fix: pre-commit

Signed-off-by: Mickael Seznec <[email protected]>

Author: mickaelseznec

setup.py

@@ -187,7 +187,8 @@ def configure(self, ext: CMakeExtension) -> None:
         # 12.6 is known to segfault when compiling flash attention,
         # instead, hack toolchain to use ptxas from 12.8
         # https://github.com/Dao-AILab/flash-attention/blob/main/hopper/setup.py#L404
-        override_files = utils.maybe_patch_compiler(get_nvcc_cuda_version() if _is_cuda() else None)
+        override_files = utils.maybe_patch_compiler(
+            get_nvcc_cuda_version() if _is_cuda() else None)
 
         with utils.OverrideFiles(override_files):
             subprocess.check_call(

vllm/setup_utils.py

@@ -1,7 +1,11 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import os
+import platform
+import tarfile
 from pathlib import Path
 from shutil import copy, copytree, which
 from tempfile import TemporaryDirectory
-import tarfile
 from typing import Optional
 
 from packaging.version import Version
@@ -81,7 +85,8 @@ def download_toolchain(nvcc_version: str, ptxas_version: str, dst_path: Path):
     )
 
 
-def maybe_patch_compiler(cuda_version: Optional[Version] = None) -> dict[Path, Path]:
+def maybe_patch_compiler(
+        cuda_version: Optional[Version] = None) -> dict[Path, Path]:
     if cuda_version is None or cuda_version != Version("12.6"):
         return {}
     assert CUDA_HOME is not None, "CUDA_HOME is not set"
@@ -110,5 +115,3 @@ def __exit__(self, exc_type, exc_value, traceback):
         for target in self.override_map:
             target.unlink()
             target.with_suffix('.backup').rename(target)
-
-