International Journal For Multidisciplinary Research

E-ISSN: 2582-2160   •   Impact Factor: 9.24

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Call for Paper Volume 7, Issue 2 (March-April 2025) Submit your research before last 3 days of April to publish your research paper in the issue of March-April.
Submit Research Paper Join as a Reviewer

Plagiarism is checked by the leading plagiarism checker TurnItIn

Call for Paper

Volume 7 Issue 2
March-April 2025

Submit your research paper

Indexing Partners
Academia Advanced Sciences Index Bielefeld Academic Search Engine CiteSeer DRJI Google Scholar Independent Search Engine & Directory Network ISI (International Scientific Indexing) Issuu Mendeley Research Networks
RefSeek ResearcherId - Thomson Reuters ResearchGate Scirus Scribd Semantic Scholar UTeM - Universiti Teknikal Malaysia Melaka Wiki for Call for Papers WorldCat Zenodo

Architectural Evolution in Distributed Training: From Parameter Servers to Zero Redundancy Systems

Author(s) Aditya Singh
Country United States
Abstract The rapid evolution of deep learning models has necessitated fundamental changes in distributed training architectures. This article comprehensively reviews the architectural transformation in distributed training systems, from the traditional parameter server approaches to modern innovations like Ring-AllReduce and pipeline parallelism. The article examines how these architectural advances, coupled with the Zero Redundancy Optimizer (ZeRO), have addressed the critical challenges of memory efficiency and hardware utilization in large-scale model training. The article further analyzes the synergy between architectural innovations and optimization algorithms, particularly focusing on Layer-wise Adaptive Moments optimizer for Batching training (LAMB) and Layer-wise Adaptive Rate Scaling (LARS), which enable stable training with large batch sizes. The article also explores various gradient compression and quantization techniques that reduce communication overhead while maintaining model quality. The analysis reveals how these combined advances have revolutionized the training of large-scale models, enabling unprecedented model sizes while maintaining computational efficiency. The article discusses emerging challenges and future directions in distributed training architectures, particularly focusing on system complexity, fault tolerance, and energy efficiency considerations.
Keywords Keywords: Distributed Training Architectures, Ring-AllReduce Networks, Pipeline Parallelism, Large-Scale Optimization, Neural Network Scaling.
Field Computer
Published In Volume 6, Issue 6, November-December 2024
Published On 2024-12-29
DOI https://doi.org/10.36948/ijfmr.2024.v06i06.34214
Short DOI https://doi.org/g8xgmm
View / Download PDF File

Share this



E-ISSN 2582-2160
CrossRef

CrossRef DOI is assigned to each research paper published in our journal.

IJFMR DOI prefix is
10.36948/ijfmr

Downloads
Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 7 Isu 2Cover Page Vol 7 Isu 1Cover Page Vol 6 Isu 6

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA Open Access

About IJFMR
Fees & Payment
Current Issue
Publication Archive 		Submit Research Paper
Track Submission Status
Publication Guidelines
Publication Ethics
Peer Review & Plagiarism 		Join as a Reviewer
Editors & Reviewers
Reviewer Referral Program
Get Reviewer Membership Certi. 		Website/Journal Policies
Usage Policy
Content Policies
Privacy Policy
Contact Us +91-9687-828-838 editor@ijfmr.com