DSV: Exploiting Dynamic Sparsity to Accelerate Large-Scale Video DiT Training

X Tan, Y Chen, Y Jiang, X Chen, K Yan, N Duan… - arxiv preprint arxiv …, 2025 - arxiv.org
Diffusion Transformers (DiTs) have shown remarkable performance in modeling and
generating high-quality videos. However, the quadratic computational complexity of 3D full …

Importance-based Token Merging for Diffusion Models

H Wu, J Xu, H Le, D Samaras - arxiv preprint arxiv:2411.16720, 2024 - arxiv.org
Diffusion models excel at high-quality image and video generation. However, a major
drawback is their high latency. A simple yet powerful way to speed them up is by merging …

AsymRnR: Video Diffusion Transformers Acceleration with Asymmetric Reduction and Restoration

W Sun, RC Tu, J Liao, Z **, D Tao - arxiv preprint arxiv:2412.11706, 2024 - arxiv.org
Video Diffusion Transformers (DiTs) have demonstrated significant potential for generating
high-fidelity videos but are computationally intensive. Existing acceleration methods include …