Release #224
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: Release | |
defaults: | |
run: | |
shell: bash -le {0} | |
on: | |
schedule: | |
- cron: '0 20 * * *' | |
release: | |
types: [ published ] | |
repository_dispatch: | |
workflow_dispatch: | |
inputs: | |
repo: | |
description: 'GitHub repo {owner}/{repo}' | |
required: false | |
default: '' | |
ref: | |
description: 'GitHub ref: Branch, Tag or Commit SHA' | |
required: false | |
default: '' | |
target: | |
description: 'CUDA Torch Python version separated by space, check http://10.0.14.199/gpu/runner/docker to get all supported combinations' | |
required: false | |
default: '' | |
max-parallel: | |
description: 'max parallel jobs' | |
required: false | |
default: '{"size": 6}' | |
upload_release: | |
description: 'upload to release (it only works with a tag ref)' | |
type: boolean | |
required: false | |
default: 'false' | |
upload_pypi: | |
description: 'upload to PyPI' | |
type: boolean | |
required: false | |
default: 'false' | |
env: | |
CUDA_DEVICE_ORDER: PCI_BUS_ID | |
ZEN4_SERVER: 10.0.14.199 | |
TORCH_CUDA_ARCH_LIST: '6.0 6.1 6.2 7.0 7.5 8.0 8.6 8.9 9.0' | |
CUDA_RELEASE: 1 | |
CI: 1 | |
GPTQMODEL_FORCE_BUILD: 1 | |
repo: ${{ github.event.inputs.repo || github.repository }} | |
ref: ${{ github.event.inputs.ref || github.ref }} | |
MAX_JOBS: 4 | |
concurrency: | |
group: ${{ github.event.inputs.ref || github.ref }}-workflow-release | |
cancel-in-progress: true | |
jobs: | |
check-vm: | |
runs-on: self-hosted | |
container: | |
image: modelcloud/gptqmodel:alpine-ci-v1 | |
outputs: | |
ip: ${{ steps.get_ip.outputs.ip }} | |
zen4_list: ${{ steps.assign.outputs.zen4_list }} | |
steps: | |
- name: Print env | |
run: | | |
echo "event name: ${{ github.event_name }}" | |
echo "repo: ${{ env.repo }}" | |
echo "ref: ${{ env.ref }}" | |
echo "max-parallel: ${{ inputs.max-parallel }}" | |
echo "upload_release: ${{ inputs.upload_release }}" | |
echo "upload_pypi: ${{ inputs.upload_pypi }}" | |
- name: Select server | |
id: get_ip | |
run: | | |
echo "ip=${ZEN4_SERVER}" >> "$GITHUB_OUTPUT" | |
echo "GPU_IP=${ZEN4_SERVER}" >> $GITHUB_ENV | |
echo "ip: $ip" | |
- name: Assign tasks | |
id: assign | |
run: | | |
input="${{ github.event.inputs.target }}" | |
if [[ -z "$input" ]]; then | |
server_lists=$(curl -s "http://${{ env.GPU_IP }}/gpu/runner/docker?json=1") | |
echo "lists=$server_lists" | |
IFS=$'+' read -r list_1 list_2 <<< "$server_lists" | |
echo "list 1: $list_1" | |
echo "list 2: $list_2" | |
else | |
echo "inputed target is: $input" | |
if [[ "$input" =~ \ ]]; then | |
echo "extracting cuda torch and python" | |
read -r cuda torch py <<< "$input" | |
task=compiler_cuda$cuda-torch$torch-python$py | |
list_1="[\"$task\"]" | |
else | |
echo "use inputed image name" | |
list_1="[\"$input\"]" | |
fi | |
list_2="[]" | |
echo "task list: $list_1" | |
fi | |
echo "zen4_list=$list_1" >> "$GITHUB_OUTPUT" | |
release-zen4: | |
strategy: | |
fail-fast: false | |
matrix: | |
tag: ${{ fromJSON(needs.check-vm.outputs.zen4_list) }} | |
max-parallel: ${{ fromJson(inputs.max-parallel).size || 6 }} | |
runs-on: [ self-hosted, zen4 ] | |
needs: | |
- check-vm | |
- release-source | |
if: needs.check-vm.outputs.zen4_list != '' && !cancelled() | |
container: | |
image: ${{ needs.check-vm.outputs.ip }}:5000/modelcloud/gptqmodel:${{ matrix.tag }} | |
timeout-minutes: 70 | |
steps: | |
- name: Checkout Codes | |
uses: actions/checkout@v4 | |
with: | |
repository: ${{ env.repo }} | |
ref: ${{ env.ref }} | |
- name: Print Env | |
run: | | |
echo "== pyenv ==" | |
pyenv versions | |
echo "== python ==" | |
python --version | |
echo "== nvcc ==" | |
#nvcc --version | |
echo "== torch ==" | |
pip show torch | |
- name: Install requirements | |
run: | | |
cuda_version=$(echo ${{ matrix.tag }} | grep -oP 'cuda\K[0-9.]+') | |
torch_version=$(echo ${{ matrix.tag }} | grep -oP 'torch\K[0-9.]+') | |
python_version=$(echo ${{ matrix.tag }} | grep -oP 'python\K[0-9.]+') | |
bash -c "$(curl -L http://${ZEN4_SERVER}/scripts/compiler/init_env.sh)" @ $cuda_version $torch_version $python_version | |
- name: Compile | |
run: python setup.py bdist_wheel | |
- name: Test install | |
run: | | |
ls -ahl dist | |
whl=$(ls -t dist/*.whl | head -n 1 | xargs basename) | |
echo "WHL_NAME=$whl" >> $GITHUB_ENV | |
twine check dist/$whl | |
uv pip install dist/$whl | |
- name: Upload wheel | |
continue-on-error: true | |
run: | | |
sha256=$(sha256sum dist/${{ env.WHL_NAME }}) | |
response=$(curl -s -F "runid=${{ github.run_id }}" -F "repo=${{ env.repo }}" -F "ref=${{ env.ref }}" -F "sha256=$sha256" -F "file=@dist/${{ env.WHL_NAME }}" http://${{ needs.check-vm.outputs.ip }}/gpu/whl/upload) | |
if [ "$response" -eq 0 ]; then | |
echo "UPLOADED=1" >> $GITHUB_ENV | |
fi | |
- name: Upload artifact | |
uses: actions/upload-artifact@v4 | |
continue-on-error: ${{ env.UPLOADED == '1' }} | |
with: | |
name: ${{ env.WHL_NAME }} | |
path: dist/${{ env.WHL_NAME }} | |
- name: Upload binaries to release | |
uses: svenstaro/upload-release-action@v2 | |
if: github.event_name == 'release' && !cancelled() | |
with: | |
repo_name: ${{ env.repo }} | |
tag: ${{ env.ref }} | |
file: dist/${{ env.WHL_NAME }} | |
file_glob: true | |
overwrite: true | |
release-source: | |
strategy: | |
fail-fast: false | |
runs-on: self-hosted | |
needs: check-vm | |
container: | |
image: ${{ needs.check-vm.outputs.ip }}:5000/modelcloud/gptqmodel:compiler_cuda124-torch2.4.1-python311 | |
env: | |
CUDA_RELEASE: 0 | |
steps: | |
- name: Checkout Codes | |
uses: actions/checkout@v4 | |
with: | |
repository: ${{ env.repo }} | |
ref: ${{ env.ref }} | |
- name: Install requirements | |
run: uv pip install build setuptools -U -i http://${{ needs.check-vm.outputs.ip }}/simple/ --trusted-host ${{ needs.check-vm.outputs.ip }} | |
- name: Compile | |
run: python -m build --no-isolation --sdist | |
- name: Check dist | |
run: | | |
ls -ahl dist | |
whl=$(ls -t dist/*.gz | head -n 1 | xargs basename) | |
echo "WHL_NAME=$whl" >> $GITHUB_ENV | |
twine check dist/$whl | |
- name: Upload to local | |
continue-on-error: true | |
run: | | |
sha256=$(sha256sum dist/${{ env.WHL_NAME }}) | |
response=$(curl -s -F "runid=${{ github.run_id }}" -F "repo=${{ env.repo }}" -F "ref=${{ env.ref }}" -F "sha256=$sha256" -F "file=@dist/${{ env.WHL_NAME }}" http://${{ needs.check-vm.outputs.ip }}/gpu/whl/upload) | |
if [ "$response" -eq 0 ]; then | |
echo "UPLOADED=1" >> $GITHUB_ENV | |
fi | |
- name: Upload to artifact | |
uses: actions/upload-artifact@v4 | |
continue-on-error: ${{ env.UPLOADED == '1' }} | |
with: | |
name: ${{ env.WHL_NAME }} | |
path: dist/${{ env.WHL_NAME }} | |
- name: Upload package to release | |
uses: svenstaro/upload-release-action@v2 | |
if: (github.event_name == 'release' || github.event.inputs.upload_release == 'true') && !cancelled() | |
with: | |
file: dist/${{ env.WHL_NAME }} | |
tag: ${{ env.ref }} | |
file_glob: true | |
overwrite: true | |
- name: Upload sdist to pypi | |
if: (github.event_name == 'release' || github.event.inputs.upload_pypi == 'true') && !cancelled() | |
env: | |
TWINE_USERNAME: "__token__" | |
TWINE_PASSWORD: ${{ secrets.PYPI_TOKEN }} | |
run: | | |
python -m twine upload dist/*gz |