LoCo: Low-Bit Communication Adaptor for Large-scale Model Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Xingyu, Lin, Zhijie, Toh, Kim-Chuan, Zhou, Pan
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910720119013376
author Xie, Xingyu
Lin, Zhijie
Toh, Kim-Chuan
Zhou, Pan
author_facet Xie, Xingyu
Lin, Zhijie
Toh, Kim-Chuan
Zhou, Pan
contents To efficiently train large-scale models, low-bit gradient communication compresses full-precision gradients on local GPU nodes into low-precision ones for higher gradient synchronization efficiency among GPU nodes. However, it often degrades training quality due to compression information loss. To address this, we propose the Low-bit Communication Adaptor (LoCo), which compensates gradients on local GPU nodes before compression, ensuring efficient synchronization without compromising training quality. Specifically, LoCo designs a moving average of historical compensation errors to stably estimate concurrent compression error and then adopts it to compensate for the concurrent gradient compression, yielding a less lossless compression. This mechanism allows it to be compatible with general optimizers like Adam and sharding strategies like FSDP. Theoretical analysis shows that integrating LoCo into full-precision optimizers like Adam and SGD does not impair their convergence speed on nonconvex problems. Experimental results show that across large-scale model training frameworks like Megatron-LM and PyTorch's FSDP, LoCo significantly improves communication efficiency, e.g., improving Adam's training speed by 14% to 40% without performance degradation on large language models like LLAMAs and MoE.
format Preprint
id arxiv_https___arxiv_org_abs_2407_04480
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LoCo: Low-Bit Communication Adaptor for Large-scale Model Training
Xie, Xingyu
Lin, Zhijie
Toh, Kim-Chuan
Zhou, Pan
Machine Learning
Optimization and Control
To efficiently train large-scale models, low-bit gradient communication compresses full-precision gradients on local GPU nodes into low-precision ones for higher gradient synchronization efficiency among GPU nodes. However, it often degrades training quality due to compression information loss. To address this, we propose the Low-bit Communication Adaptor (LoCo), which compensates gradients on local GPU nodes before compression, ensuring efficient synchronization without compromising training quality. Specifically, LoCo designs a moving average of historical compensation errors to stably estimate concurrent compression error and then adopts it to compensate for the concurrent gradient compression, yielding a less lossless compression. This mechanism allows it to be compatible with general optimizers like Adam and sharding strategies like FSDP. Theoretical analysis shows that integrating LoCo into full-precision optimizers like Adam and SGD does not impair their convergence speed on nonconvex problems. Experimental results show that across large-scale model training frameworks like Megatron-LM and PyTorch's FSDP, LoCo significantly improves communication efficiency, e.g., improving Adam's training speed by 14% to 40% without performance degradation on large language models like LLAMAs and MoE.
title LoCo: Low-Bit Communication Adaptor for Large-scale Model Training
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2407.04480