PAROAttention: Pattern-Aware ReOrdering for Efficient Sparse and Quantized Attention in Visual Generation Models
Abstract
PAROAttention reorganizes visual attention patterns to enable efficient sparsification and quantization, reducing memory and computational costs with minimal impact on performance.
In visual generation, the quadratic complexity of attention mechanisms results in high memory and computational costs, especially for longer token sequences required in high-resolution image or multi-frame video generation. To address this, prior research has explored techniques such as sparsification and quantization. However, these techniques face significant challenges under low density and reduced bitwidths. Through systematic analysis, we identify that the core difficulty stems from the dispersed and irregular characteristics of visual attention patterns. Therefore, instead of introducing specialized sparsification and quantization design to accommodate such patterns, we propose an alternative strategy: *reorganizing* the attention pattern to alleviate the challenges. Inspired by the local aggregation nature of visual feature extraction, we design a novel **Pattern-Aware token ReOrdering (PARO)** technique, which unifies the diverse attention patterns into a hardware-friendly block-wise pattern. This unification substantially simplifies and enhances both sparsification and quantization. We evaluate the performance-efficiency trade-offs of various design choices and finalize a methodology tailored for the unified pattern. Our approach, **PAROAttention**, achieves video and image generation with lossless metrics, and nearly identical results from full-precision (FP) baselines, while operating at notably lower density (~20%-30%) and bitwidth (**INT8/INT4**), achieving a **1.9x** to **2.7x** end-to-end latency speedup.
Community
We propose PAROAttention: A simple yet effective approach to enhance the efficiency of attention in visual generative models. By taking a novel alternative approach of "reorganize" the attention pattern through "Pattern-Aware token ReOrder", it simutaneously improves the attention sparsification and quantization, achieving superior performance preservation and hardware efficiency. PAROAttention achieves 20% density (5x sparse), and INT4 quantization for both QK and PV computing on mainstream video (CogVideo, Wan) and image generation (Flux) models, speedup attention computing by 3-10x and 2-4x end-to-end acceleration while maintaining generation quality.
- Please refer to Project Page: https://a-suozhang.xyz/paroattn.github.io/ for more details
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- FPSAttention: Training-Aware FP8 and Sparsity Co-Design for Fast Video Diffusion (2025)
- Sparse-vDiT: Unleashing the Power of Sparse Attention to Accelerate Video Diffusion Transformers (2025)
- Re-ttention: Ultra Sparse Visual Generation via Attention Statistical Reshape (2025)
- DraftAttention: Fast Video Diffusion via Low-Resolution Attention Guidance (2025)
- Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation (2025)
- Training-Free Efficient Video Generation via Dynamic Token Carving (2025)
- SALE : Low-bit Estimation for Efficient Sparse Attention in Long-context LLM Prefilling (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper