GTP-ViT: Efficient Vision Transformers via Graph-Based Token Propagation | ComputerVisionFoundation Videos | Podwise