This repo contains visualization tools for the paper 'Adversarial Policies Beat Superhuman Go AIs' (website)
Python package containing general tools used throughout the repo, in particular the code to parse sgf files.
Python notebooks used to produce figures for the paper.
Source for the website that accompanies the paper.
Source for a data visualization web app used internally to help with research and monitoring training runs.
KataGoVisualizer/
├── .circleci/
├── Dockerfile
├── go_attack_utils/
│ ├── src/
│ │ └── sgf_parser/
│ │ └── game_info.py
│ └── setup.py
├── notebooks/
│ ├── .devcontainer/
│ ├── notebooks/
│ ├── Dockerfile
│ ├── docker-compose.yml
│ └── requirements.txt
├── sgf-viewer/
│ ├── prepare_data/
│ │ └── prepare_data.py
│ ├── public/
│ ├── src/
│ │ ├── components/
│ │ ├── App.svelte
│ │ └── content.ts
│ ├── index.html
│ └── package.json
└── streamlit_app/
├── Pipfile
├── Pipfile.lock
├── requirements.txt
├── parsing_server.py
├── streamlit_app.py
├── tests/
└── components/
└── subcomponents/