L2PF -- Learning to Prune Faster

TL;DR

This paper introduces Learning to Prune Faster, a multi-task method that accelerates CNN pruning, achieving high compression with minimal accuracy loss and significantly reducing GPU hours needed for optimization.

Contribution

It presents a novel try-and-learn approach that speeds up CNN pruning convergence and reduces computational resources compared to existing methods.

Findings

Achieved 3.84x compression ratio on ResNet20

Reduced GPU hours by 1.71x over state-of-the-art

Minimal accuracy degradation during pruning

Abstract

Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible, and as such, it is not feasible to spend redundant GPU hours during compression. In this context, we present Learning to Prune Faster which details a multi-task, try-and-learn method, discretely learning redundant filters of the CNN and a continuous action of how long the layers have to be fine-tuned. This allows us to significantly speed up the convergence process of learning how to find an embedded-friendly filter-wise pruned CNN. For ResNet20, we have achieved a compression ratio of 3.84 x with minimal accuracy degradation. Compared to the state-of-the-art pruning method, we reduced the GPU hours by 1.71 x.