9
group: ${{ github.workflow }}-${{ github.ref }}
10
cancel-in-progress: true
18
# The type of runner that the job will run on
19
runs-on: [self-hosted, amd, mi200]
21
# Steps represent a sequence of tasks that will be executed as part of the job
23
# Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
24
- uses: actions/checkout@v3
27
uses: ./.github/workflows/setup-venv
29
- name: Install pytorch
31
pip install -U --cache-dir $TORCH_CACHE torch torchvision --index-url https://download.pytorch.org/whl/rocm5.6
32
python -c "import torch; print('torch:', torch.__version__, torch)"
33
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"
35
- name: Install transformers
37
git clone https://github.com/huggingface/transformers
39
# if needed switch to the last known good SHA until transformers@master is fixed
40
# git checkout 1cc453d33
41
git rev-parse --short HEAD
44
- name: Install (ROCm) apex
46
git clone https://github.com/ROCmSoftwarePlatform/apex.git
48
git checkout torch_2.1_higher
49
CURRENT_VER=$(git rev-parse HEAD)
50
INSTALLED_VER=$(cat /blob/amd-apex/.venv_installed_version)
51
if [[ "$CURRENT_VER" != "$INSTALLED_VER" ]]; then
52
pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings="--global-option=--cpp_ext" --config-settings="--global-option=--cuda_ext" --target=/blob/amd-apex/ --upgrade .
53
git rev-parse HEAD > /blob/amd-apex/.venv_installed_version
55
echo PYTHONPATH=$PYTHONPATH:/blob/amd-apex/ >> $GITHUB_ENV
56
# Runs a set of commands using the runners shell
57
- name: Install deepspeed
59
pip install .[dev,1bit,autotuning]
60
#python -c "from deepspeed.env_report import cli_main; cli_main()"
63
- name: Python environment
67
# Runs a set of commands using the runners shell
70
unset TORCH_CUDA_ARCH_LIST # only jit compile for current arch
72
pytest $PYTEST_OPTS -n 4 --verbose unit/
73
pytest $PYTEST_OPTS -m 'sequential' unit/
75
- name: Open GitHub issue if nightly CI fails
76
if: ${{ failure() && (github.event_name == 'schedule') }}
77
uses: JasonEtco/create-an-issue@v2
79
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
81
filename: .github/ISSUE_TEMPLATE/ci_failure_report.md