Suggestion from Tom:
I am wondering as well whether we could make a super minimal mock dataset to use to run the model in GitHub Actions, we should be able to build the container, then start a new job where the action runner is using the docker image we just created