Skip to content

Latest commit

 

History

History
124 lines (83 loc) · 3.52 KB

2.1-LINUX_TUTORIAL.md

File metadata and controls

124 lines (83 loc) · 3.52 KB

Linux Compilation Tutorial

The tutorial requires some basic knowledge about Linux OS!

Though it is unclear why it needs to run in Linux... Anyway in case somebody needs it_(:з」∠)_

If you meet any other problems, please feel free to submit issue to us.

Note Linux build of MAA is still under discussion, some of the content might be outdated, please follow the script in GitHub workflow file

Compiling MAA itself only

Download build depends

  • Ubuntu/Debian

    sudo apt install gcc-12 g++-12 cmake zlib1g-dev

  • Arch Linux

    sudo pacman -S --needed gcc cmake zlib

Build MAA

git clone https://github.com/MaaAssistantArknights/MaaAssistantArknights.git
cd MaaAssistantArknights
mkdir -p build
CC=gcc-12 CXX=g++-12 cmake -B build \
    -DINSTALL_THIRD_LIBS=ON \
    -DINSTALL_RESOURCE=ON \
    -DINSTALL_PYTHON=ON
cmake --build build

cmake would download opencv and fastdeploy automatically before compiling by default. Run the following command to install MAA to the target directory after build

cmake --install build --prefix <target_directory>

MAA should be run by setting LD_LIBRARY_PATH, do not attempt to install MAA into / with superuser privilege.

Compiling/specifying third-party libraries

Prebuilt opencv and fastdeploy downloaded by cmake may not compatible with all system, if you ran into abi version issue, the following instructions may help.

OpenCV

Install with package manager of your system, e.g.:

  • Ubuntu/Debian

    sudo apt install libopencv libopencv-dev
# libopencv-dev 可以在编译完成后移除

  • Arch Linux

    sudo pacman -S opencv

Note that MAA on Windows were developed under OpenCV 4.5.3, if you would like to use specific version or statically link to OpenCV, try to compile from source .

FastDeploy

  1. Compile and install C++ SDK according to the tutorial, turn on ONNX Runtime (Ort) and Vision switch. Both CPU and GPU version are ok.

    git clone https://github.com/PaddlePaddle/FastDeploy.git
cd FastDeploy
mkdir build && cd build
cmake .. -DENABLE_ORT_BACKEND=ON \
         -DENABLE_PADDLE_BACKEND=ON \
         -DENABLE_OPENVINO_BACKEND=OFF \
         -DCMAKE_INSTALL_PREFIX=${PWD}/compiled_fastdeploy_sdk \
         -DENABLE_VISION=ON \
         -DENABLE_TEXT=OFF \
         -DOPENCV_DIRECTORY=/
make -j 12
make install

    in which -DCMAKE_INSTALL_PREFIX is the directory to install into, -DOPENCV_DIRECTORY should be set to install directory of OpenCV.

MAA

Similar to above with additional cmake argument

cmake -B build \
    -DINSTALL_THIRD_LIBS=OFF \
    -DOPENCV_DIRECTORY=path_to_opencv_install_prefix \
    -DFASTDEPLOY_DIRECTORY=path_to_compiled_fastdeploy_sdk \
    ...
cmake --build build

Installing and packaging

Dependencies might depends on compilation argument of opencv and fastdeploy, check manually after cmake --install, copy files if needed.

Integration Documentation

Maybe not a doc

Python

Refer to the implementation of __main__ in Python demo

C

Refer to the implementation of CppSample

C sharp

Refer to the implementation of MaaWpfGui