Skip to content

huawei-ascend-npu

huawei-ascend-npu #214

name: huawei-ascend-npu
defaults:
run:
shell: bash -ieo pipefail {0}
on:
workflow_dispatch:
pull_request:
paths:
- '.github/workflows/huawei-ascend-npu.yml'
- 'requirements/**'
schedule:
- cron: "0 0 * * *"
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
permissions:
contents: read
issues: write
jobs:
unit-tests:
runs-on: [self-hosted, ascend, npu]
container:
image: ascendai/cann
ports:
- 80
volumes:
- /usr/local/dcmi:/usr/local/dcmi
- /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
- /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/
- /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info
- /etc/ascend_install.info:/etc/ascend_install.info
options: --network host
--name deepspeed_unit-tests
--device /dev/davinci4
--device /dev/davinci_manager
--device /dev/devmm_svm
--device /dev/hisi_hdc
--shm-size "20g"
--entrypoint /bin/bash
env:
PT_HPU_LAZY_MODE: 0
TORCHINDUCTOR_COMPILE_THREADS: 1
TEST_LIST: |
test_accelerator.py
test_autotuning.py
test_compression.py
steps:
- uses: actions/checkout@v4
- name: Install pytorch
run: |
npu-smi info
apt-get update
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
source /root/.bashrc
pip install torch==2.2.0 torchvision==0.17.0 torch_npu==2.2.0 numpy==1.26.4 cloudpickle tornado ml-dtypes
python << EOF
if __name__ == '__main__':
import torch
import torch_npu
torch_npu.npu.set_device("npu:0")
print(f"Device Name: {torch.npu.get_device_name(0)}")
print(f"Device Count: {torch.npu.device_count()}")
print(f"Device Available: {torch.npu.is_available()}")
EOF
- name: Install transformers
run: |
source /root/.bashrc
echo "y" | apt-get install git
git clone https://github.com/huggingface/transformers
cd transformers
git rev-parse --short HEAD
pip install .
- name: Install deepspeed
run: |
source /root/.bashrc
git clone --depth=1 https://github.com/microsoft/DeepSpeed.git
pip install .[1bit,npu,autotuning,inf]
ds_report
- name: Python environment
run: |
source /root/.bashrc
pip list
- name: Unit tests
run: |
source /root/.bashrc
unset TORCH_CUDA_ARCH_LIST # only jit compile for current arch
cd tests/unit
pytest --verbose accelerator/*
pytest --verbose autotuning/*
pytest --verbose checkpoint/test_reshape_checkpoint.py
pytest --verbose checkpoint/test_moe_checkpoint.py
pytest --verbose checkpoint/test_shared_weights.py
pytest --verbose launcher/test_ds_arguments.py launcher/test_run.py
pytest --verbose model_parallelism/*
pytest --verbose moe/test_moe_tp.py
pytest --verbose monitor/*
pytest --verbose utils/*
pytest --verbose runtime/test_ds_config_model.py
pytest --verbose runtime/pipe/test_pipe_schedule.py
pytest --verbose runtime/zero/test_zero_config.py
pytest --verbose runtime/zero/test_zero_tiled.py
pytest --verbose runtime/zero/test_zeropp.py
pytest --verbose runtime/test_autocast.py
pytest --verbose runtime/test_data.py
pytest --verbose runtime/test_runtime_utils.py
pytest --verbose runtime/activation_checkpointing/*
pytest --verbose runtime/utils/*
pytest --verbose runtime/zero/test_zero_dynamic_class.py