PipeDiT: Accelerating Diffusion Transformers in Video Generation with Task Pipelining and Model Decoupling.

Sijie Wang, Qiang Wang, Shaohuai Shi

Browse the full AAAI paper archive.