This project is a development environment for creating, testing, and comparing simple chess bots built with different neural architectures using PyTorch, Python-Chess, and DearPyGui. The environment is designed to facilitate easy experimentation with various neural network models and to evaluate their performance in chess games.
- Modular Design: Easily integrate various neural architectures (e.g., CNNs, MLPs) with PyTorch for training chess bots.
- Python-Chess Integration: Utilize the Python-Chess library to handle the game logic, move validation, and game state management.
- Graphical User Interface: Use DearPyGui to create a user-friendly interface for configuring the bot's parameters and running simulations.
- Training and Evaluation: Train the bots using PyTorch with custom datasets, or load pre-trained models. Evaluate performance against human players or other bots.
- PyTorch (for creating and training neural network models)
- Python-Chess (for chess game logic)
- DearPyGui (for the graphical interface)
-
Clone the repository: git clone https://github.com/Mathwizard1/reimagined-octosquare.git
-
Install dependencies: pip install -r requirements.txt
-
Run the application: python gui.py
- Configure Bots: Launch the application using python main.py. Use the DearPyGui interface to select or configure different bots, neural architectures, and game parameters.
Better trained weights Test bots against each other Create Elo system Additional neural architectures like Transformers for better decision-making. More sophisticated evaluation metrics. Support for human players to challenge AI bots via the GUI.
If you wish to contribute to this project, feel free to fork the repository, make your changes, and submit a pull request.