An OpenAI gym environment for models to learn playing SudokuTetris, or also known as BlockuDoku (I have no affiliation to Easybrain, the creators of BlockuDoku).
Install the requirements
pip install -r requirements.txt
Then just do
python3 index.py
which will run it with a random agent for 200 episodes.
The action space is discrete vector of shape (N_PIECES*N_BOARD_TILES,) - (19*81,) = 1539. Which ones are available at a given time are included in the info dictionary, the last in the tuple sequence returned by the step-function.
If you like it, don't hesitate to star this repository.
Enjoy!