Saved in:
Bibliographic Details
Main Authors: Liu, Jing, Koike-Akino, Toshiaki, Wang, Ye, Mansour, Hassan, Brand, Matthew
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.10205
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912737202798592
author Liu, Jing
Koike-Akino, Toshiaki
Wang, Ye
Mansour, Hassan
Brand, Matthew
author_facet Liu, Jing
Koike-Akino, Toshiaki
Wang, Ye
Mansour, Hassan
Brand, Matthew
contents To address the enormous size of Large Language Models (LLMs), model compression methods, such as quantization and pruning, are often deployed, especially on edge devices. In this work, we focus on layer-wise post-training quantization and pruning. Drawing connections between activation-aware weight pruning and sparse approximation problems, and motivated by the success of Iterative Hard Thresholding (IHT), we propose a unified method for Activation-aware Weight pruning and quantization via Projected gradient descent (AWP). Our experiments demonstrate that AWP outperforms state-of-the-art LLM pruning and quantization methods. Theoretical convergence guarantees of the proposed method for pruning are also provided.
format Preprint
id arxiv_https___arxiv_org_abs_2506_10205
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AWP: Activation-Aware Weight Pruning and Quantization with Projected Gradient Descent
Liu, Jing
Koike-Akino, Toshiaki
Wang, Ye
Mansour, Hassan
Brand, Matthew
Machine Learning
To address the enormous size of Large Language Models (LLMs), model compression methods, such as quantization and pruning, are often deployed, especially on edge devices. In this work, we focus on layer-wise post-training quantization and pruning. Drawing connections between activation-aware weight pruning and sparse approximation problems, and motivated by the success of Iterative Hard Thresholding (IHT), we propose a unified method for Activation-aware Weight pruning and quantization via Projected gradient descent (AWP). Our experiments demonstrate that AWP outperforms state-of-the-art LLM pruning and quantization methods. Theoretical convergence guarantees of the proposed method for pruning are also provided.
title AWP: Activation-Aware Weight Pruning and Quantization with Projected Gradient Descent
topic Machine Learning
url https://arxiv.org/abs/2506.10205