xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-21 07:25:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] use flash attn on sm90 (#22933)

Browse Source 
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Yongye Zhu 2025-08-14 19:37:22 -04:00 committed by GitHub
parent 919234fe17
commit 39cd09dc86
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/platforms/cuda.py
View File

@ -316,7 +316,7 @@ class CudaPlatformBase(Platform):
# FlashAttention is the default for SM 8.0+ GPUs
if cls.has_device_capability(80):
if has_sink:
if has_sink and not cls.is_device_capability(90):
logger.info_once("Using Triton backend on V1 engine.")
return TRITON_ATTN_VLLM_V1
if is_default_backend_supported := is_attn_backend_supported(