biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Misc][XPU] Use None as device capacity for XPU (#14932)

Browse Source 
Signed-off-by: yan ma <yan.ma@intel.com>
This commit is contained in:
Yan Ma
2025-03-17 16:22:14 +08:00
committed by GitHub
parent b539222d4e
commit 9b87a579aa
1 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/platforms/xpu.py
View File

@@ -37,10 +37,11 @@ class XPUPlatform(Platform):
return "vllm.attention.backends.ipex_attn.IpexAttnBackend" return "vllm.attention.backends.ipex_attn.IpexAttnBackend"
@staticmethod @staticmethod
def get_device_capability(device_id: int = 0) -> DeviceCapability: def get_device_capability(
major, minor, *_ = torch.xpu.get_device_capability( device_id: int = 0) -> Optional[DeviceCapability]:
device_id)['version'].split('.') # capacity format differs from cuda's and will cause unexpected
return DeviceCapability(major=int(major), minor=int(minor)) # failure, so use None directly
return None
@staticmethod @staticmethod
def get_device_name(device_id: int = 0) -> str: def get_device_name(device_id: int = 0) -> str: