7
- '.github/workflows/nv-inference.yml'
9
- 'deepspeed/__init__.py'
10
- 'deepspeed/inference/**'
11
- '!deepspeed/inference/v2/**' # exclude v2 dir
12
- 'tests/unit/inference/**'
13
- '!tests/unit/inference/v2/**' # exclude v2 tests dir
20
group: ${{ github.workflow }}-${{ github.ref }}
21
cancel-in-progress: true
25
runs-on: [self-hosted, nvidia, cu116, v100]
28
- uses: actions/checkout@v3
31
uses: ./.github/workflows/setup-venv
33
- name: Install pytorch
35
pip install -U --cache-dir $TORCH_CACHE torch==2.1.2 torchvision==0.16.2 --index-url https://download.pytorch.org/whl/cu118
36
python -c "import torch; print('torch:', torch.__version__, torch)"
37
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"
39
- name: Install transformers
41
git clone https://github.com/huggingface/transformers
43
#git checkout f370bebdc
44
git rev-parse --short HEAD
47
- name: Install deepspeed
49
pip install .[dev,1bit,autotuning,inf,triton]
52
- name: Python environment
58
unset TORCH_CUDA_ARCH_LIST # only jit compile for current arch
60
#pytest $PYTEST_OPTS -m 'seq_inference' unit/ --torch_ver="2.1" --cuda_ver="11.8"
61
pytest $PYTEST_OPTS -m 'inference_ops' unit/ --torch_ver="2.1" --cuda_ver="11.8"
62
pytest $PYTEST_OPTS --forked -n 4 -m 'inference' unit/ --torch_ver="2.1" --cuda_ver="11.8"