Training-free and Adaptive Sparse Attention for Efficient Long Video Generation

• URL: http://arxiv.org/abs/2502.21079v1
• Date: Fri, 28 Feb 2025 14:11:20 GMT
• Title: Training-free and Adaptive Sparse Attention for Efficient Long Video Generation
• Authors: Yifei Xia, Suhan Ling, Fangcheng Fu, Yujie Wang, Huixia Li, Xuefeng Xiao, Bin Cui,
• Abstract summary: generating high-fidelity long videos with Diffusion Transformers (DiTs) is often hindered by significant latency.<n>We propose AdaSpa, the first Dynamic Pattern and Online Precise Search sparse attention method.<n>AdaSpa is implemented as an adaptive, plug-and-play solution and can be integrated seamlessly with existing DiTs.
• Score: 31.615453637053793
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Generating high-fidelity long videos with Diffusion Transformers (DiTs) is often hindered by significant latency, primarily due to the computational demands of attention mechanisms. For instance, generating an 8-second 720p video (110K tokens) with HunyuanVideo takes about 600 PFLOPs, with around 500 PFLOPs consumed by attention computations. To address this issue, we propose AdaSpa, the first Dynamic Pattern and Online Precise Search sparse attention method. Firstly, to realize the Dynamic Pattern, we introduce a blockified pattern to efficiently capture the hierarchical sparsity inherent in DiTs. This is based on our observation that sparse characteristics of DiTs exhibit hierarchical and blockified structures between and within different modalities. This blockified approach significantly reduces the complexity of attention computation while maintaining high fidelity in the generated videos. Secondly, to enable Online Precise Search, we propose the Fused LSE-Cached Search with Head-adaptive Hierarchical Block Sparse Attention. This method is motivated by our finding that DiTs' sparse pattern and LSE vary w.r.t. inputs, layers, and heads, but remain invariant across denoising steps. By leveraging this invariance across denoising steps, it adapts to the dynamic nature of DiTs and allows for precise, real-time identification of sparse indices with minimal overhead. AdaSpa is implemented as an adaptive, plug-and-play solution and can be integrated seamlessly with existing DiTs, requiring neither additional fine-tuning nor a dataset-dependent profiling. Extensive experiments validate that AdaSpa delivers substantial acceleration across various models while preserving video quality, establishing itself as a robust and scalable approach to efficient video generation.

Related papers

• Generating, Fast and Slow: Scalable Parallel Video Generation with Video Interface Networks [21.710127132217526]
  We introduce a new paradigm called Video Interface Networks (VINs), which augment DiTs with an abstraction module to enable parallel inference of video chunks. VINs encode global semantics from the noisy input of local chunks and the encoded representations, in turn, guide DiTs in denoising chunks in parallel. Our approach attains state-of-the-art motion smoothness while using 25-40% fewer FLOPs than full generation.
  arXiv  Detail & Related papers  (2025-03-21T21:13:02Z)
• Accelerating Vision Diffusion Transformers with Skip Branches [47.07564477125228]
  Diffusion Transformers (DiT) are an emerging image and video generation model architecture.<n>DiT's practical deployment is constrained by computational complexity and redundancy in the sequential denoising process.<n>We introduce Skip-DiT, which converts standard DiT into Skip-DiT with skip branches to enhance feature smoothness.<n>We also introduce Skip-Cache which utilizes the skip branches to cache DiT features across timesteps at the inference time.
  arXiv  Detail & Related papers  (2024-11-26T17:28:10Z)
• Identity-Preserving Text-to-Video Generation by Frequency Decomposition [52.19475797580653]
  Identity-preserving text-to-video (IPT2V) generation aims to create high-fidelity videos with consistent human identity.<n>This paper pushes the technical frontier of IPT2V in two directions that have not been resolved in literature.<n>We propose ConsisID, a tuning-free DiT-based controllable IPT2V model to keep human identity consistent in the generated video.
  arXiv  Detail & Related papers  (2024-11-26T13:58:24Z)
• Adaptive Caching for Faster Video Generation with Diffusion Transformers [52.73348147077075]
  Diffusion Transformers (DiTs) rely on larger models and heavier attention mechanisms, resulting in slower inference speeds. We introduce a training-free method to accelerate video DiTs, termed Adaptive Caching (AdaCache) We also introduce a Motion Regularization (MoReg) scheme to utilize video information within AdaCache, controlling the compute allocation based on motion content.
  arXiv  Detail & Related papers  (2024-11-04T18:59:44Z)
• SparseTem: Boosting the Efficiency of CNN-Based Video Encoders by Exploiting Temporal Continuity [15.872209884833977]
  We propose a memory-efficient scheduling method to eliminate memory overhead and an online adjustment mechanism to minimize accuracy degradation. SparseTem achieves speedup of 1.79x for EfficientDet and 4.72x for CRNN, with minimal accuracy drop and no additional memory overhead.
  arXiv  Detail & Related papers  (2024-10-28T07:13:25Z)
• Dynamic Diffusion Transformer [67.13876021157887]
  Diffusion Transformer (DiT) has demonstrated superior performance but suffers from substantial computational costs. We propose Dynamic Diffusion Transformer (DyDiT), an architecture that dynamically adjusts its computation along both timestep and spatial dimensions during generation. With 3% additional fine-tuning, our method reduces the FLOPs of DiT-XL by 51%, accelerates generation by 1.73, and achieves a competitive FID score of 2.07 on ImageNet.
  arXiv  Detail & Related papers  (2024-10-04T14:14:28Z)
• SIGMA:Sinkhorn-Guided Masked Video Modeling [69.31715194419091]
  Sinkhorn-guided Masked Video Modelling ( SIGMA) is a novel video pretraining method. We distribute features of space-time tubes evenly across a limited number of learnable clusters. Experimental results on ten datasets validate the effectiveness of SIGMA in learning more performant, temporally-aware, and robust video representations.
  arXiv  Detail & Related papers  (2024-07-22T08:04:09Z)
• Transform-Equivariant Consistency Learning for Temporal Sentence Grounding [66.10949751429781]
  We introduce a novel Equivariant Consistency Regulation Learning framework to learn more discriminative representations for each video. Our motivation comes from that the temporal boundary of the query-guided activity should be consistently predicted. In particular, we devise a self-supervised consistency loss module to enhance the completeness and smoothness of the augmented video.
  arXiv  Detail & Related papers  (2023-05-06T19:29:28Z)
• Multi-Density Attention Network for Loop Filtering in Video Compression [9.322800480045336]
  We propose a on-line scaling based multi-density attention network for loop filtering in video compression. Experimental results show that 10.18% bit-rate reduction at the same video quality can be achieved over the latest Versatile Video Coding (VVC) standard.
  arXiv  Detail & Related papers  (2021-04-08T05:46:38Z)
• Efficient Two-Stream Network for Violence Detection Using Separable Convolutional LSTM [0.0]
  We propose an efficient two-stream deep learning architecture leveraging Separable Convolutional LSTM (SepConvLSTM) and pre-trained MobileNet. SepConvLSTM is constructed by replacing convolution operation at each gate of ConvLSTM with a depthwise separable convolution. Our model outperforms the accuracy on the larger and more challenging RWF-2000 dataset by more than a 2% margin.
  arXiv  Detail & Related papers  (2021-02-21T12:01:48Z)
• Perceptron Synthesis Network: Rethinking the Action Scale Variances in Videos [48.57686258913474]
  Video action recognition has been partially addressed by the CNNs stacking of fixed-size 3D kernels. We propose to learn the optimal-scale kernels from the data. An textitaction perceptron synthesizer is proposed to generate the kernels from a bag of fixed-size kernels.
  arXiv  Detail & Related papers  (2020-07-22T14:22:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.