DOE OSTI · code-33387
DiHydrogen
Abstract
DiHydrogen is the second version of the Hydrogen fork of the well-known distributed linear algebra library, Elemental. DiHydrogen is a GPU-accelerated distributed multilinear algebra interface with a particular emphasis on the needs of the scalable distributed deep learning training and inference. DiHydrogen is part of the Livermore Big Artificial Neural Network (LBANN) software stack.
Keep this discovery
Explore connections, maps & timelines
Maruyama, Naoya, Essen, BrianVan, Dryden, NikoliJ, Benson, ThomasR, Moon, TimothyY, Oyama, Yosuke. 2020-12-16. DiHydrogen. https://doi.org/10.11578/dc.20200106.1
Cite the original work for its findings. Save a collection to share your selection of sources.