A Computational-Graph Partitioning Method for Training Memory-Constrained DNNs

Autor: Qararyah, Fareed, Wahib, Mohamed, Dikbayır, Doğa, Belviranli, Mehmet Esat, Unat, Didem

Many state-of-the-art Deep Neural Networks (DNNs) have substantial memory requirements. Limited device memory becomes a bottleneck when training those models. We propose ParDNN, an automatic, generic, and non-intrusive partitioning strategy for DNNs

Externí odkaz: http://arxiv.org/abs/2008.08636

Efficient Execution of Scientific Applications on Heterogeneous Architectures

Autor: Belviranli, Mehmet Esat

Publikováno v: Belviranli, Mehmet Esat. (2016). Efficient Execution of Scientific Applications on Heterogeneous Architectures. UC Riverside: Computer Science. Retrieved from: http://www.escholarship.org/uc/item/8kn3j3pd

Today's heterogeneous architectures bring together multiple general purpose CPUs, domain specific GPUs and FPGAs to provide dramatic speedup for many applications. However, the challenge lies in utilizing these heterogeneous processors to optimize ov

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=od_______325::98118128ed8a4c90c7bc4cfb8c228f2d
http://www.escholarship.org/uc/item/8kn3j3pd