THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Minghao, Basat, Ran Ben, Vargaftik, Shay, Lao, ChonLam, Xu, Kevin, Mitzenmacher, Michael, Yu, Minlan
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910354475319296
author Li, Minghao
Basat, Ran Ben
Vargaftik, Shay
Lao, ChonLam
Xu, Kevin
Mitzenmacher, Michael
Yu, Minlan
author_facet Li, Minghao
Basat, Ran Ben
Vargaftik, Shay
Lao, ChonLam
Xu, Kevin
Mitzenmacher, Michael
Yu, Minlan
contents Deep neural networks (DNNs) are the de facto standard for essential use cases, such as image classification, computer vision, and natural language processing. As DNNs and datasets get larger, they require distributed training on increasingly larger clusters. A main bottleneck is the resulting communication overhead where workers exchange model updates (i.e., gradients) on a per-round basis. To address this bottleneck and accelerate training, a widely-deployed approach is compression. However, previous deployments often apply bi-directional compression schemes by simply using a uni-directional gradient compression scheme in each direction. This results in significant computational overheads at the parameter server and increased compression error, leading to longer training and lower accuracy. We introduce Tensor Homomorphic Compression (THC), a novel bi-directional compression framework that enables the direct aggregation of compressed values and thus eliminating the aforementioned computational overheads. Moreover, THC is compatible with in-network aggregation (INA), which allows for further acceleration. Our evaluation shows that training representative vision and language models with THC reaches target accuracy by 1.40x to 1.47x faster using INA and 1.28x to 1.33x faster using a software PS compared with state-of-the-art systems.
format Preprint
id arxiv_https___arxiv_org_abs_2302_08545
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression
Li, Minghao
Basat, Ran Ben
Vargaftik, Shay
Lao, ChonLam
Xu, Kevin
Mitzenmacher, Michael
Yu, Minlan
Machine Learning
Artificial Intelligence
Networking and Internet Architecture
Deep neural networks (DNNs) are the de facto standard for essential use cases, such as image classification, computer vision, and natural language processing. As DNNs and datasets get larger, they require distributed training on increasingly larger clusters. A main bottleneck is the resulting communication overhead where workers exchange model updates (i.e., gradients) on a per-round basis. To address this bottleneck and accelerate training, a widely-deployed approach is compression. However, previous deployments often apply bi-directional compression schemes by simply using a uni-directional gradient compression scheme in each direction. This results in significant computational overheads at the parameter server and increased compression error, leading to longer training and lower accuracy. We introduce Tensor Homomorphic Compression (THC), a novel bi-directional compression framework that enables the direct aggregation of compressed values and thus eliminating the aforementioned computational overheads. Moreover, THC is compatible with in-network aggregation (INA), which allows for further acceleration. Our evaluation shows that training representative vision and language models with THC reaches target accuracy by 1.40x to 1.47x faster using INA and 1.28x to 1.33x faster using a software PS compared with state-of-the-art systems.
title THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression
topic Machine Learning
Artificial Intelligence
Networking and Internet Architecture
url https://arxiv.org/abs/2302.08545