From 5a24c8a2a23529cf83b3a79b69d95a24d4af0663 Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot <bot-opensource@localaisrl.com>
Date: Tue, 15 Sep 2026 15:30:23 +0200
Subject: [PATCH] fix(qwen-tts): restore CUDA 13 SDPA fallback (#12067)

The CUDA 13 profile resolves PyTorch 2.14. Its headers cannot build the
current FlashAttention source release under C++17, so the full backend
matrix cannot publish the image.

Leave FlashAttention optional on CUDA 13. The backend already retries
model loading with SDPA when FlashAttention is unavailable.

Assisted-by: Codex:gpt-5

Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
---
 backend/python/qwen-tts/install.sh                      | 7 -------
 backend/python/qwen-tts/requirements-cublas13-after.txt | 2 --
 backend/python/qwen-tts/test.sh                         | 9 ---------
 3 files changed, 18 deletions(-)
 delete mode 100644 backend/python/qwen-tts/requirements-cublas13-after.txt

diff --git a/backend/python/qwen-tts/install.sh b/backend/python/qwen-tts/install.sh
index 9c8bcf83f..b7d487873 100755
--- a/backend/python/qwen-tts/install.sh
+++ b/backend/python/qwen-tts/install.sh
@@ -10,11 +10,4 @@ else
     source $backend_dir/../common/libbackend.sh
 fi
 
-# CUDA 13 has no prebuilt FlashAttention wheel, so the fallback source build
-# exceeds the CI runner's memory when ninja compiles multiple units at once.
-if [ "x${BUILD_PROFILE}" = "xcublas13" ]; then
-    export MAX_JOBS="${MAX_JOBS:-1}"
-    export NVCC_THREADS="${NVCC_THREADS:-1}"
-fi
-
 installRequirements
diff --git a/backend/python/qwen-tts/requirements-cublas13-after.txt b/backend/python/qwen-tts/requirements-cublas13-after.txt
deleted file mode 100644
index c8ba665b2..000000000
--- a/backend/python/qwen-tts/requirements-cublas13-after.txt
+++ /dev/null
@@ -1,2 +0,0 @@
-ninja
-flash-attn
diff --git a/backend/python/qwen-tts/test.sh b/backend/python/qwen-tts/test.sh
index 97ea2faf9..eb59f2aaf 100755
--- a/backend/python/qwen-tts/test.sh
+++ b/backend/python/qwen-tts/test.sh
@@ -2,15 +2,6 @@
 set -e
 
 backend_dir=$(dirname $0)
-
-for cuda_version in 12 13; do
-    grep -qx "flash-attn" "$backend_dir/requirements-cublas${cuda_version}-after.txt"
-done
-
-grep -q 'BUILD_PROFILE.*cublas13' "$backend_dir/install.sh"
-grep -q 'MAX_JOBS.*1' "$backend_dir/install.sh"
-grep -q 'NVCC_THREADS.*1' "$backend_dir/install.sh"
-
 if [ -d $backend_dir/common ]; then
     source $backend_dir/common/libbackend.sh
 else