Windows & Conda Setup

We use Conda to run APT using GPUs locally on a Windows machine.

1. Download and install Anaconda Python 3 version for Windows: https://www.anaconda.com/distribution/.
2. Download and install CUDA 10.0.
3. Start an Anaconda prompt (Start Menu -> Anaconda3 -> Anaconda Prompt).
4. Update Conda. Within the Anaconda terminal, type:
   conda update conda
conda upgrade --all
5. Create the APT environment from the provided environment file. We provide a preconfigured environment (yml) file in the APT repo, in the condaenv subfolder. Within the Anaconda terminal, type:
   conda env create -f c:\path\to\APT\install\apt_conda_environment.yml
   If this errors due to a previous/existing APT Conda environment, use the force flag:
   conda env create -f c:\path\to\APT\install\apt_conda_environment.yml --force

In APT, you should be able to choose "Local (Conda)" as the backend to use this APT conda environment.

Test that this worked by running Track -> GPU/Backend Configuration -> Test backend configuration.

Notes/Troubleshooting

• We discourage starting MATLAB from within a command shell or Anaconda prompt that has activated the APT Conda environment. This has sometimes caused unusual problems with libraries and dependencies. Instead, start MATLAB the usual way (and outside of any Conda environment), eg by double-clicking its program icon.