1 paper · 1 filter
Steven Eliuk, Cameron Upright, Anthony Skjellum
A new scalable parallel math library, dMath, is presented in this paper that demonstrates leading scaling when using intranode, or internode, hybrid-parallelism for deep-learning.…