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

[V1][Metrics] Allow V1 AsyncLLM to use custom logger (#14661)

Browse Source 
Signed-off-by: Zijing Liu <liuzijing2014@gmail.com>
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Mark McLoughlin <markmc@redhat.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Zijing Liu
2025-04-25 22:05:40 -07:00
committed by GitHub
parent 54271bb766
commit 53e8cf53a4
4 changed files with 118 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
vllm/v1/engine/async_llm.py
View File

@@ -1,6 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
import asyncio
import logging
from collections.abc import AsyncGenerator, Mapping
from copy import copy
from typing import Optional, Union
@@ -33,8 +32,8 @@ from vllm.v1.engine.output_processor import (OutputProcessor,
from vllm.v1.engine.parallel_sampling import ParentRequest
from vllm.v1.engine.processor import Processor
from vllm.v1.executor.abstract import Executor
from vllm.v1.metrics.loggers import (LoggingStatLogger, PrometheusStatLogger,
StatLoggerBase)
from vllm.v1.metrics.loggers import (StatLoggerBase, StatLoggerFactory,
setup_default_loggers)
from vllm.v1.metrics.stats import IterationStats, SchedulerStats
logger = init_logger(__name__)
@@ -52,7 +51,28 @@ class AsyncLLM(EngineClient):
use_cached_outputs: bool = False,
log_requests: bool = True,
start_engine_loop: bool = True,
stat_loggers: Optional[list[StatLoggerFactory]] = None,
) -> None:
"""
Create an AsyncLLM.
Args:
vllm_config: global configuration.
executor_class: an Executor impl, e.g. MultiprocExecutor.
log_stats: Whether to log stats.
usage_context: Usage context of the LLM.
mm_registry: Multi-modal registry.
use_cached_outputs: Whether to use cached outputs.
log_requests: Whether to log requests.
start_engine_loop: Whether to start the engine loop.
stat_loggers: customized stat loggers for the engine.
If not provided, default stat loggers will be used.
PLEASE BE AWARE THAT STAT LOGGER IS NOT STABLE
IN V1, AND ITS BASE CLASS INTERFACE MIGHT CHANGE.
Returns:
None
"""
if not envs.VLLM_USE_V1:
raise ValueError(
"Using V1 AsyncLLMEngine, but envs.VLLM_USE_V1=False. "
@@ -66,15 +86,12 @@ class AsyncLLM(EngineClient):
self.log_stats = log_stats
# Set up stat loggers; independent set for each DP rank.
self.stat_loggers: list[list[StatLoggerBase]] = []
if self.log_stats:
for i in range(vllm_config.parallel_config.data_parallel_size):
loggers: list[StatLoggerBase] = []
if logger.isEnabledFor(logging.INFO):
loggers.append(LoggingStatLogger(engine_index=i))
loggers.append(
PrometheusStatLogger(vllm_config, engine_index=i))
self.stat_loggers.append(loggers)
self.stat_loggers: list[list[StatLoggerBase]] = setup_default_loggers(
vllm_config=vllm_config,
log_stats=self.log_stats,
engine_num=vllm_config.parallel_config.data_parallel_size,
custom_stat_loggers=stat_loggers,
)
# Tokenizer (+ ensure liveness if running in another process).
self.tokenizer = init_tokenizer_from_configs(
@@ -118,7 +135,7 @@ class AsyncLLM(EngineClient):
vllm_config: VllmConfig,
start_engine_loop: bool = True,
usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
stat_loggers: Optional[dict[str, StatLoggerBase]] = None,
stat_loggers: Optional[list[StatLoggerFactory]] = None,
disable_log_requests: bool = False,
disable_log_stats: bool = False,
) -> "AsyncLLM":
@@ -129,17 +146,12 @@ class AsyncLLM(EngineClient):
"AsyncLLMEngine.from_vllm_config(...) or explicitly set "
"VLLM_USE_V1=0 or 1 and report this issue on Github.")
# FIXME(rob): refactor VllmConfig to include the StatLoggers
# include StatLogger in the Oracle decision.
if stat_loggers is not None:
raise ValueError("Custom StatLoggers are not yet supported on V1. "
"Explicitly set VLLM_USE_V1=0 to disable V1.")
# Create the LLMEngine.
return cls(
vllm_config=vllm_config,
executor_class=Executor.get_class(vllm_config),
start_engine_loop=start_engine_loop,
stat_loggers=stat_loggers,
log_requests=not disable_log_requests,
log_stats=not disable_log_stats,
usage_context=usage_context,
@@ -151,6 +163,7 @@ class AsyncLLM(EngineClient):
engine_args: AsyncEngineArgs,
start_engine_loop: bool = True,
usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
stat_loggers: Optional[list[StatLoggerFactory]] = None,
) -> "AsyncLLM":
"""Create an AsyncLLM from the EngineArgs."""
@@ -166,6 +179,7 @@ class AsyncLLM(EngineClient):
log_stats=not engine_args.disable_log_stats,
start_engine_loop=start_engine_loop,
usage_context=usage_context,
stat_loggers=stat_loggers,
)
def __del__(self):