MetaMix: Meta-state Precision Searcher for Mixed-precision Activation Quantization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Han-Byul, Lee, Joo Hyung, Yoo, Sungjoo, Kim, Hong-Seok
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911831898980352
author Kim, Han-Byul
Lee, Joo Hyung
Yoo, Sungjoo
Kim, Hong-Seok
author_facet Kim, Han-Byul
Lee, Joo Hyung
Yoo, Sungjoo
Kim, Hong-Seok
contents Mixed-precision quantization of efficient networks often suffer from activation instability encountered in the exploration of bit selections. To address this problem, we propose a novel method called MetaMix which consists of bit selection and weight training phases. The bit selection phase iterates two steps, (1) the mixed-precision-aware weight update, and (2) the bit-search training with the fixed mixed-precision-aware weights, both of which combined reduce activation instability in mixed-precision quantization and contribute to fast and high-quality bit selection. The weight training phase exploits the weights and step sizes trained in the bit selection phase and fine-tunes them thereby offering fast training. Our experiments with efficient and hard-to-quantize networks, i.e., MobileNet v2 and v3, and ResNet-18 on ImageNet show that our proposed method pushes the boundary of mixed-precision quantization, in terms of accuracy vs. operations, by outperforming both mixed- and single-precision SOTA methods.
format Preprint
id arxiv_https___arxiv_org_abs_2311_06798
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MetaMix: Meta-state Precision Searcher for Mixed-precision Activation Quantization
Kim, Han-Byul
Lee, Joo Hyung
Yoo, Sungjoo
Kim, Hong-Seok
Machine Learning
Computer Vision and Pattern Recognition
Mixed-precision quantization of efficient networks often suffer from activation instability encountered in the exploration of bit selections. To address this problem, we propose a novel method called MetaMix which consists of bit selection and weight training phases. The bit selection phase iterates two steps, (1) the mixed-precision-aware weight update, and (2) the bit-search training with the fixed mixed-precision-aware weights, both of which combined reduce activation instability in mixed-precision quantization and contribute to fast and high-quality bit selection. The weight training phase exploits the weights and step sizes trained in the bit selection phase and fine-tunes them thereby offering fast training. Our experiments with efficient and hard-to-quantize networks, i.e., MobileNet v2 and v3, and ResNet-18 on ImageNet show that our proposed method pushes the boundary of mixed-precision quantization, in terms of accuracy vs. operations, by outperforming both mixed- and single-precision SOTA methods.
title MetaMix: Meta-state Precision Searcher for Mixed-precision Activation Quantization
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2311.06798