Docker installation¶

Step 1 Install Docker on your machine by following the docker installation instructions.

Note - You can install Community Edition (CE)

Step 2 [Optional] Post installation steps to manage Docker as a non-root user.

Follow the four steps in this docker documentation to allow managing docker containers without sudo.

Build¶

Please use a Native build with gcc 4 as explained below, higher compiler versions currently cause test failures on ARM

The following command will build a container with dependencies and tools and then compile MXNet for ARMv7. The resulting artifact will be located in build/mxnet-x.x.x-py2.py3-none-any.whl, copy this file to your Raspberry Pi.

ci/build.py -p armv7

Install¶

Create a virtualenv and install the package we created previously.

virtualenv -p `which python3` mxnet_py3
source mxnet_py3/bin/activate
pip install mxnet-x.x.x-py2.py3-none-any.whl

Native Build

Installing MXNet is a two-step process:

1. Build the shared library from the MXNet C++ source code.
2. Install the supported language-specific packages for MXNet.

Step 1 Build the Shared Library

On Raspbian versions Wheezy and later, you need the following dependencies:

• Git (to pull code from GitHub)
• libblas (for linear algebraic operations)
• libopencv (for computer vision operations. This is optional if you want to save RAM and Disk Space)
• A C++ compiler that supports C++ 11. The C++ compiler compiles and builds MXNet source code. Supported compilers include the following:
  • G++ (4.8 or later). Make sure to use gcc 4 and not 5 or 6 as there are known bugs with these compilers.
  • Clang (3.9 - 6)

Install these dependencies using the following commands in any directory:

    sudo apt-get update
    sudo apt-get -y install git cmake ninja-build build-essential g++-4.9 c++-4.9 liblapack* libblas* libopencv* libopenblas* python3-dev virtualenv

Clone the MXNet source code repository using the following git command in your home directory:

    git clone https://github.com/apache/incubator-mxnet.git --recursive
    cd incubator-mxnet

Build:

    mkdir -p build && cd build
    cmake \
        -DUSE_SSE=OFF \
        -DUSE_CUDA=OFF \
        -DUSE_OPENCV=ON \
        -DUSE_OPENMP=ON \
        -DUSE_MKL_IF_AVAILABLE=OFF \
        -DUSE_SIGNAL_HANDLER=ON \
        -DCMAKE_BUILD_TYPE=Release \
        -GNinja ..
    ninja -j$(nproc)

Some compilation units require memory close to 1GB, so it’s recommended that you enable swap as explained below and be cautious about increasing the number of jobs when building (-j)

Executing these commands start the build process, which can take up to a couple hours, and creates a file called libmxnet.so in the build directory.

If you are getting build errors in which the compiler is being killed, it is likely that the compiler is running out of memory (especially if you are on Raspberry Pi 1, 2 or Zero, which have less than 1GB of RAM), this can often be rectified by increasing the swapfile size on the Pi by editing the file /etc/dphys-swapfile and changing the line CONF_SWAPSIZE=100 to CONF_SWAPSIZE=1024, then running:

  sudo /etc/init.d/dphys-swapfile stop
  sudo /etc/init.d/dphys-swapfile start
  free -m # to verify the swapfile size has been increased

Step 2 Install MXNet Python Bindings

To install Python bindings run the following commands in the MXNet directory:

    cd python
    pip install --upgrade pip
    pip install -e .

Note that the -e flag is optional. It is equivalent to --editable and means that if you edit the source files, these changes will be reflected in the package installed.

Alternatively you can create a whl package installable with pip with the following command:

ci/docker/runtime_functions.sh build_wheel python/ $(realpath build)

You are now ready to run MXNet on your Raspberry Pi device. You can get started by following the tutorial on Real-time Object Detection with MXNet On The Raspberry Pi.

Note - Because the complete MXNet library takes up a significant amount of the Raspberry Pi’s limited RAM, when loading training data or large models into memory, you might have to turn off the GUI and terminate running processes to free RAM.