Install

There are two ways of installing HOLLy that we have included in the repository. The first is to use a virtual environment such as Conda, PyEnv or Virtual Env. The second is to use a GPU enabled container with docker.

CUDA and GPU acceleration

HOLLy relies on having some form of GPU acceleration. Please refer to the Pytorch Getting Started docs. Typically, if you can run the program nvidia-smi successfully, you are off to a good start.

via a virtual environment install

Once you’ve picked your choice of virtual environment, you can run pip to install the required packages. In the example below, I’m using Miniconda:

conda create -n holly python
conda activate holly
pip install -r requirements.txt

via the included docker

There is a dockerfile included in docker/holly that will create a container that you can use to train our network.

Assuming you have docker installed, you can create the docker container like so:

docker build -t holly docker/holly

With this setup, you can run a test example : usage