EGI, the federation of computing and storage resource providers, invited SZTAKI’s developer Krisztián Póra to talk about the Horovod distributed deep learning framework and the reference architecture and service(s) SZTAKI based on this solution. EGI is committed to delivering advanced computing and data analytics services for research and innovation in Europe.
The Horovod reference architecture is already available on the ELKH Cloud, and the Distributed Deep Learning by Horovod service is also ready to use via the EOSC marketplace. The European Open Science Cloud is developing a multi-disciplinary environment based on FAIR principles, where researchers can publish, find and re-use data, tools and services, enabling them to enhance their scientific work.
The presentation addresses the theoretical foundations of distributed deep learning (what, why, how), the Horovod framework, the ELKH Cloud infrastructure, the Horovod reference architecture, the benchmarking activities performed on the reference architecture, and finally the EOSC service portfolio. The aim of the presented achievements is to aid researchers in the swift and efficient utilization of cloud resources, by means of providing them with a well-tested digital research environment built for the task.
The webinar consists of a presentation and a demonstration, its target audience is infrastructure providers and (potential) researchers/users of (distributed) deep learning.
The presentation is available on the EGI youtube channel.
