Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Taowen, Andronic, Marta, Gündüz, Deniz, Constantinides, George A.
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2511.00874
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917055137054720
author Liu, Taowen
Andronic, Marta
Gündüz, Deniz
Constantinides, George A.
author_facet Liu, Taowen
Andronic, Marta
Gündüz, Deniz
Constantinides, George A.
contents LLM training is resource-intensive. Quantized training improves computational and memory efficiency but introduces quantization noise, which can hinder convergence and degrade model accuracy. Stochastic Rounding (SR) has emerged as a theoretically attractive alternative to deterministic rounding, offering unbiased gradient estimates. However, its interaction with other training factors -- especially batch size -- remains under explored. In this paper, we present a theoretical and empirical study of mini-batch stochastic gradient descent (SGD) with SR, showing that increased batch sizes can compensate for reduced precision during back-propagation. Furthermore, we show that quantizing weights and activations impacts gradient variance in distinct ways. Our experiments validate these theoretical insights.
format Preprint
id arxiv_https___arxiv_org_abs_2511_00874
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Training with Fewer Bits: Unlocking Edge LLMs Training with Stochastic Rounding
Liu, Taowen
Andronic, Marta
Gündüz, Deniz
Constantinides, George A.
Machine Learning
Numerical Analysis
LLM training is resource-intensive. Quantized training improves computational and memory efficiency but introduces quantization noise, which can hinder convergence and degrade model accuracy. Stochastic Rounding (SR) has emerged as a theoretically attractive alternative to deterministic rounding, offering unbiased gradient estimates. However, its interaction with other training factors -- especially batch size -- remains under explored. In this paper, we present a theoretical and empirical study of mini-batch stochastic gradient descent (SGD) with SR, showing that increased batch sizes can compensate for reduced precision during back-propagation. Furthermore, we show that quantizing weights and activations impacts gradient variance in distinct ways. Our experiments validate these theoretical insights.
title Training with Fewer Bits: Unlocking Edge LLMs Training with Stochastic Rounding
topic Machine Learning
Numerical Analysis
url https://arxiv.org/abs/2511.00874