MPDiT: Multi-Patch Global-to-Local Transformer Architecture For Efficient Flow Matching and Diffusion Model
This work addresses computational bottlenecks in diffusion and flow-matching models for generative AI, representing an incremental improvement in architecture efficiency.
The paper tackles the computational inefficiency of isotropic Diffusion Transformers (DiTs) by introducing a multi-patch hierarchical design that reduces GFLOPs by up to 50% while maintaining good generative performance on ImageNet.
Transformer architectures, particularly Diffusion Transformers (DiTs), have become widely used in diffusion and flow-matching models due to their strong performance compared to convolutional UNets. However, the isotropic design of DiTs processes the same number of patchified tokens in every block, leading to relatively heavy computation during training process. In this work, we introduce a multi-patch transformer design in which early blocks operate on larger patches to capture coarse global context, while later blocks use smaller patches to refine local details. This hierarchical design could reduces computational cost by up to 50\% in GFLOPs while achieving good generative performance. In addition, we also propose improved designs for time and class embeddings that accelerate training convergence. Extensive experiments on the ImageNet dataset demonstrate the effectiveness of our architectural choices. Code is released at \url{https://github.com/quandao10/MPDiT}