Smooth Cache: A Universal Inference Acceleration Technique for Diffusion Transformers | IEEE Conference Publication | IEEE Xplore