Accelerating Deep Neural Networks via Semi-Structured Activation Sparsity

TL;DR

This paper introduces a semi-structured activation sparsity method for deep neural networks that enables significant inference speedups with minimal accuracy loss, suitable for embedded devices.

Contribution

The authors propose a novel semi-structured sparsity technique and a training procedure that considers activation positions, improving inference speed with minimal accuracy degradation.

Findings

Achieves 1.25x speedup on ResNet18 with 1.1% accuracy drop.

Effective across image classification and object detection models.

Enhances structured pruning methods for better latency-accuracy trade-offs.

Abstract

The demand for efficient processing of deep neural networks (DNNs) on embedded devices is a significant challenge limiting their deployment. Exploiting sparsity in the network's feature maps is one of the ways to reduce its inference latency. It is known that unstructured sparsity results in lower accuracy degradation with respect to structured sparsity but the former needs extensive inference engine changes to get latency benefits. To tackle this challenge, we propose a solution to induce semi-structured activation sparsity exploitable through minor runtime modifications. To attain high speedup levels at inference time, we design a sparse training procedure with awareness of the final position of the activations while computing the General Matrix Multiplication (GEMM). We extensively evaluate the proposed solution across various models for image classification and object detection tasks. Remarkably, our approach yields a speed improvement of $1.25 \times$ with a minimal accuracy drop of $1.1\%$ for the ResNet18 model on the ImageNet dataset. Furthermore, when combined with a state-of-the-art structured pruning method, the resulting models provide a good latency-accuracy trade-off, outperforming models that solely employ structured pruning techniques.