EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix sgl_kernel import failure on devices other than CUDA (#10610)

Browse Source
This commit is contained in:
Zaili Wang
2025-09-19 02:38:02 +08:00
committed by GitHub
parent 1344ebc833
commit 6fd4816d9f
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docs/platforms/cpu_server.md
View File

@@ -63,7 +63,7 @@ is required to enable SGLang service with CPU engine.
conda create -n sgl-cpu python=3.12 -y conda create -n sgl-cpu python=3.12 -y
conda activate sgl-cpu conda activate sgl-cpu
# Optional: Set PyTorch CPU as primary pip install channel to avoid installing CUDA version # Set PyTorch CPU as primary pip install channel to avoid installing the larger CUDA-enabled version and prevent potential runtime issues.
pip config set global.index-url https://download.pytorch.org/whl/cpu pip config set global.index-url https://download.pytorch.org/whl/cpu
pip config set global.extra-index-url https://pypi.org/simple pip config set global.extra-index-url https://pypi.org/simple

2
sgl-kernel/python/sgl_kernel/__init__.py
View File

@@ -23,7 +23,7 @@ def _find_cuda_home():
return cuda_home return cuda_home
if torch.version.hip is None: if torch.version.cuda is not None:
cuda_home = Path(_find_cuda_home()) cuda_home = Path(_find_cuda_home())
if (cuda_home / "lib").is_dir(): if (cuda_home / "lib").is_dir():