Anaconda on the AIR-T


Anaconda is a great virtual environment for developing Python applications and it is fully supported by the AIR-T and AirStack. This tutorial will show you how to create an AirStack Conda environment on the AIR-T. Note that GNU Radio is not yet available in this environment, but the SoapySDR radio drivers and GPU acceleration for signal processing and deep learning supported.

For AirStack 0.4.0+, Anaconda comes pre-installed with a built-in local conda channels for the AIR-T drivers. For older AirStack versions, see the Installing Anaconda on AIR-T Prior to AirStack 0.4.0 at the end of this article.

Create AirStack Conda Environment

We will now create a conda environment called airstack using the airstack.yml file provided by Deepwave. You may modify this to fit your application.

wget docs.deepwavedigital.com/Open_Source/files/airstack.yml
conda env create -f airstack.yml

Testing a New Conda Environment

Copy the conda_test.py script to your AIR-T to test full operation.

conda activate airstack
wget docs.deepwavedigital.com/Open_Source/files/conda_test.py
python conda_test.py

Removing the Conda Environment

To remove the AirStack Conda environment, run the following command:

conda remove --name airstack --all


Installing Anaconda on AIR-T Prior to AirStack 0.4.0

Anaconda is pre-installed on AirStack v0.4.0+. If you are using an earlier version of AirStack, we recommend the conda4aarch64 with instructions provided here. The instructions are reproduced below with some slight modifications.

1. Install Anaconda:

wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge-pypy3-Linux-aarch64.sh
chmod +x Miniforge-pypy3-Linux-aarch64.sh
./Miniforge-pypy3-Linux-aarch64.sh

When prompted with the following question answer yes

Do you wish the installer to initialize Miniforge3? [yes|no]

2. (Optional) At this point, the conda base environment will be activated whenever a terminal is opened. We recommend removing this behavior by executing the following:

$ conda config --set auto_activate_base false

Now the conda environment is only activated when you type conda activate <env_name>.

3. Create a airstack.yml file with the content below.

name: airstack
channels:
  - conda-forge
  - nvidia
  - defaults
  - numba
  - c4aarch64
dependencies:
  - python=3.6
  - scipy
  - numpy
  - matplotlib
  - ipython
  - pip

4. Open a terminal and create the new conda environment from the airstack.yml file.

conda env create -f airstack.yml

5. Install AirStack Drivers in Conda Environment by adding symbolic links to the AIR-T radio drivers from within a new conda environment. The radio will not work inside the environment unless these steps are followed, so you must do this for each new conda environment that you create. As a reminder, this is not necessary for AirStack 0.4.0+

Note: you must have your Conda environment activated before proceeding.

conda activate airstack
wget docs.deepwavedigital.com/Open_Source/files/conda_post_install.sh
bash conda_post_install.sh

The conda_post_install.sh shell script will create the necessary symbolic links to bring the radio drivers into the AirStack environment.


Last update: August 25, 2020