Saved in:
Bibliographic Details
Main Authors: Nair, Pranav, Datta, Puranjay, Dean, Jeff, Jain, Prateek, Kusupati, Aditya
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.06786
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916640511229952
author Nair, Pranav
Datta, Puranjay
Dean, Jeff
Jain, Prateek
Kusupati, Aditya
author_facet Nair, Pranav
Datta, Puranjay
Dean, Jeff
Jain, Prateek
Kusupati, Aditya
contents Quantizing model weights is critical for reducing the communication and inference costs of large models. However, quantizing models -- especially to low precisions like int4 or int2 -- requires a trade-off in model quality; int2, in particular, is known to severely degrade model quality. Consequently, practitioners are often forced to maintain multiple models with different quantization levels or serve a single model that best satisfies the quality-latency trade-off. On the other hand, integer data types, such as int8, inherently possess a nested (Matryoshka) structure where smaller bit-width integers, like int4 or int2, are nested within the most significant bits. Leveraging this insight, in this paper, we propose Matryoshka Quantization (MatQuant), a novel multi-scale quantization technique that alleviates the aforementioned challenge. This technique allows us to train and maintain a single quantized model but serve it with the precision demanded by the deployment. Furthermore, leveraging MatQuant's co-training and co-distillation regularization, int2 precision models extracted by MatQuant outperform standard int2 quantization by up to to 4% and 7% with OmniQuant and QAT as base algorithms respectively. Finally, we demonstrate that by using an extra bit to represent outliers, a model with an effective precision of 2.05-bit gives an additional 6% improvement with OmniQuant as the base algorithm.
format Preprint
id arxiv_https___arxiv_org_abs_2502_06786
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Matryoshka Quantization
Nair, Pranav
Datta, Puranjay
Dean, Jeff
Jain, Prateek
Kusupati, Aditya
Machine Learning
Artificial Intelligence
Quantizing model weights is critical for reducing the communication and inference costs of large models. However, quantizing models -- especially to low precisions like int4 or int2 -- requires a trade-off in model quality; int2, in particular, is known to severely degrade model quality. Consequently, practitioners are often forced to maintain multiple models with different quantization levels or serve a single model that best satisfies the quality-latency trade-off. On the other hand, integer data types, such as int8, inherently possess a nested (Matryoshka) structure where smaller bit-width integers, like int4 or int2, are nested within the most significant bits. Leveraging this insight, in this paper, we propose Matryoshka Quantization (MatQuant), a novel multi-scale quantization technique that alleviates the aforementioned challenge. This technique allows us to train and maintain a single quantized model but serve it with the precision demanded by the deployment. Furthermore, leveraging MatQuant's co-training and co-distillation regularization, int2 precision models extracted by MatQuant outperform standard int2 quantization by up to to 4% and 7% with OmniQuant and QAT as base algorithms respectively. Finally, we demonstrate that by using an extra bit to represent outliers, a model with an effective precision of 2.05-bit gives an additional 6% improvement with OmniQuant as the base algorithm.
title Matryoshka Quantization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2502.06786