AsymRnR: Video Diffusion Transformers Acceleration with Asymmetric Reduction and Restoration
Wenhao Sun, Rong-Cheng Tu, Jingyi Liao, Zhao Jin, Dacheng Tao
摘要
HunyuanVideo +AsymRnR (30% faster) (original) avg. latency -703s avg. latency -913s prompt: A person is knitting ... FastVideo-Hunyuan +AsymRnR (24% faster) (original) avg. latency -128s avg. latency -159s prompt: A teddy bear is swimming in the ocean ... Figure 1. Quality and speed comparison between baseline models, HunyuanVideo (Team, 2024c) and FastVideo-Hunyuan (Team, 2024a), with our AsymRnR. Our approach enables training-free, lossless acceleration for state-of-the-art video diffusion transformers.
问 Lune
尽管问，回答会引用原文。
Lune 会读完全文和相关的 Paper，每个回答都附上所依据的原文段落。
引用它的顶会 Paper5
- VORTA: Efficient Video Diffusion via Routing Sparse AttentionWenhao Sun, Rong-Cheng Tu, Yifu Ding, Jingyi Liao 等NeurIPS 2025 · 25 次引用
- DDiT: Dynamic Patch Scheduling for Efficient Diffusion TransformersDahye Kim, Deepti Ghadiyaram, Raghudeep GaddeCVPR 2026 · 3 次引用
- Self-Supervised Direct Preference Optimization for Text-to-Image Diffusion ModelsLiang Peng, Boxi Wu, Haoran Cheng, Yibo Zhao 等NeurIPS 2025 · 2 次引用
- Dlfr-Gen: Diffusion-Based Video Generation With Dynamic Latent Frame RateZhihang Yuan, Rui Xie, Yuzhang Shang, Hanling Zhang 等ICCV 2025 · 1 次引用
- SODA: Sensitivity-Oriented Dynamic Acceleration for Diffusion TransformerTong Shao, Yusen Fu, Guoying Sun, Jingde Kong 等CVPR 2026 · 1 次引用
它引用的 Paper29
- Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 次引用
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser 等CVPR 2022 · 13,123 次引用
- Denoising Diffusion Implicit ModelsJiaming Song, Chenlin Meng, Stefano ErmonICLR 2021 · 11,743 次引用
- Scalable Diffusion Models with TransformersWilliam Peebles, Saining XieICCV 2023 · 5,568 次引用
- Scaling Rectified Flow Transformers for High-Resolution Image SynthesisPatrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari 等ICML 2024 · 3,620 次引用
相关 Paper
- LESA: Learnable Stage-Aware Predictors for Diffusion Model AccelerationPeiliang Cai, Jiacheng Liu, Haowen Xu, Xinyu Wang 等CVPR 2026 · 4 次引用
- ScalingCache: Extreme Acceleration of DiTs through Difference Scaling and Dynamic Interval CachingLihui Gu, Jingbin He, Lianghao Su, Kang He 等ICLR 2026
- AdaCluster: Adaptive Query-Key Clustering for Sparse Attention in Video GenerationHaoyue Tan, Shengnan Wang, Yulin Qiao, Juncheng Zhang 等CVPR 2026 · 5 次引用
- Video-SVD: Efficient Video Diffusion via Orthogonal Basis CompositionZhang Wan, Yu Li, Tianze Huang, Haochen Li 等ICML 2026
- QuantCache: Adaptive Importance-Guided Quantization with Hierarchical Latent and Layer Caching for Video GenerationJunyi Wu, Zhiteng Li, Zheng Hui, Yulun Zhang 等ICCV 2025 · 20 次引用
