xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 03:26:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

use_aot_compile should respect VLLM_DISABLE_COMPILE_CACHE (#27698)

Browse Source 
Signed-off-by: Boyuan Feng <boyuan@meta.com>
This commit is contained in:
Boyuan Feng 2025-10-29 10:08:54 -07:00 committed by GitHub
parent 7568a282b9
commit a9fe0793f2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 11 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/envs.py
View File

@ -247,10 +247,19 @@ def maybe_convert_bool(value: str | None) -> bool | None:
return bool(int(value))
def disable_compile_cache() -> bool:
return bool(int(os.getenv("VLLM_DISABLE_COMPILE_CACHE", "0")))
def use_aot_compile() -> bool:
from vllm.utils.torch_utils import is_torch_equal_or_newer
default_value = "1" if is_torch_equal_or_newer("2.10.0.dev") else "0"
default_value = (
"1"
if is_torch_equal_or_newer("2.10.0.dev") and not disable_compile_cache()
else "0"
)
return os.environ.get("VLLM_USE_AOT_COMPILE", default_value) == "1"
@ -963,9 +972,7 @@ environment_variables: dict[str, Callable[[], Any]] = {
"VLLM_LOG_BATCHSIZE_INTERVAL": lambda: float(
os.getenv("VLLM_LOG_BATCHSIZE_INTERVAL", "-1")
),
"VLLM_DISABLE_COMPILE_CACHE": lambda: bool(
int(os.getenv("VLLM_DISABLE_COMPILE_CACHE", "0"))
),
"VLLM_DISABLE_COMPILE_CACHE": disable_compile_cache,
# If set, vllm will run in development mode, which will enable
# some additional endpoints for developing and debugging,
# e.g. `/reset_prefix_cache`