A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 18:30 UTC
[Misc][XPU] Use None as device capacity for XPU (#14932)
Signed-off-by: yan ma <yan.ma@intel.com>
Yan Ma committed
Mar 17, 2025 at 16:22 UTC
9b87a579aaf82338d5304219350932abae9b19ac
1 file changed
+5
-4
vllm/platforms/xpu.py
+5
-4
@@ -37,10 +37,11 @@ class XPUPlatform(Platform):
37
return "vllm.attention.backends.ipex_attn.IpexAttnBackend"
38
39
@staticmethod
40
- def get_device_capability(device_id: int = 0) -> DeviceCapability:
41
- major, minor, *_ = torch.xpu.get_device_capability(
42
- device_id)['version'].split('.')
43
- return DeviceCapability(major=int(major), minor=int(minor))
40
+ def get_device_capability(
41
+ device_id: int = 0) -> Optional[DeviceCapability]:
42
+ # capacity format differs from cuda's and will cause unexpected
43
+ # failure, so use None directly
44
+ return None
45
46
@staticmethod
47
def get_device_name(device_id: int = 0) -> str: