xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-09 06:51:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

Disable the torch.compile cache checks when VLLM_DISABLE_COMPILE_CACHE=1 (#16573)

Browse Source 
Signed-off-by: Lu Fang <lufang@fb.com>
This commit is contained in:
Lu Fang 2025-04-26 09:17:58 -07:00 committed by GitHub
parent 10fd1d7380
commit 4d17e20310
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
vllm/compilation/compiler_interface.py
View File

@ -11,6 +11,7 @@ import torch
import torch._inductor.compile_fx
import torch.fx as fx
import vllm.envs as envs
from vllm.config import VllmConfig
from vllm.utils import is_torch_equal_or_newer
@ -317,10 +318,14 @@ class InductorAdaptor(CompilerInterface):
inner_compile=hijacked_compile_fx_inner,
config_patches=current_config)
assert hash_str is not None, (
"failed to get the hash of the compiled graph")
assert file_path is not None, (
"failed to get the file path of the compiled graph")
# We treat VLLM_DISABLE_COMPILE_CACHE as the overall switch for torch
# compilation cache. So turn off the checks if we disable the
# compilation cache.
if not envs.VLLM_DISABLE_COMPILE_CACHE:
assert hash_str is not None, (
"failed to get the hash of the compiled graph")
assert file_path is not None, (
"failed to get the file path of the compiled graph")
return compiled_graph, (hash_str, file_path)
def load(self,