Transformer Encoder: The ViT model consists of multiple transformer encoder layers that process the sequence of patch embeddings using self-attention mechanisms to capture global dependencies.