async def abort_requests(raw_request: Request) -> JSONResponse:
|
||||
|
||||
Empty/missing ``request_ids`` aborts all in-flight requests.
|
||||
"""
|
||||
|
||||
engine = engine_client(raw_request)
|
||||
|
||||
try:
|
||||
@@ -139,7 +139,6 @@ async def abort_requests(raw_request: Request) -> JSONResponse:
|
||||
@router.get("/is_paused")
|
||||
async def is_paused(raw_request: Request) -> JSONResponse:
|
||||
"""Return the current pause status."""
|
||||
|
||||
engine = engine_client(raw_request)
|
||||
|
||||
try:
|
||||
@@ -234,9 +233,12 @@ async def get_world_size(
|
||||
"""Get the world size from the parallel config.
|
||||
|
||||
Args:
|
||||
raw_request: The incoming FastAPI request, used to reach the engine
|
||||
client on the app state.
|
||||
include_dp: If True (default), returns the world size including
|
||||
data parallelism (TP * PP * DP). If False, returns the world
|
||||
size without data parallelism (TP * PP).
|
||||
|
||||
"""
|
||||
parallel_config = engine_client(raw_request).vllm_config.parallel_config
|
||||
if include_dp:
|
||||
|
||||
@@ -20,8 +20,7 @@ def set_scaling_elastic_ep(value):
|
||||
|
||||
|
||||
class ScalingMiddleware:
|
||||
"""
|
||||
Middleware that checks if the model is currently scaling and
|
||||
"""Middleware that checks if the model is currently scaling and
|
||||
returns a 503 Service Unavailable response if it is.
|
||||
|
||||
This middleware applies to all HTTP requests and prevents
|
||||
|
||||
@@ -117,7 +117,7 @@ class BaseServing:
|
||||
def _base_request_id(
|
||||
raw_request: Request | None, default: str | None = None
|
||||
) -> str | None:
|
||||
"""Pulls the request id to use from a header, if provided"""
|
||||
"""Pulls the request id to use from a header, if provided."""
|
||||
if raw_request is not None and (
|
||||
(req_id := raw_request.headers.get("X-Request-Id")) is not None
|
||||
):
|
||||
|
||||
@@ -126,7 +126,8 @@ def _format_error(err: object) -> str:
|
||||
|
||||
def _is_internal_loc_segment(segment: str) -> bool:
|
||||
"""True if `segment` is a Pydantic-internal wrapper/union-branch
|
||||
marker rather than a user-meaningful field name or list index."""
|
||||
marker rather than a user-meaningful field name or list index.
|
||||
"""
|
||||
if _BRACKETED_INTERNAL_RE.search(segment):
|
||||
return True
|
||||
return segment.lower() in _INTERNAL_LOC_MARKERS
|
||||
|
||||
@@ -28,8 +28,7 @@ async def vllm_error_handler(req: Request, exc: VLLMError):
|
||||
async def engine_error_handler(
|
||||
req: Request, exc: EngineDeadError | EngineGenerateError
|
||||
):
|
||||
"""
|
||||
VLLM V1 AsyncLLM catches exceptions and returns
|
||||
"""VLLM V1 AsyncLLM catches exceptions and returns
|
||||
only two types: EngineGenerateError and EngineDeadError.
|
||||
|
||||
EngineGenerateError is raised by the per request generate()
|
||||
@@ -51,7 +50,6 @@ async def engine_error_handler(
|
||||
will not automatically shut down. Instead, we use the watchdog
|
||||
background task for check for errored state.
|
||||
"""
|
||||
|
||||
if req.app.state.args.log_error_stack:
|
||||
logger.exception(
|
||||
"Engine Exception caught. Request id: %s",
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
vLLM Exception handlers are registered in four layers:
|
||||
"""vLLM Exception handlers are registered in four layers:
|
||||
1. framework errors raised by FastAPI/Starlette
|
||||
2. vLLM-specific errors dispatched via a single ``VLLMError`` handler
|
||||
3. fallback handlers for raw exceptions not yet migrated to ``VLLMError``
|
||||
|
||||
@@ -55,7 +55,6 @@ class PrometheusResponse(Response):
|
||||
|
||||
def attach_router(app: FastAPI):
|
||||
"""Mount prometheus metrics to a FastAPI app."""
|
||||
|
||||
registry = get_prometheus_registry()
|
||||
|
||||
# `response_class=PrometheusResponse` is needed to return an HTTP response
|
||||
|
||||
@@ -63,7 +63,8 @@ def _attach_router(app: FastAPI):
|
||||
@router.post("/v1/load_lora_adapter", dependencies=[Depends(validate_json_request)])
|
||||
async def load_lora_adapter(request: LoadLoRAAdapterRequest, raw_request: Request):
|
||||
"""Handle POST /v1/load_lora_adapter: load a LoRA adapter into the
|
||||
serving engine."""
|
||||
serving engine.
|
||||
"""
|
||||
handler: OpenAIServingModels = models(raw_request)
|
||||
response = await handler.load_lora_adapter(request)
|
||||
if isinstance(response, ErrorResponse):
|
||||
@@ -85,7 +86,8 @@ def _attach_router(app: FastAPI):
|
||||
request: UnloadLoRAAdapterRequest, raw_request: Request
|
||||
):
|
||||
"""Handle POST /v1/unload_lora_adapter: unload a LoRA adapter from
|
||||
the serving engine."""
|
||||
the serving engine.
|
||||
"""
|
||||
handler: OpenAIServingModels = models(raw_request)
|
||||
response = await handler.unload_lora_adapter(request)
|
||||
if isinstance(response, ErrorResponse):
|
||||
|
||||
@@ -12,8 +12,7 @@ GUARDED_PREFIX = ("/v1", "/v2", "/inference", "/cohere")
|
||||
|
||||
|
||||
class AuthenticationMiddleware:
|
||||
"""
|
||||
Pure ASGI middleware that authenticates each request by checking
|
||||
"""Pure ASGI middleware that authenticates each request by checking
|
||||
if the Authorization Bearer token exists and equals anyof "{api_key}".
|
||||
|
||||
Notes
|
||||
@@ -21,6 +20,7 @@ class AuthenticationMiddleware:
|
||||
There are two cases in which authentication is skipped:
|
||||
1. The HTTP method is OPTIONS.
|
||||
2. The request path doesn't start with GUARDED_PREFIX (e.g. /health).
|
||||
|
||||
"""
|
||||
|
||||
def __init__(self, app: ASGIApp, tokens: list[str]) -> None:
|
||||
|
||||
@@ -94,7 +94,6 @@ def _extract_content_from_chunk(chunk_data: dict) -> str:
|
||||
|
||||
def _log_streaming_response(response, response_body: list) -> None:
|
||||
"""Log streaming response with robust SSE parsing."""
|
||||
|
||||
sse_decoder = SSEDecoder()
|
||||
chunk_count = 0
|
||||
|
||||
|
||||
@@ -8,8 +8,7 @@ from starlette.types import ASGIApp, Message, Receive, Scope, Send
|
||||
|
||||
|
||||
class XRequestIdMiddleware:
|
||||
"""
|
||||
Middleware the set's the X-Request-Id header for each response
|
||||
"""Middleware the set's the X-Request-Id header for each response
|
||||
to a random uuid4 (hex) value if the header isn't already
|
||||
present in the request, otherwise use the provided request id.
|
||||
"""
|
||||
@@ -25,8 +24,7 @@ class XRequestIdMiddleware:
|
||||
request_headers = Headers(scope=scope)
|
||||
|
||||
async def send_with_request_id(message: Message) -> None:
|
||||
"""
|
||||
Custom send function to mutate the response headers
|
||||
"""Custom send function to mutate the response headers
|
||||
and append X-Request-Id to it.
|
||||
"""
|
||||
if message["type"] == "http.response.start":
|
||||
|
||||
@@ -30,7 +30,8 @@ EndpointFn = Callable[[RequestType, Request], Awaitable[Any]]
|
||||
def _snapshot_handler_levels() -> list[tuple[logging.Handler, int]]:
|
||||
"""Snapshot handler levels for loggers that may be affected by third-party
|
||||
logging configuration side effects (e.g. model_hosting_container_standards
|
||||
calling configure_root_logger() at import time)."""
|
||||
calling configure_root_logger() at import time).
|
||||
"""
|
||||
loggers = [logging.getLogger(), logging.getLogger("vllm")]
|
||||
seen: set[int] = set()
|
||||
snapshot: list[tuple[logging.Handler, int]] = []
|
||||
@@ -73,7 +74,8 @@ def _attach_router(
|
||||
model_config: ModelConfig | None = None,
|
||||
):
|
||||
"""Register the SageMaker hosting routes (/ping, /invocations) on the
|
||||
app."""
|
||||
app.
|
||||
"""
|
||||
import model_hosting_container_standards.sagemaker as sagemaker_standards
|
||||
|
||||
router = APIRouter()
|
||||
@@ -92,7 +94,7 @@ def _attach_router(
|
||||
@router.get("/ping", response_class=Response)
|
||||
@sagemaker_standards.register_ping_handler
|
||||
async def ping(raw_request: Request) -> Response:
|
||||
"""Ping check. Endpoint required for SageMaker"""
|
||||
"""Ping check. Endpoint required for SageMaker."""
|
||||
return await health(raw_request)
|
||||
|
||||
@router.post(
|
||||
|
||||
@@ -181,8 +181,7 @@ class DetokenizeResponse(OpenAIBaseModel):
|
||||
|
||||
|
||||
class TokenizerInfoResponse(OpenAIBaseModel):
|
||||
"""
|
||||
Response containing tokenizer configuration
|
||||
"""Response containing tokenizer configuration
|
||||
equivalent to tokenizer_config.json
|
||||
"""
|
||||
|
||||
|
||||
@@ -34,7 +34,7 @@ VLLM_SUBCMD_PARSER_EPILOG = (
|
||||
|
||||
|
||||
async def listen_for_disconnect(request: Request) -> None:
|
||||
"""Returns if a disconnect message is received"""
|
||||
"""Returns if a disconnect message is received."""
|
||||
while True:
|
||||
message = await request.receive()
|
||||
if message["type"] == "http.disconnect":
|
||||
|
||||
@@ -33,7 +33,8 @@ def set_default_fingerprint_mode(
|
||||
custom_value: str | None = None,
|
||||
) -> None:
|
||||
"""Configure the fingerprint mode for subsequent ``get_system_fingerprint``
|
||||
calls. Called once at server startup."""
|
||||
calls. Called once at server startup.
|
||||
"""
|
||||
global _DEFAULT_MODE, _CUSTOM_VALUE
|
||||
_DEFAULT_MODE = mode
|
||||
_CUSTOM_VALUE = custom_value
|
||||
@@ -41,7 +42,8 @@ def set_default_fingerprint_mode(
|
||||
|
||||
def get_system_fingerprint(vllm_config: Any) -> str | None:
|
||||
"""Return the fingerprint for ``vllm_config`` using the mode configured by
|
||||
``set_default_fingerprint_mode``."""
|
||||
``set_default_fingerprint_mode``.
|
||||
"""
|
||||
return build_system_fingerprint(vllm_config, _DEFAULT_MODE, _CUSTOM_VALUE)
|
||||
|
||||
|
||||
|
||||
@@ -1,8 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Utility functions that create ORCA endpoint load report response headers.
|
||||
"""
|
||||
"""Utility functions that create ORCA endpoint load report response headers."""
|
||||
|
||||
import json
|
||||
from collections.abc import Mapping
|
||||
@@ -16,23 +14,22 @@ logger = init_logger(__name__)
|
||||
def create_orca_header(
|
||||
metrics_format: str, named_metrics: list[tuple[str, float]]
|
||||
) -> Mapping[str, str] | None:
|
||||
"""
|
||||
Creates ORCA headers named 'endpoint-load-metrics' in the specified format
|
||||
"""Creates ORCA headers named 'endpoint-load-metrics' in the specified format
|
||||
and adds custom metrics to named_metrics.
|
||||
ORCA headers format description: https://docs.google.com/document/d/1C1ybMmDKJIVlrbOLbywhu9iRYo4rilR-cT50OTtOFTs/edit?tab=t.0
|
||||
ORCA proto https://github.com/cncf/xds/blob/main/xds/data/orca/v3/orca_load_report.proto
|
||||
|
||||
Parameters:
|
||||
- metrics_format (str): The format of the header ('TEXT', 'JSON').
|
||||
- named_metrics (List[Tuple[str, float]]): List of tuples with metric names
|
||||
and their corresponding double values.
|
||||
Args:
|
||||
metrics_format (str): The format of the header ('TEXT', 'JSON').
|
||||
named_metrics (List[Tuple[str, float]]): List of tuples with metric names
|
||||
and their corresponding double values.
|
||||
|
||||
Returns:
|
||||
- Optional[Mapping[str,str]]: A dictionary with header key as
|
||||
'endpoint-load-metrics' and values as the ORCA header strings with
|
||||
format prefix and data in with named_metrics in.
|
||||
"""
|
||||
|
||||
"""
|
||||
if metrics_format.lower() not in ["text", "json"]:
|
||||
logger.warning(
|
||||
"Warning: `%s` format is not supported in the ORCA response header",
|
||||
@@ -71,15 +68,17 @@ def create_orca_header(
|
||||
|
||||
|
||||
def get_named_metrics_from_prometheus() -> list[tuple[str, float]]:
|
||||
"""
|
||||
Collects current metrics from Prometheus and returns some of them
|
||||
"""Collects current metrics from Prometheus and returns some of them
|
||||
in the form of the `named_metrics` list for `create_orca_header()`.
|
||||
|
||||
Parameters:
|
||||
Parameters
|
||||
----------
|
||||
- None
|
||||
|
||||
Returns:
|
||||
Returns
|
||||
-------
|
||||
- list[tuple[str, float]]: List of tuples of metric names and their values.
|
||||
|
||||
"""
|
||||
named_metrics: list[tuple[str, float]] = []
|
||||
# Map from prometheus metric names to ORCA named metrics.
|
||||
@@ -98,20 +97,20 @@ def get_named_metrics_from_prometheus() -> list[tuple[str, float]]:
|
||||
|
||||
|
||||
def metrics_header(metrics_format: str) -> Mapping[str, str] | None:
|
||||
"""
|
||||
Creates ORCA headers named 'endpoint-load-metrics' in the specified format.
|
||||
"""Creates ORCA headers named 'endpoint-load-metrics' in the specified format.
|
||||
Metrics are collected from Prometheus using `get_named_metrics_from_prometheus()`.
|
||||
|
||||
ORCA headers format description: https://docs.google.com/document/d/1C1ybMmDKJIVlrbOLbywhu9iRYo4rilR-cT50OTtOFTs/edit?tab=t.0
|
||||
ORCA proto https://github.com/cncf/xds/blob/main/xds/data/orca/v3/orca_load_report.proto
|
||||
|
||||
Parameters:
|
||||
- metrics_format (str): The format of the header ('TEXT', 'JSON').
|
||||
Args:
|
||||
metrics_format (str): The format of the header ('TEXT', 'JSON').
|
||||
|
||||
Returns:
|
||||
- Optional[Mapping[str,str]]: A dictionary with header key as
|
||||
'endpoint-load-metrics' and values as the ORCA header strings with
|
||||
format prefix and data in with named_metrics in.
|
||||
|
||||
"""
|
||||
if not metrics_format:
|
||||
return None
|
||||
|
||||
@@ -20,7 +20,6 @@ def maybe_filter_parallel_tool_calls(
|
||||
choice: _ChatCompletionResponseChoiceT, request: ChatCompletionRequest
|
||||
) -> _ChatCompletionResponseChoiceT:
|
||||
"""Filter to first tool call only when parallel_tool_calls is explicitly False."""
|
||||
|
||||
if request.parallel_tool_calls is not False:
|
||||
return choice
|
||||
|
||||
|
||||
Reference in New Issue
Block a user