Compile on Windows from Source Code¶

Direct native source code compilation¶

Please strictly follow the following instructions step by step

1. Install the necessary tools i.e. cmake, git and python:

   CMake requires version 3.17 and above, and add to the ring Environment variables.

   Python requires version 3.8 and above, which can be downloaded from the official website.

   • After installing python, please check whether the python version is the expected version by python-version, because you may have more than one python installed on your computer. You can handle conflicts of multiple pythons by changing the order of the environment variables.

   numpy, protobuf, wheel are needed to be installed. Use the ‘pip’ command.

   • To Install numpy package you can use command

     pip install numpy
     

   • To Install protobuf package you can use command

     pip install protobuf
     

   • To Install Wheel package you can use command

     pip install wheel
     

   Git can be downloaded on the official website and added to the environment variable.

2. Clone the PaddlePaddle source code to the Paddle subdirectories of the current directory and go to the Paddle subdirectories:

   git clone https://github.com/PaddlePaddle/Paddle.git
   
   cd Paddle
   

3. Switch to develop branch for compilation:

   git checkout develop
   

   Note: Paddle supports Python version 3.8 and above.

4. Create a directory called build and enter it:

   mkdir build
   
   cd build
   

5. Execute cmake:

   For details on the compilation options, see the compilation options list. On Windows, you can compile by Ninja(recommended) or Visual Studio IDE, as follow:

   • 1）Compile by Ninja(recommended) method:

     Firstly, install ninja:

     pip install ninja
     

     Then, search “x64 Native Tools Command Prompt for VS” in Windows search bar, run it as Administrator. Here is the cmake command:

     cmake .. -GNinja -DWITH_GPU=OFF -DWITH_UNITY_BUILD=ON
     

   • 2）Compile by Visual Studio IDE method:

     cmake .. -G "Visual Studio 15 2017" -A x64 -T host=x64 -DWITH_GPU=OFF -DWITH_UNITY_BUILD=ON
     

     In the above command, change to -DWITH_GPU=ON to compile the GPU version Paddle.

     Note:

     1. If more than one CUDA are installed, the latest installed CUDA will be used. If you need to specify a CUDA version, you’ll need to set environment variables and CMake options. for example:

     set CUDA_TOOLKIT_ROOT_DIR=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.2
     set PATH=%CUDA_TOOLKIT_ROOT_DIR:/=\%\bin;%CUDA_TOOLKIT_ROOT_DIR:/=\%\libnvvp;%PATH%
     cmake .. -GNinja -DWITH_GPU=ON -DCUDA_TOOLKIT_ROOT_DIR="%CUDA_TOOLKIT_ROOT_DIR%" -DWITH_UNITY_BUILD=ON
     

     1. If more than one Python are installed, the latest installed Python will be used by default, and you can choose the Python version by -DPYTHON_EXECUTABLE . for example:

     cmake .. -GNinja -DWITH_GPU=ON -DPYTHON_EXECUTABLE=C:\\Python38\\python.exe  -DWITH_UNITY_BUILD=ON
     

6. Execute compile:

   • 1. For Ninja method(recommended), run ninja all , it will begin to compile.

   • 1. For Visual Studio IDE method, use Visual Studio to open paddle.sln file, select the platform x64, configure with Release, click the button, it will begin to compile.

7. After compilation successfully, go to the \paddle\build\python\dist directory and find the generated .whl package:

   cd \paddle\build\python\dist
   

8. Install the generated .whl package:

   pip install -U (whl package name)
   

Congratulations, you have completed the process of compiling PaddlePaddle successfully!

Verify installation¶

After the compilation and installation is completed, you can use python to enter the Python interface, input

import paddle

and then

paddle.utils.run_check()

to verify that the installation was successful.

If PaddlePaddle is installed successfully! appears, it means the compilation and installation was successful.

How to uninstall¶

Please use the following command to uninstall PaddlePaddle:

• CPU version of PaddlePaddle :

  pip uninstall paddlepaddle
  

• GPU version of PaddlePaddle :

  pip uninstall paddlepaddle-gpu