ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Saxena, Utkarsh, Sharify, Sayeh, Roy, Kaushik, Wang, Xin
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929696614121472
author Saxena, Utkarsh
Sharify, Sayeh
Roy, Kaushik
Wang, Xin
author_facet Saxena, Utkarsh
Sharify, Sayeh
Roy, Kaushik
Wang, Xin
contents Post-training quantization (PTQ) of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time. Quantization of all weight, activation and key-value (KV) cache tensors to 4-bit without significantly degrading generalizability is challenging, due to the high quantization error caused by extreme outliers in activations. To tackle this problem, we propose ResQ, a PTQ method that pushes further the state-of-the-art. By means of principal component analysis (PCA), it identifies a low-rank subspace (in practice 1/8 of the hidden dimension) in which activation variances are highest, and keep the coefficients within this subspace in high precision, e.g. 8-bit, while quantizing the rest to 4-bit. Within each subspace, invariant random rotation is applied to further suppress outliers. We show that this is a provably optimal mixed precision quantization scheme that minimizes error. With the Llama and Qwen2.5 families of models, we demonstrate that ResQ outperforms recent uniform and mixed precision PTQ methods on a variety of benchmarks, achieving up to 33\% lower perplexity on Wikitext than the next best method SpinQuant, and upto 3\times speedup over 16-bit baseline. Code is available at https://github.com/utkarsh-dmx/project-resq.
format Preprint
id arxiv_https___arxiv_org_abs_2412_14363
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals
Saxena, Utkarsh
Sharify, Sayeh
Roy, Kaushik
Wang, Xin
Machine Learning
Computation and Language
Post-training quantization (PTQ) of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time. Quantization of all weight, activation and key-value (KV) cache tensors to 4-bit without significantly degrading generalizability is challenging, due to the high quantization error caused by extreme outliers in activations. To tackle this problem, we propose ResQ, a PTQ method that pushes further the state-of-the-art. By means of principal component analysis (PCA), it identifies a low-rank subspace (in practice 1/8 of the hidden dimension) in which activation variances are highest, and keep the coefficients within this subspace in high precision, e.g. 8-bit, while quantizing the rest to 4-bit. Within each subspace, invariant random rotation is applied to further suppress outliers. We show that this is a provably optimal mixed precision quantization scheme that minimizes error. With the Llama and Qwen2.5 families of models, we demonstrate that ResQ outperforms recent uniform and mixed precision PTQ methods on a variety of benchmarks, achieving up to 33\% lower perplexity on Wikitext than the next best method SpinQuant, and upto 3\times speedup over 16-bit baseline. Code is available at https://github.com/utkarsh-dmx/project-resq.
title ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2412.14363