xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-16 05:35:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[XPU]Fix crash due to removed VLLM_USE_V1 attribute (#28520)

Browse Source 
Signed-off-by: chaojun-zhang <chaojun.zhang@intel.com>
This commit is contained in:
Chaojun Zhang 2025-11-12 18:20:55 +08:00 committed by GitHub
parent d3ade61e42
commit a4730c1b4f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/platforms/xpu.py
View File

@ -65,7 +65,6 @@ class XPUPlatform(Platform):
if use_sparse: if use_sparse:
raise NotImplementedError("Sparse Attention is not supported on XPU.") raise NotImplementedError("Sparse Attention is not supported on XPU.")
use_v1 = envs.VLLM_USE_V1
if not use_v1: if not use_v1:
raise ValueError("XPU backend only supports V1.") raise ValueError("XPU backend only supports V1.")
if selected_backend == AttentionBackendEnum.TRITON_ATTN: if selected_backend == AttentionBackendEnum.TRITON_ATTN:
@ -115,7 +114,9 @@ class XPUPlatform(Platform):
@classmethod @classmethod
def get_vit_attn_backend( def get_vit_attn_backend(
cls, head_size: int, dtype: torch.dtype cls, head_size: int, dtype: torch.dtype
) -> AttentionBackendEnum: ) -> "AttentionBackendEnum":
from vllm.attention.backends.registry import AttentionBackendEnum
return AttentionBackendEnum.FLASH_ATTN return AttentionBackendEnum.FLASH_ATTN
@classmethod @classmethod