FlexiDiT: Your Diffusion Transformer Can Easily Generate High-Quality Samples with Less Compute
This addresses the resource-intensive inference problem for users of diffusion models, offering a practical improvement that is incremental but impactful.
The paper tackles the high computational cost of Diffusion Transformers during inference by introducing a dynamic compute allocation strategy, resulting in a 40% reduction in FLOPs for image generation and up to 75% for video generation without quality loss.
Despite their remarkable performance, modern Diffusion Transformers are hindered by substantial resource requirements during inference, stemming from the fixed and large amount of compute needed for each denoising step. In this work, we revisit the conventional static paradigm that allocates a fixed compute budget per denoising iteration and propose a dynamic strategy instead. Our simple and sample-efficient framework enables pre-trained DiT models to be converted into \emph{flexible} ones -- dubbed FlexiDiT -- allowing them to process inputs at varying compute budgets. We demonstrate how a single \emph{flexible} model can generate images without any drop in quality, while reducing the required FLOPs by more than $40$\% compared to their static counterparts, for both class-conditioned and text-conditioned image generation. Our method is general and agnostic to input and conditioning modalities. We show how our approach can be readily extended for video generation, where FlexiDiT models generate samples with up to $75$\% less compute without compromising performance.