Skip to content

[Fix] method always using cuda #4921

[Fix] method always using cuda

[Fix] method always using cuda #4921

Workflow file for this run

name: backend-ncnn
on:
push:
paths-ignore:
- "demo/**"
- "tools/**"
pull_request:
paths-ignore:
- "demo/**"
- "tools/**"
- "docs/**"
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
jobs:
test_onnx2ncnn:
runs-on: ubuntu-20.04
strategy:
matrix:
python-version: [3.7]
steps:
- name: Checkout repository
uses: actions/checkout@v3
with:
submodules: 'recursive'
- name: Set up Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
- name: Install dependencies
run: |
sudo apt update
sudo apt install wget gcc-multilib g++-multilib wget libprotobuf-dev protobuf-compiler
python -m pip install cmake onnx
- name: Install ncnn
run: |
wget https://github.com/Tencent/ncnn/archive/refs/tags/20220420.tar.gz
tar xf 20220420.tar.gz
pushd ncnn-20220420
mkdir build && pushd build
cmake -DCMAKE_INSTALL_PREFIX=$(pwd)/../install -DNCNN_BUILD_TESTS=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
cmake --build . -j2
make install
popd && popd
- name: Install mmdeploy with ncnn backend
run: |
mkdir -p build && pushd build
export LD_LIBRARY_PATH=/home/runner/work/mmdeploy/mmdeploy/ncnn-20220420/install/lib/:$LD_LIBRARY_PATH
cmake -DMMDEPLOY_TARGET_BACKENDS=ncnn -Dncnn_DIR=/home/runner/work/mmdeploy/mmdeploy/ncnn-20220420/install/lib/cmake/ncnn/ ..
make mmdeploy_onnx2ncnn -j2
popd
- name: Test onnx2ncnn
run: |
echo $(pwd)
ln -s build/bin/mmdeploy_onnx2ncnn ./
python .github/scripts/test_onnx2ncnn.py --run 1
build_ncnn:
runs-on: ubuntu-20.04
strategy:
matrix:
python-version: [3.7]
steps:
- name: Checkout repository
uses: actions/checkout@v3
with:
submodules: 'recursive'
- name: Set up Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
- name: Install mmdeploy
run: |
python -m pip install torch==1.8.2 torchvision==0.9.2 --extra-index-url https://download.pytorch.org/whl/lts/1.8/cpu
python -m pip install mmcv-lite
python tools/scripts/build_ubuntu_x64_ncnn.py 8
python -c 'import mmdeploy.apis.ncnn as ncnn_api; assert ncnn_api.is_available(with_custom_ops=True)'
test_ncnn_ptq:
runs-on: [self-hosted, linux-3090]
container:
image: openmmlab/mmdeploy:ubuntu20.04-cuda11.3
options: "--gpus=all --ipc=host"
steps:
- name: Checkout repository
uses: actions/checkout@v3
with:
submodules: recursive
- name: Install dependencies
run: |
apt-get update
apt-get install ninja-build -y
python3 -V
python3 -m pip install openmim
python3 -m pip install -r requirements.txt
python3 -m mim install $(cat requirements/codebases.txt | grep mmpretrain)
python3 -m pip install numpy==1.22.0
python3 -m pip list
- name: Install mmdeploy
run: |
rm -rf .eggs && python3 -m pip install -e .
python3 tools/check_env.py
- name: Install ppq
run: |
git clone -b v0.6.6 --depth 1 https://github.com/openppl-public/ppq
cd ppq
python3 -m pip install -r requirements.txt
python3 setup.py install
- name: Test ncnn + ppq pipeline
run: |
export PYTHONPATH=${PWD}/ppq:${PYTHONPATH}
export LD_LIBRARY_PATH="/root/workspace/mmdeploy/build/lib:${LD_LIBRARY_PATH}"
export LD_LIBRARY_PATH="/root/workspace/mmdeploy/mmdeploy/lib:${LD_LIBRARY_PATH}"
export work_dir=./work_dir
mkdir -p $work_dir
export model_cfg=$work_dir/resnet18_8xb32_in1k.py
export deploy_cfg=configs/mmpretrain/classification_ncnn-int8_static.py
export checkpoint=$work_dir/resnet18_8xb32_in1k_20210831-fbbb1da6.pth
export input_img=tests/data/tiger.jpeg
python3 -m mim download mmpretrain --config resnet18_8xb32_in1k --dest $work_dir
python3 tools/torch2onnx.py $deploy_cfg $model_cfg $checkpoint $input_img --work-dir $work_dir
wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.1.0/dataset.tar
tar xvf dataset.tar
python3 tools/onnx2ncnn_quant_table.py \
--onnx $work_dir/end2end.onnx \
--deploy-cfg $deploy_cfg \
--model-cfg $model_cfg \
--out-onnx $work_dir/quant.onnx \
--out-table $work_dir/ncnn.table \
--image-dir ./dataset
ls -sha $work_dir/quant.onnx
cat $work_dir/ncnn.table