DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformers
This addresses the high computational cost of DiTs for image and video generation, though it is an incremental improvement focused on inference efficiency.
The paper tackles the computational inefficiency of Diffusion Transformers (DiTs) by proposing dynamic tokenization, a test-time strategy that varies patch sizes based on content complexity and denoising timestep, achieving up to 3.52× speedup on FLUX-1.Dev and 3.2× on Wan 2.1 without compromising quality.
Diffusion Transformers (DiTs) have achieved state-of-the-art performance in image and video generation, but their success comes at the cost of heavy computation. This inefficiency is largely due to the fixed tokenization process, which uses constant-sized patches throughout the entire denoising phase, regardless of the content's complexity. We propose dynamic tokenization, an efficient test-time strategy that varies patch sizes based on content complexity and the denoising timestep. Our key insight is that early timesteps only require coarser patches to model global structure, while later iterations demand finer (smaller-sized) patches to refine local details. During inference, our method dynamically reallocates patch sizes across denoising steps for image and video generation and substantially reduces cost while preserving perceptual generation quality. Extensive experiments demonstrate the effectiveness of our approach: it achieves up to $3.52\times$ and $3.2\times$ speedup on FLUX-1.Dev and Wan $2.1$, respectively, without compromising the generation quality and prompt adherence.