Parallel Computing Cluster for Solving Computational Problems in Data: A Framework for High-Throughput Data Processing - Couverture souple

Mutyala Rao, Chavala

 
9786209340734: Parallel Computing Cluster for Solving Computational Problems in Data: A Framework for High-Throughput Data Processing
Couverture souple
ISBN 10 : 6209340733 ISBN 13 : 9786209340734
Editeur : LAP LAMBERT Academic Publishing, 2025

Synopsis

Thisbook tackles a critical bottleneck in large-scale AI: the slow and communication-heavy training of massive Deep Neural Networks (DNNs) on multi-GPU systems. It addresses the trade-off between two main parallelization methods. Data parallelism suffers from severe communication overhead for large models, while pipelined model parallelism (like PipeDream) offers up to 8.91x speedup for large Fully Connected/Recurrent Neural Networks but causes "weight staleness," degrading model accuracy. To resolve this, the paper introduces SpecTrain, a novel technique. SpecTrain uses the momentum from optimizers to predict future weight updates, allowing pipelined computation with accurate, non-stale weights. This enables the high GPU utilization and speed of pipelining while maintaining the training robustness and final accuracy of synchronous methods.

Les informations fournies dans la section « Synopsis » peuvent faire référence à une autre édition de ce titre.

Éditeur
LAP LAMBERT Academic Publishing
Date d'édition
2025
Langue
anglais
ISBN 10 
6209340733
ISBN 13 
9786209340734
Reliure
Broché
Nombre de pages
56
Coordonnées du fabricant
non disponible
Personne responsable
KreGa Onlinehandel GmbH
4950000000000
info@krega-onlinehandel.de

Birkenweg 31
Brakel
33034
Allemagne