PyTorch is an open source machine learning library. It is widely used for applications such as computer vision and natural language processing.

Procedure for Installing PyTorch and PyTorch Vision on Raspberry Pi 4B

• 1. First, update the Raspbian OS (on the terminal):

     1. • sudo apt-get update

        • sudo apt-get dist-upgrade

  2. Install dependencies (on the terminal):

     1. • sudo apt install libopenblas-dev libblas-dev m4 cmake cython python3-dev python3-yaml python3-setuptools

        • sudo apt-get install libavutil-dev libavcodec-dev libavformat-dev libswscale-dev

  3. You can install PyTorch and PyTorch Vision using the pre-compiled Python wheel files:

     1. • First, download wheel files from https://github.com/sungjuGit/Pytorch-and-Vision-for-Raspberry-Pi-4B

        • sudo pip3 install replace_with_name_for_pytorch_wheel.whl

        • sudo pip3 install replace_with_name_for_pytorchvision_wheel.whl

  4. Or, if you would like to get the latest version by building PyTorch from the source yourself, which may take 8~10 hours (otherwise skip this step):

     1. • mkdir pytorch_install

        • cd pytorch_install

        • git clone --recursive https://github.com/pytorch/pytorch

        • cd pytorch

        • git submodule update --remote third_party/protobuf

        • export NO_CUDA=1

        • export NO_DISTRIBUTED=1

        • export NO_MKLDNN=1

        • export NO_NNPACK=1

        • export NO_QNNPACK=1

        • python3 setup.py build

        • sudo -E python3 setup.py install

        • cd ..

        • mkdir pytorch_vision_install && cd pytorch_vision_install

        • git clone --recursive https://github.com/pytorch/vision

        • cd vision

        • sudo -E python3 setup.py install

  5. Check the installation by running a simple test and confirming that Python outputs a tensor (5x3 matrix) with random number elements:

     1. • cd

        • python3

        • (in the python shell)>>> from __future__ import print_function

        • (in the python shell)>>> import torch

        • (in the python shell)>>> a = torch.rand(5,3)

        • (in the python shell)>>> print (a)