This repository contains the presentation and code for my talk "How to teach space invaders to your computer" at PyCon.DE 2018 held on 25.10.2018 in Karlsruhe.
presentation/How_to_teach_space_invaders_to_your_computer.pdf is a pdf version of the slides I used during my presentation. The original slides have been interactive, while the pdf is not. The videos of the Space Invaders gameplay are also located in the presentation directory. Other interactive content can be found online, links to the sources are given in the slides.
The repo contains two interactive notebooks.
analyse_training.ipynb contains the code that generated parts of the presentation content, i.e. the videos and plots about the training progress.
run_training.ipynb contains the code for running the training of autoencoder and agent.
Most of the actual training logic is located in training.py while additional supporting code can be found in helpers.py and extractor.py. The results of the training will be placed in a subdirectory under results and actually the results used in the presentation are stored there too.
The notebooks have been executed inside a docker container which contains all required packages and dependencies.
See https://docs.docker.com/get-started/ for an introduction to docker.
See also https://docs.docker.com/compose/overview/ for an introduction to docker-compose.
To handle the containers you may want to execute the following commands inside the root directory of the repository:
Use docker-compose up -d to start the container. Afterwards you can visit http://localhost:8888 in your browser to access the notebook server.
Use docker-compose logs | grep token= | cut -d = -f 2 to retrieve the token required for logging in at juypter.
Use docker-compose down to stop and remove the container.
The training can also be executed on a GPU, to do so overwrite the docker-compose.yml file with the version in gpu_mode/docker-compose.yml. You will need to have nvidia docker installed.
This has been tested and design for linux. If you wish to execute the code under windows you probably need to adapt the docker-compose.yml file especially the volume mount.