To launch locally, clone this repo then simply run docker-compose up -d in the project root folder. Then go to localhost:8888 in a web browser and open runner.ipynb. You can also docker exec into the running container and run python runner.py.
Using OpenAI Gym's monitor inside docker may or may not work for you based on your setup. On Windows, VcXsrv works quite well.
Calling solver.play(is_train=False, is_random=False) tells the agent to play a single episode in the environment without random actions or any additional training. Use this with one of the persisted learners in the repo to observe fully-trained performance.