Update CI/CD

Signed-off-by: SimJeg <sjegou@nvidia.com>

Author: SimJeg

.github/workflows/test.yml

@@ -5,23 +5,19 @@ on:
     branches:
       - main
       - "pull-request/[0-9]+"
+  workflow_dispatch:  # Allows manual triggering from any branch
 
 jobs:
   test:
     runs-on: linux-amd64-gpu-l4-latest-1
     steps:
       - uses: actions/checkout@v3
-      - name: Setup Python
-        uses: actions/setup-python@v4
-        with:
-          python-version: "3.12"
 
-      - name: Check CUDA runtime
+      - name: Verify environment
         run: |
           nvidia-smi
-          echo "Checking for CUDA libraries..."
-          ls /usr/local/cuda*/lib64/libcudart.so* 2>/dev/null || ls /usr/lib/x86_64-linux-gnu/libcuda.so* 2>/dev/null || echo "No CUDA libs found in standard paths"
-          ldconfig -p | grep -i cuda || true
+          python3 --version
+          which python3
 
       - name: Install uv
         uses: astral-sh/setup-uv@v6
@@ -30,8 +26,8 @@ jobs:
 
       - name: Install dependencies
         run: |
-          uv pip install torch==2.9.0
           uv sync --all-groups
+          uv pip install torch==2.10.0
 
       - run: make test
         env:

Makefile

@@ -42,7 +42,7 @@ reports:
 .PHONY: test
 test: reports
 	$(UV) pip install optimum-quanto
-	$(UV) pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.9cxx11abiTRUE-cp312-cp312-linux_x86_64.whl
+	$(UV) pip install flash-attn --no-build-isolation --find-links https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/expanded_assets/v0.7.12
 	PYTHONPATH=. \
 	$(UV) run pytest \
 		--cov-report xml:reports/coverage.xml \