python3 vamperouge.py {adresseIP} {port}
pytorch and numpy are required.
Related to Alpha Zero
- MCTS implemented in
mcts.py - DNN implemented in
model.py - self-play implemented in
self_play.pyandarena.py
Related to Vampires VS Werewolves
- game logic implemented in
game.py - TCP client implemented in
tcp_client.py
Run python3 main.py to start training Vamperouge (don't run this on your machine, it uses a LOT of RAM).
This implementation of the Alpha Zero algorithm was inspired by:
- https://github.com/suragnair/alpha-zero-general (main reference)
- http://web.stanford.edu/~surag/posts/alphazero.html
- https://arxiv.org/pdf/1712.01815.pdf
- https://discovery.ucl.ac.uk/id/eprint/10045895/1/agz_unformatted_nature.pdf
- https://medium.com/oracledevs/lessons-from-implementing-alphazero-7e36e9054191
- https://medium.com/applied-data-science/how-to-build-your-own-alphazero-ai-using-python-and-keras-7f664945c188
Also, for game-specific things, credits to: