This repository provides a PyTorch implementation of the Graph Attention Network architecture introduced by Veličković et al. (2018). The GAT applies self-attention mechanisms to graph nodes to learn importance weights between connected vertices, enabling inductive learning on graph data. It includes training code for transductive learning tasks (Cora dataset), benchmark performance metrics, and both dense and sparse matrix implementations. The project is forked from the classic PyGCN repository and references the original TensorFlow implementation.