Back to Search
Start Over
Combinatorial Tiling for Sparse Neural Networks
- Source :
- HPEC, Proceedings 2020 IEEE High Performance Extreme Computing Conference (HPEC), 1. IEEE Computer Society Publications, STARTPAGE=1;TITLE=Proceedings 2020 IEEE High Performance Extreme Computing Conference (HPEC)
- Publication Year :
- 2020
-
Abstract
- Sparse deep neural networks (DNNs) emerged as the result of search for networks with less storage and lower computational complexity. The sparse DNN inference is the task of using such trained DNN networks to classify a batch of input data. We propose an efficient, hybrid model- and data-parallel DNN inference using hypergraph models and partitioners. We exploit tiling and weak synchronization to increase cache reuse, hide load imbalance, and hide synchronization costs. Finally, a blocking approach allows application of this new hybrid inference procedure for deep neural networks. We initially experiment using the hybrid tiled inference approach only, using the first five layers of networks from the IEEE HPEC 2019 Graph Challenge, and attain up to 2 x speedup versus a data-parallel baseline.
- Subjects :
- Hypergraph
Load modeling
Theoretical computer science
Speedup
Computational complexity theory
Artificial neural network
Computer science
Inference
010103 numerical & computational mathematics
02 engineering and technology
Synchronization
Blocking (statistics)
01 natural sciences
020204 information systems
Synchronization (computer science)
Sparse matrices
Task analysis
Taverne
0202 electrical engineering, electronic engineering, information engineering
Graph (abstract data type)
0101 mathematics
Neural networks
Subjects
Details
- Language :
- English
- Database :
- OpenAIRE
- Journal :
- HPEC, Proceedings 2020 IEEE High Performance Extreme Computing Conference (HPEC), 1. IEEE Computer Society Publications, STARTPAGE=1;TITLE=Proceedings 2020 IEEE High Performance Extreme Computing Conference (HPEC)
- Accession number :
- edsair.doi.dedup.....ff9a5986a493c4ade1c399ffcdff5062