Contains solvers and a framework for easily building new ones. The SolverBase -class takes care of
starting a new game and handling the interactions with it. At its simplest, a new solver can just
inherit from the base class and implement only the logic method, that contains the solver logic.
This solver randomly clicks on cells.
This solver makes decisions based on 3x3 subgrids. If it determines that all the unopened cells in a 3x3 subgrid are mines it flags them, if it thinks that all of them are safe, it opens them. If it can't deduct anymore safe or flaggable cells, it opens a cell that has the lowest probability of uncovering a mine.
This solver is is powered by a reinforcement learning model. DQL is used to train a model to select the safest cell to open.
Clone the repository and run
pip install -e .
with a fresh virtual environment to install the dependencies. If you want to run or train the reinforcement learning models, follow these instructions to install pytorch.
Run the command
msolver --help
for information on the available commands, and
msolver command --help
for information on individual commands.