Skip to Main content Skip to Navigation

Combinatorial Tiling for Sparse Neural Networks

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 synchronisation 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 2x speedup versus a data-parallel baseline.
Complete list of metadatas
Contributor : Bora Uçar <>
Submitted on : Monday, August 3, 2020 - 3:02:43 PM
Last modification on : Monday, September 28, 2020 - 9:12:02 AM
Long-term archiving on: : Monday, November 30, 2020 - 1:41:13 PM


  • HAL Id : hal-02910997, version 1


Filip Pawłowski, Rob Bisseling, Bora Uçar, Albert-Jan Yzelman. Combinatorial Tiling for Sparse Neural Networks. [Research Report] RR-9357, Inria - Research Centre Grenoble – Rhône-Alpes. 2020. ⟨hal-02910997v1⟩



Record views


Files downloads