NORDICS20 - Assets

Deep Learning on AWS

Amazon Web Services Resources EMEA

Issue link:

Contents of this Issue


Page 49 of 50

Amazon Web Services Deep Learning on AWS Page 45 If you belong to an organization that has decided to standardize infrastructure on Kubernetes, you can use Amazon EKS as an infrastructure layer. You can add open source components such as Kubeflow to simplify build, train, and deployment of deep learning models on Amazon EKS. Amazon EKS is a persistent cluster. The cluster can scale-in and scale-out. However, it is a general-purpose compute platform that requires tuning to make it efficient and performant for deep learning jobs. It requires advanced expertise and skills to manage and operate it. Optionally, with Kubernetes and Kubeflow stack, you can use Amazon SageMaker Ground Truth for data labeling and annotation and Amazon SageMaker Neo for model optimization. Conclusion This guide provided you with a comprehensive overview of all the technology building blocks, solution and patterns used when deploying deep learning on AWS. Additionally, we also highlighted some advanced use cases that may be of interest to you. We understand that research in deep learning is advancing fast and the tools used in deep learning domain are constantly changing and evolving. The current state of options offered by the AWS landscape, ranging from fully managed to DIY, cover both the definitive and exploratory requirements of most deep learning projects. However, it is possible that you encounter a use case or a requirement that may require deeper engagement and conversations to address special requirements of your projects. For these special requirements or for clarification on content published in this guide, reach out to your respective AWS account teams. Contributors Contributors to this document include: • Vikrant Kahlir, Solutions Architect, Strategic Accounts • Christian Williams, Machine Learning Specialist, AWS Solutions Architecture • Amr Ragab, HPC Global Consultant, AWS Professional Services

Articles in this issue

Links on this page

view archives of NORDICS20 - Assets - Deep Learning on AWS