Efficient Inferencing of Compressed Deep Neural Networks

TL;DR

This paper introduces parallel algorithms for efficient inference on compressed deep neural networks, especially Huffman-encoded models, improving throughput by 15-25% under memory constraints.

Contribution

It presents novel parallel inference algorithms tailored for compressed models, focusing on Huffman encoding and variable batch sizes to enhance performance.

Findings

Achieves 15-25% inference throughput improvement on AlexNet.

Maintains memory and latency constraints during inference.

Provides algorithms applicable to low-memory environments.

Abstract

Large number of weights in deep neural networks makes the models difficult to be deployed in low memory environments such as, mobile phones, IOT edge devices as well as "inferencing as a service" environments on cloud. Prior work has considered reduction in the size of the models, through compression techniques like pruning, quantization, Huffman encoding etc. However, efficient inferencing using the compressed models has received little attention, specially with the Huffman encoding in place. In this paper, we propose efficient parallel algorithms for inferencing of single image and batches, under various memory constraints. Our experimental results show that our approach of using variable batch size for inferencing achieves 15-25\% performance improvement in the inference throughput for AlexNet, while maintaining memory and latency constraints.