Compilation and Installation
Procedure
- Use PuTTY to log in to the server as the root user.
- Go to the installation directory of the main program.
cd /path/to/paddle
- Decompress the installation package.
git clone https://github.com/PaddlePaddle/Paddle.git cd Paddle git checkout develop ulimit -n 4096 tar xvf v2.3.1.tar.gz cd Paddle-2.3.1
- Modify the configuration file.
sed -i "144c\ #-Werror " cmake/flags.cmake sed -i "227c\ set(GPU_COMMON_FLAGS -Wall -Wextra ${GPU_COMMON_FLAGS}) " cmake/flags.cmake - Perform the compilation.
mkdir build cd build git init cmake .. -DWITH_GPU=OFF -DWITH_TESTING=OFF -DCMAKE_BUILD_TYPE=Release -DPY_VERSION=3 -DWITH_MKL=OFF -DWITH_AVX=OFF -DON_INFER=ON -DWITH_XBAYAK=OFF -DWITH_ARM=ON git config --global http.postBuffer 524288000 make TARGET=ARMV8 -j$(nproc)
Note: In this step, you need to git clone the dependencies and then they are automatically installed. Ensure that your computer is connected to the Internet. This step may take a long time.
After the compilation is complete, a whl file is generated in python/dist of the current directory.

- Perform the installation.
cd python/dist pip3 install -U *.whl
Parent topic: paddle 2.3.1 Porting Guide (Kylin V10)