From 23fe09621081e1127a9d510607209084131a004c Mon Sep 17 00:00:00 2001 From: Junya Morioka Date: Sat, 17 May 2025 03:39:14 +0900 Subject: [PATCH 1/3] add self-host runner workflow --- .github/workflows/{test.yml => test-self-hosted.yml} | 7 +------ self-hosted-runner/Dockerfile | 2 +- self-hosted-runner/compose.yml | 2 ++ self-hosted-runner/env.template | 3 ++- 4 files changed, 6 insertions(+), 8 deletions(-) rename .github/workflows/{test.yml => test-self-hosted.yml} (55%) diff --git a/.github/workflows/test.yml b/.github/workflows/test-self-hosted.yml similarity index 55% rename from .github/workflows/test.yml rename to .github/workflows/test-self-hosted.yml index 51bebad..218c41f 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test-self-hosted.yml @@ -1,4 +1,4 @@ -name: "Docker and container hooks test" +name: "Test self-hosted runner" on: workflow_dispatch: @@ -10,10 +10,5 @@ jobs: container: image: docker:dind steps: - - name: Show docker info - run: | - docker -v - docker info - docker ps - run: ls - run: pwd \ No newline at end of file diff --git a/self-hosted-runner/Dockerfile b/self-hosted-runner/Dockerfile index 6a940fc..63bff4c 100644 --- a/self-hosted-runner/Dockerfile +++ b/self-hosted-runner/Dockerfile @@ -17,7 +17,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends\ software-properties-common \ && rm -rf /var/lib/apt/lists/* -# Install Docker cli +# Install Docker RUN install -m 0755 -d /etc/apt/keyrings \ && curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc \ && chmod a+r /etc/apt/keyrings/docker.asc \ diff --git a/self-hosted-runner/compose.yml b/self-hosted-runner/compose.yml index 7790c26..a1c4162 100644 --- a/self-hosted-runner/compose.yml +++ b/self-hosted-runner/compose.yml @@ -12,3 +12,5 @@ services: TARGET_ARCH: x64 volumes: - /var/run/docker.sock:/var/run/docker.sock + group_add: + - "${DOCKER_GID}" diff --git a/self-hosted-runner/env.template b/self-hosted-runner/env.template index bc955d4..a6bfb2a 100644 --- a/self-hosted-runner/env.template +++ b/self-hosted-runner/env.template @@ -1 +1,2 @@ -PERSONAL_ACCESS_TOKEN= \ No newline at end of file +PERSONAL_ACCESS_TOKEN= +DOCKER_GID= \ No newline at end of file From d73166e8b3a788a23ae0f0179ca391a78e1ad369 Mon Sep 17 00:00:00 2001 From: Junya Morioka Date: Sat, 17 May 2025 03:41:46 +0900 Subject: [PATCH 2/3] add self-host runner workflow --- .github/workflows/test-self-hosted.yml | 86 +++++++++++++++++++++++++- 1 file changed, 85 insertions(+), 1 deletion(-) diff --git a/.github/workflows/test-self-hosted.yml b/.github/workflows/test-self-hosted.yml index 218c41f..e37a4b8 100644 --- a/.github/workflows/test-self-hosted.yml +++ b/.github/workflows/test-self-hosted.yml @@ -11,4 +11,88 @@ jobs: image: docker:dind steps: - run: ls - - run: pwd \ No newline at end of file + - run: pwd + + test-build: + runs-on: + - self-hosted + timeout-minutes: 1000 + strategy: + fail-fast: false + matrix: + flash-attn-version: ["2.4.3"] + python-version: ["3.12"] + torch-version: ["2.7.0"] + # https://developer.nvidia.com/cuda-toolkit-archive + cuda-version: ["12.8.1"] + steps: + - uses: actions/checkout@v4 + - uses: actions/setup-python@v5 + with: + python-version: ${{ matrix.python-version }} + - uses: Jimver/cuda-toolkit@master + with: + cuda: ${{ matrix.cuda-version }} + linux-local-args: '["--toolkit"]' + method: "network" + + - name: Set CUDA and PyTorch versions + run: | + echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda-version }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV + echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch-version }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV + echo "CACHE_KEY=cuda-ext-${{ matrix.flash-attn-version }}-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-cuda${{ matrix.cuda-version }}" >> $GITHUB_ENV + + - name: Install build dependencies + run: | + sudo apt install -y ninja-build clang + pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil + + - name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }} + run: | + export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \ + support_cuda_versions = { \ + '2.0': [117, 118], \ + '2.1': [118, 121], \ + '2.2': [118, 121], \ + '2.3': [118, 121], \ + '2.4': [118, 121, 124], \ + '2.5': [118, 121, 124], \ + '2.6': [118, 124, 126], \ + '2.7': [118, 126, 128], \ + }; \ + target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \ + cuda_version = int(env['MATRIX_CUDA_VERSION']); \ + closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \ + print(closest_version) \ + ") + pip install --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION} + nvcc --version + python -V + python -c "import torch; print('PyTorch:', torch.__version__)" + python -c "import torch; print('CUDA:', torch.version.cuda)" + python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)" + + - name: Checkout flash-attn + run: | + git clone https://github.com/Dao-AILab/flash-attention.git -b "v${{ matrix.flash-attn-version }}" + cd flash-attention + + - name: Build wheels + timeout-minutes: 800 + run: | + export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH + export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH + export MAX_JOBS=2 + export NVCC_THREADS=2 + export FLASH_ATTENTION_FORCE_BUILD="TRUE" + cd flash-attention + python setup.py bdist_wheel --dist-dir=dist + base_wheel_name=$(basename $(ls dist/*.whl | head -n 1)) + wheel_name=$(echo $base_wheel_name | sed "s/${{ matrix.flash-attn-version }}/${{ matrix.flash-attn-version }}+cu${{ env.MATRIX_CUDA_VERSION }}torch${{ env.MATRIX_TORCH_VERSION }}/") + mv dist/$base_wheel_name dist/$wheel_name + echo "wheel_name=$wheel_name" >> $GITHUB_ENV + + - name: Install Test + run: | + pip install flash-attention/dist/${{ env.wheel_name }} + python -c "import flash_attn; print(flash_attn.__version__)" \ No newline at end of file From aa994ec6e569934c7aa0d0c42a1ff6ec8c72af4c Mon Sep 17 00:00:00 2001 From: Junya Morioka Date: Sat, 17 May 2025 03:43:50 +0900 Subject: [PATCH 3/3] change workflow container image --- .github/workflows/test-self-hosted.yml | 2 ++ 1 file changed, 2 insertions(+) diff --git a/.github/workflows/test-self-hosted.yml b/.github/workflows/test-self-hosted.yml index e37a4b8..c79f7e3 100644 --- a/.github/workflows/test-self-hosted.yml +++ b/.github/workflows/test-self-hosted.yml @@ -25,6 +25,8 @@ jobs: torch-version: ["2.7.0"] # https://developer.nvidia.com/cuda-toolkit-archive cuda-version: ["12.8.1"] + container: + image: ubuntu:22.04 steps: - uses: actions/checkout@v4 - uses: actions/setup-python@v5