Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912343510745088 |
|---|---|
| author | Hassani, Ali Zhou, Fengzhe Kane, Aditya Huang, Jiannan Chen, Chieh-Yun Shi, Min Walton, Steven Hoehnerbach, Markus Thakkar, Vijay Isaev, Michael Zhang, Qinsheng Xu, Bing Wu, Haicheng Hwu, Wen-mei Liu, Ming-Yu Shi, Humphrey |
| author_facet | Hassani, Ali Zhou, Fengzhe Kane, Aditya Huang, Jiannan Chen, Chieh-Yun Shi, Min Walton, Steven Hoehnerbach, Markus Thakkar, Vijay Isaev, Michael Zhang, Qinsheng Xu, Bing Wu, Haicheng Hwu, Wen-mei Liu, Ming-Yu Shi, Humphrey |
| contents | Many sparse attention mechanisms such as Neighborhood Attention have typically failed to consistently deliver speedup over the self attention baseline. This is largely due to the level of complexity in attention infrastructure, and the rapid evolution of AI hardware architecture. At the same time, many state-of-the-art foundational models, particularly in computer vision, are heavily bound by attention, and need reliable sparsity to escape the O(n^2) complexity. In this paper, we study a class of promising sparse attention mechanisms that focus on locality, and aim to develop a better analytical model of their performance improvements. We first introduce Generalized Neighborhood Attention (GNA), which can describe sliding window, strided sliding window, and blocked attention. We then consider possible design choices in implementing these approaches, and create a simulator that can provide much more realistic speedup upper bounds for any given setting. Finally, we implement GNA on top of a state-of-the-art fused multi-headed attention (FMHA) kernel designed for the NVIDIA Blackwell architecture in CUTLASS. Our implementation can fully realize the maximum speedup theoretically possible in many perfectly block-sparse cases, and achieves an effective utilization of 1.3 petaFLOPs/second in FP16. In addition, we plug various GNA configurations into off-the-shelf generative models, such as Cosmos-7B, HunyuanVideo, and FLUX, and show that it can deliver 28% to 46% end-to-end speedup on B200 without any fine-tuning. We will open source our simulator and Blackwell kernels directly through the NATTEN project. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_16922 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light Hassani, Ali Zhou, Fengzhe Kane, Aditya Huang, Jiannan Chen, Chieh-Yun Shi, Min Walton, Steven Hoehnerbach, Markus Thakkar, Vijay Isaev, Michael Zhang, Qinsheng Xu, Bing Wu, Haicheng Hwu, Wen-mei Liu, Ming-Yu Shi, Humphrey Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Many sparse attention mechanisms such as Neighborhood Attention have typically failed to consistently deliver speedup over the self attention baseline. This is largely due to the level of complexity in attention infrastructure, and the rapid evolution of AI hardware architecture. At the same time, many state-of-the-art foundational models, particularly in computer vision, are heavily bound by attention, and need reliable sparsity to escape the O(n^2) complexity. In this paper, we study a class of promising sparse attention mechanisms that focus on locality, and aim to develop a better analytical model of their performance improvements. We first introduce Generalized Neighborhood Attention (GNA), which can describe sliding window, strided sliding window, and blocked attention. We then consider possible design choices in implementing these approaches, and create a simulator that can provide much more realistic speedup upper bounds for any given setting. Finally, we implement GNA on top of a state-of-the-art fused multi-headed attention (FMHA) kernel designed for the NVIDIA Blackwell architecture in CUTLASS. Our implementation can fully realize the maximum speedup theoretically possible in many perfectly block-sparse cases, and achieves an effective utilization of 1.3 petaFLOPs/second in FP16. In addition, we plug various GNA configurations into off-the-shelf generative models, such as Cosmos-7B, HunyuanVideo, and FLUX, and show that it can deliver 28% to 46% end-to-end speedup on B200 without any fine-tuning. We will open source our simulator and Blackwell kernels directly through the NATTEN project. |
| title | Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2504.16922 |