Coquelin, Daniel und Debus, Charlotte und Götz, Markus und von der Lehr, Fabrice und Kahn, James und Siggel, Martin und Streit, Achim (2021) Accelerating Neural Network Training with Distributed Asynchronous and Selective Optimization (DASO). [sonstige Veröffentlichung]
![]() |
PDF
441kB |
Offizielle URL: https://arxiv.org/abs/2104.05588
Kurzfassung
With increasing data and model complexities, the time required to train neural networks has become prohibitively large. To address the exponential rise in training time, users are turning to data parallel neural networks (DPNN) to utilize large-scale distributed resources on computer clusters. Current DPNN approaches implement the network parameter updates by synchronizing and averaging gradients across all processes with blocking communication operations. This synchronization is the central algorithmic bottleneck. To combat this, we introduce the Distributed Asynchronous and Selective Optimization (DASO) method which leverages multi-GPU compute node architectures to accelerate network training. DASO uses a hierarchical and asynchronous communication scheme comprised of node-local and global networks while adjusting the global synchronization rate during the learning process. We show that DASO yields a reduction in training time of up to 34% on classical and state-of-the-art networks, as compared to other existing data parallel training methods.
elib-URL des Eintrags: | https://elib.dlr.de/146819/ | ||||||||||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Dokumentart: | sonstige Veröffentlichung | ||||||||||||||||||||||||||||||||
Titel: | Accelerating Neural Network Training with Distributed Asynchronous and Selective Optimization (DASO) | ||||||||||||||||||||||||||||||||
Autoren: |
| ||||||||||||||||||||||||||||||||
Datum: | 12 April 2021 | ||||||||||||||||||||||||||||||||
Erschienen in: | ArXiV | ||||||||||||||||||||||||||||||||
Referierte Publikation: | Nein | ||||||||||||||||||||||||||||||||
Open Access: | Ja | ||||||||||||||||||||||||||||||||
Seitenanzahl: | 12 | ||||||||||||||||||||||||||||||||
Status: | veröffentlicht | ||||||||||||||||||||||||||||||||
Stichwörter: | Computer Science, Machine Learning, Neural Network, Optimization, Distributed Training, Data parallelism | ||||||||||||||||||||||||||||||||
HGF - Forschungsbereich: | Luftfahrt, Raumfahrt und Verkehr | ||||||||||||||||||||||||||||||||
HGF - Programm: | Raumfahrt | ||||||||||||||||||||||||||||||||
HGF - Programmthema: | Technik für Raumfahrtsysteme | ||||||||||||||||||||||||||||||||
DLR - Schwerpunkt: | Raumfahrt | ||||||||||||||||||||||||||||||||
DLR - Forschungsgebiet: | R SY - Technik für Raumfahrtsysteme | ||||||||||||||||||||||||||||||||
DLR - Teilgebiet (Projekt, Vorhaben): | R - Aufgaben SISTEC | ||||||||||||||||||||||||||||||||
Standort: | Köln-Porz | ||||||||||||||||||||||||||||||||
Institute & Einrichtungen: | Institut für Simulations- und Softwaretechnik > High Performance Computing Institut für Softwaretechnologie Institut für Softwaretechnologie > High-Performance Computing | ||||||||||||||||||||||||||||||||
Hinterlegt von: | von der Lehr, Fabrice | ||||||||||||||||||||||||||||||||
Hinterlegt am: | 09 Dez 2021 09:05 | ||||||||||||||||||||||||||||||||
Letzte Änderung: | 16 Dez 2021 13:33 |
Nur für Mitarbeiter des Archivs: Kontrollseite des Eintrags