Skip to content

GitLab

Unverified Commit 39d82005 authored by Rabi Mishra's avatar Rabi Mishra Committed by GitHub
Browse files

fix(rocm): add early return in get_flash_attn_version for ROCm (#31286) 


Signed-off-by: default avatarrabi <ramishra@redhat.com>
parent 25eef3dc
Hide whitespace changes
Inline Side-by-side
Showing with 3 additions and 0 deletions
vllm/attention/utils/fa_utils.py
View file @ 39d82005
...@@ -40,6 +40,9 @@ def get_flash_attn_version(requires_alibi: bool = False) -> int | None: ...@@ -40,6 +40,9 @@ def get_flash_attn_version(requires_alibi: bool = False) -> int | None:
if current_platform.is_xpu(): if current_platform.is_xpu():
return 2 return 2
if current_platform.is_rocm():
# ROCm doesn't use vllm_flash_attn; return None to skip fa_version arg
return None
try: try:
from vllm.vllm_flash_attn.flash_attn_interface import ( from vllm.vllm_flash_attn.flash_attn_interface import (
fa_version_unsupported_reason, fa_version_unsupported_reason,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment