Skip to content

Commit 1952ca2

Browse files
fix(qwen-tts): install flash-attn on CUDA 13
Add the missing cublas13 post-install requirements hook so the CUDA 13 backend uses FlashAttention instead of falling back to SDPA. Keep CUDA packaging parity covered by the backend test script. Assisted-by: Codex:gpt-5
1 parent fb54d0f commit 1952ca2

2 files changed

Lines changed: 6 additions & 0 deletions

File tree

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1 @@
1+
flash-attn

backend/python/qwen-tts/test.sh

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2,6 +2,11 @@
22
set -e
33

44
backend_dir=$(dirname $0)
5+
6+
for cuda_version in 12 13; do
7+
grep -qx "flash-attn" "$backend_dir/requirements-cublas${cuda_version}-after.txt"
8+
done
9+
510
if [ -d $backend_dir/common ]; then
611
source $backend_dir/common/libbackend.sh
712
else

0 commit comments

Comments
 (0)