Chess bot based on the transformer architecture

This project explores whether a transformer architecture can serve as a viable chess bot by training an 11M-parameter model on elite human games from Lichess. The model alone achieves around 1500 Elo, making it interesting for AI/ML enthusiasts curious about the limits of transformer-based game agents without traditional search algorithms. It offers a lightweight, data-driven alternative to conventional engines, highlighting both the potential and current constraints of pure neural network play.