Installation on Windows

Environment Preparation

  • Windows 7/8/10 Pro/Enterprise(64bit)
    • GPU Version support CUDA 9.0/9.1/9.2/10.0/10.1, and only support single GPU
  • Python version 2.7.15+/3.5.1+/3.6/3.7(64bit)
  • pip version 9.0.1+(64bit)

Precautions

  • Confirm the Python you install PaddlePaddle is the version you expected, because your computer may have more than one python, use the following command:

    python --version

    • If you are using Python 2, the output should be 2.7.15+
    • If you are using Python 3, the output should be 3.5.1+/3.6+/3.7+
  • If Python doesn't match your expected version, use the following command to see if Python's path is where you expect it to be:

    where python

    • If you are using Python 2, The installation directory for python2.7 should be on the first line
    • If you are using Python 3, The installation directory for python3.5.1+/3.6+/3.7+ should be on the first line
    • You can adjust it in any of the following ways:
      • Use specific Python paths to execute commands(e.g. C:\Python36\python.exe corresponding to Python 3,C:\Python27\python.exe corresponding to Python 2)
      • By modifying the environment variable, set your expected installation path in the first order (please modify it in control panel -> system properties -> environment variable -> path)
  • Confirm whether the pip version meets the requirements. The pip version is required to be 9.0.1+

    python -m ensurepip

    python -m pip --version

  • Confirm that Python and pip is 64 bit,and the processor architecture is x86_64(or x64、Intel 64、AMD64)architecture. Currently, PaddlePaddle doesn't support arm64 architecture. The first line of output from the following command should be "64bit", and the second line should be "x86_64", "x64" or "AMD64":

    python -c "import platform;print(platform.architecture()[0]);print(platform.machine())"

  • The installation package provided by default requires the computer to support MKL. If your environment does not support MKL, please download the openblas version of the installation package in here

  • Nccl, distributed and other related functions are not supported in the current version.

CPU or GPU

  • If your computer doesn’t have NVIDIA® GPU, please install the CPU version of PaddlePaddle
  • If your computer has NVIDIA® GPU, and it satisfies the following requirements, we recommend you to install the GPU version of PaddlePaddle
    • CUDA Toolkit 9.0/10.0 with cuDNN v7.3+
    • GPU's computing capability exceeds 1.0

Note: currently, the official Windows installation package only support CUDA 9.0/10.0 with single GPU, and don't include CUDA 9.1/9.2/10.1. if you need to use, please compile by yourself through the source code.

Please refer to the NVIDIA official documents for the installation process and the configuration methods of CUDA and cuDNN.

Installation Method

There are 3 ways to install PaddlePaddle on Windows:

We would like to introduce the pip installation here.

Installation steps

  • CPU version of PaddlePaddle:
    • python -m pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple(Baidu source is recommended) or python -m pip install paddlepaddle -i https://pypi.tuna.tsinghua.edu.cn/simple
  • GPU version of PaddlePaddle:
    • python -m pip install paddlepaddle-gpu -i https://mirror.baidu.com/pypi/simple(Baidu source is recommended) or python -m pip install paddlepaddle-gpu -i https://pypi.tuna.tsinghua.edu.cn/simple

There is a checking function below for verifyig whether the installation is successful. If you have any further questions, please check the FAQ.

Notice:

  • python -m pip install paddlepaddle-gpu -i https://pypi.tuna.tsinghua.edu.cn/simple This command will install PaddlePaddle that supports CUDA 10.0(with cuDNN v7.3+). If you have different requirements for the version of CUDA or cuDNN, you can use command python -m pip install paddlepaddle-gpu==[version number] -i https://pypi.tuna.tsinghua.edu.cn/simple to install, and you can see the version here. For the corresponding relations between PaddlePaddle and CUDA, cuDNN version, please see the installer package list

Installation Verification

After completing the installation process, you can use python to enter python interface and input import paddle.fluid as fluid and then fluid.install_check.run_check() to check whether the installation is successful.

If you see Your Paddle Fluid is installed succesfully!, your installation is verified successful.

Uninstall PaddlePaddle

  • CPU version of PaddlePaddle: python -m pip uninstall paddlepaddle
  • GPU version of PaddlePaddle: python -m pip uninstall paddlepaddle-gpu