Pytorchsweeper is a Minesweeper solver implemented via a convolutional neural net using PyTorch.
Included is a simple implementation of the game Minesweeper using numpy, in game.py, and several utility functions and classes in utils.py and ais.py.
The implementation of Pytorchsweeper is detailed in the three jupyter notebooks. This includes the generation of training data, the definition and training of the network, and the testing of the network. Running the first two notebooks in order will create a trained model, however a prepared version of the model is also included, in prepared_pytorchsweeper_model.pt.