Saved in:
Bibliographic Details
Main Authors: Huang, Chenghuan, Xu, Zhigeng, Sun, Chong, Li, Chen, Ma, Ziyang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.14022
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • Multi-scale deformable attention (MSDA) is a flexible and powerful feature extraction mechanism for visual tasks, but its random-access grid sampling strategy poses significant optimization challenges, especially on domain-specific accelerators such as NPUs. In this work, we present a co-design approach that systematically rethinks memory access and computation strategies for MSDA on the Ascend NPU architecture. With this co-design approach, our implementation supports both efficient forward and backward computation, is fully adapted for training workloads, and incorporates a suite of hardware-aware optimizations. Extensive experiments show that our solution achieves up to $5.9\times$ (forward), $8.9\times$ (backward), and $7.3\times$ (end-to-end training) speedup over the grid sample-based baseline, and $1.9\times$, $2.4\times$, and $2.0\times$ acceleration over the latest vendor library, respectively.