SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers
This addresses the problem of slow inference for DiT models, enabling real-time applications and broader accessibility, though it is incremental as it builds on existing DiT architectures.
The paper tackles the computational expense of Diffusion Transformers (DiT) during inference by introducing SmoothCache, a model-agnostic acceleration technique that caches and reuses features based on layer similarity across timesteps, achieving speed-ups of 8% to 71% while maintaining or improving generation quality across image, video, and audio tasks.
Diffusion Transformers (DiT) have emerged as powerful generative models for various tasks, including image, video, and speech synthesis. However, their inference process remains computationally expensive due to the repeated evaluation of resource-intensive attention and feed-forward modules. To address this, we introduce SmoothCache, a model-agnostic inference acceleration technique for DiT architectures. SmoothCache leverages the observed high similarity between layer outputs across adjacent diffusion timesteps. By analyzing layer-wise representation errors from a small calibration set, SmoothCache adaptively caches and reuses key features during inference. Our experiments demonstrate that SmoothCache achieves 8% to 71% speed up while maintaining or even improving generation quality across diverse modalities. We showcase its effectiveness on DiT-XL for image generation, Open-Sora for text-to-video, and Stable Audio Open for text-to-audio, highlighting its potential to enable real-time applications and broaden the accessibility of powerful DiT models.