xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-23 23:14:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Core][Easy] Use envs.__getattr__ for all Unify to environment variable access (#26810)

Browse Source 
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
This commit is contained in:
Jialin Ouyang 2025-10-14 18:41:18 -07:00 committed by GitHub
parent 2dcd12d357
commit 07ca70af8d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 10 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/multimodal/cache.py
View File

@ -10,12 +10,12 @@ from typing import TYPE_CHECKING, Generic, TypeAlias, TypeVar, cast
import torch
from typing_extensions import override
import vllm.envs as envs
from vllm.distributed.device_communicators.shm_object_storage import (
MsgpackSerde,
SingleWriterShmObjectStorage,
SingleWriterShmRingBuffer,
)
from vllm.envs import VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME
from vllm.logger import init_logger
from vllm.utils import GiB_bytes, MiB_bytes
from vllm.utils.cache import CacheInfo, LRUCache
@ -436,7 +436,7 @@ class ShmObjectStoreSenderCache(BaseMultiModalProcessorCache):
ring_buffer = SingleWriterShmRingBuffer(
data_buffer_size=int(mm_config.mm_processor_cache_gb * GiB_bytes),
name=VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME,
name=envs.VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME,
create=True, # sender is the writer
)
self._shm_cache = SingleWriterShmObjectStorage(
@ -678,7 +678,7 @@ class ShmObjectStoreReceiverCache(BaseMultiModalReceiverCache):
ring_buffer = SingleWriterShmRingBuffer(
data_buffer_size=int(mm_config.mm_processor_cache_gb * GiB_bytes),
name=VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME,
name=envs.VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME,
create=False, # Server is a reader
)
self._shm_cache = SingleWriterShmObjectStorage(

4
vllm/transformers_utils/utils.py
View File

@ -8,7 +8,7 @@ from os import PathLike
from pathlib import Path
from typing import Any
from vllm.envs import VLLM_MODEL_REDIRECT_PATH
import vllm.envs as envs
from vllm.logger import init_logger
logger = init_logger(__name__)
@ -86,7 +86,7 @@ def maybe_model_redirect(model: str) -> str:
:return: maybe redirect to a local folder
"""
model_redirect_path = VLLM_MODEL_REDIRECT_PATH
model_redirect_path = envs.VLLM_MODEL_REDIRECT_PATH
if not model_redirect_path:
return model

6
vllm/utils/gc_utils.py
View File

@ -7,7 +7,7 @@ from collections import Counter
from contextlib import suppress
from typing import Any
from vllm.envs import VLLM_GC_DEBUG
import vllm.envs as envs
from vllm.logger import init_logger
logger = init_logger(__name__)
@ -36,7 +36,7 @@ class GCDebugConfig:
self.top_objects = json_conf.get("top_objects", -1)
except Exception:
self.enabled = False
logger.error("Failed to parse VLLM_GC_DEBUG(%s)", VLLM_GC_DEBUG)
logger.error("Failed to parse VLLM_GC_DEBUG(%s)", envs.VLLM_GC_DEBUG)
logger.info("GC Debug Config. %s", str(self))
def __repr__(self) -> str:
@ -93,7 +93,7 @@ def maybe_attach_gc_debug_callback() -> None:
"""
Attached a callback for GC debug when VLLM_GC_DEBUG is enabled.
"""
config = GCDebugConfig(VLLM_GC_DEBUG)
config = GCDebugConfig(envs.VLLM_GC_DEBUG)
if config.enabled:
debugger: GCDebugger = GCDebugger(config)

5
vllm/v1/engine/async_llm.py
View File

@ -16,7 +16,6 @@ from vllm.config import VllmConfig
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.engine.protocol import EngineClient
from vllm.entrypoints.utils import _validate_truncation_size
from vllm.envs import VLLM_V1_OUTPUT_PROC_CHUNK_SIZE
from vllm.inputs import PromptType
from vllm.logger import init_logger
from vllm.lora.request import LoRARequest
@ -483,12 +482,12 @@ class AsyncLLM(EngineClient):
# Split outputs into chunks of at most
# VLLM_V1_OUTPUT_PROC_CHUNK_SIZE, so that we don't block the
# event loop for too long.
if num_outputs <= VLLM_V1_OUTPUT_PROC_CHUNK_SIZE:
if num_outputs <= envs.VLLM_V1_OUTPUT_PROC_CHUNK_SIZE:
slices = (outputs.outputs,)
else:
slices = np.array_split(
outputs.outputs,
cdiv(num_outputs, VLLM_V1_OUTPUT_PROC_CHUNK_SIZE),
cdiv(num_outputs, envs.VLLM_V1_OUTPUT_PROC_CHUNK_SIZE),
)
for i, outputs_slice in enumerate(slices):