Improve your shapes and intuition in the game of Go. Tailered to your weaknesses from your own games. Using human supervised learning for questions and answers of your skill level.

Active Development: 2024-03-26 - 2024-03-29
Last Change: 2025-12-09
Highlights: Machine Learning

X = Game move
O = HSL move

1. KataGo with a model.
2. A HSL model. For example this model from lightvector. Make sure it's a ckpt file.
3. Python. Version 3.10.6 works for me.
4. Run pip install -r requirements.txt to install the necessary python packages.
5. Install PyTorch.
6. Make a copy of Reviewer-TEMPLATE.sh/bat and remove -TEMPLATE from the name. Change the parameters in the script to your own. If you have a dedicated GPU, try cuda:0 as the device. If that doesn't work you can switch to cpu which is slower but will always work.

1. Run Reviewer.sh/bat.
2. Set the color you were in the game you want to review.
3. Optionally change the HSL sliders.
4. Drag & drop your game sgf.
5. Wait until the console says === REVIEW DONE ===

There are also some options you can tweak at the top in hsl_reviewer.py.

1. Run Trainer.sh/bat.
2. Load a card and think of a move.
3. Click on the board to see the answer.
4. Delete the card if you feel the answer is easy.
5. Load the next card.