arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.10520cs.CVcs.AIcs.LG
收藏

基于自适应激活的结构化剪枝

Adaptive Activation-based Structured Pruning

Arizona State University(亚利桑那州立大学);Facebook(脸书)

Kaiqi Zhao, Animesh Jain, Ming Zhao

摘要:

修剪是一种很有前途的压缩复杂深度学习模型的方法,以便将它们部署在资源受限的边缘设备上。然而,许多现有的修剪解决方案是基于非结构化修剪,这产生的模型不能有效地运行在商用硬件上,并需要用户手动探索和调整修剪过程,这是耗时的,往往导致次优的结果。为了解决这些限制,本文提出了一种自适应的,基于激活的,结构化的修剪方法,自动和有效地生成小,准确,硬件效率高的模型,满足用户的需求。首先,它提出了使用基于激活的注意力特征图的迭代结构化修剪,以有效地识别和修剪不重要的过滤器。然后,它提出了自适应修剪策略,自动满足修剪目标的精度关键,内存受限,延迟敏感的任务。综合评估表明,该方法在CIFAR-10和ImageNet数据集上的性能大大优于最先进的结构化剪枝工作。例如,在使用CIFAR-10的ResNet-56上,在没有任何准确性下降的情况下,我们的方法实现了最大的参数减少(79.11%),优于相关工作22.81%至66.07%,以及最大的FLOPs减少(70.13%),优于相关工作14.13%至26.53%。

英文摘要:

Pruning is a promising approach to compress complex deep learning models in order to deploy them on resource-constrained edge devices. However, many existing pruning solutions are based on unstructured pruning, which yields models that cannot efficiently run on commodity hardware and require users to manually explore and tune the pruning process, which is time-consuming and often leads to sub-optimal results. To address these limitations, this paper presents an adaptive, activation-based, structured pruning approach to automatically and efficiently generate small, accurate, and hardware-efficient models that meet user requirements. First, it proposes iterative structured pruning using activation-based attention feature maps to effectively identify and prune unimportant filters. Then, it proposes adaptive pruning policies for automatically meeting the pruning objectives of accuracy-critical, memory-constrained, and latency-sensitive tasks. A comprehensive evaluation shows that the proposed method can substantially outperform the state-of-the-art structured pruning works on CIFAR-10 and ImageNet datasets. For example, on ResNet-56 with CIFAR-10, without any accuracy drop, our method achieves the largest parameter reduction (79.11%), outperforming the related works by 22.81% to 66.07%, and the largest FLOPs reduction (70.13%), outperforming the related works by 14.13% to 26.53%.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑