xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-19 12:17:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ROCm] remove unsupported archs from rocm triton flash-attention supported list (#17536)

Browse Source 
Signed-off-by: Hongxia Yang <hongxia.yang@amd.com>
This commit is contained in:
Hongxia Yang 2025-05-01 10:54:25 -04:00 committed by GitHub
parent 98060b001d
commit 28566d73b3
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/attention/ops/triton_flash_attention.py
View File

@ -650,7 +650,7 @@ def get_general_autotune_configs():
def has_cdna_target(): def has_cdna_target():
ROCM_CDNA_TARGETS = ["gfx940", "gfx941", "gfx942", "gfx90a", "gfx908"] ROCM_CDNA_TARGETS = ["gfx942", "gfx90a", "gfx908"]
return triton.runtime.driver.active.get_current_target( return triton.runtime.driver.active.get_current_target(
).arch in ROCM_CDNA_TARGETS ).arch in ROCM_CDNA_TARGETS