A Data-Free Analytical Quantization Scheme for Deep Learning Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Luqman, Ahmed, Qazi, Khuzemah, Patterson, Murray, Khan, Malik Jahan, Khan, Imdadullah
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911143345258496
author Luqman, Ahmed
Qazi, Khuzemah
Patterson, Murray
Khan, Malik Jahan
Khan, Imdadullah
author_facet Luqman, Ahmed
Qazi, Khuzemah
Patterson, Murray
Khan, Malik Jahan
Khan, Imdadullah
contents Despite the success of CNN models on a variety of Image classification and segmentation tasks, their extensive computational and storage demands pose considerable challenges for real-world deployment on resource-constrained devices. Quantization is one technique that aims to alleviate these large storage requirements and speed up the inference process by reducing the precision of model parameters to lower-bit representations. In this paper, we introduce a novel post-training quantization method for model weights. Our method finds optimal clipping thresholds and scaling factors along with mathematical guarantees that our method minimizes quantization noise. Empirical results on real-world datasets demonstrate that our quantization scheme significantly reduces model size and computational requirements while preserving model accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2412_07391
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Data-Free Analytical Quantization Scheme for Deep Learning Models
Luqman, Ahmed
Qazi, Khuzemah
Patterson, Murray
Khan, Malik Jahan
Khan, Imdadullah
Computer Vision and Pattern Recognition
Machine Learning
Despite the success of CNN models on a variety of Image classification and segmentation tasks, their extensive computational and storage demands pose considerable challenges for real-world deployment on resource-constrained devices. Quantization is one technique that aims to alleviate these large storage requirements and speed up the inference process by reducing the precision of model parameters to lower-bit representations. In this paper, we introduce a novel post-training quantization method for model weights. Our method finds optimal clipping thresholds and scaling factors along with mathematical guarantees that our method minimizes quantization noise. Empirical results on real-world datasets demonstrate that our quantization scheme significantly reduces model size and computational requirements while preserving model accuracy.
title A Data-Free Analytical Quantization Scheme for Deep Learning Models
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2412.07391