NORDICS20 - Assets

Deep Learning on AWS

Amazon Web Services Resources EMEA

Issue link:

Contents of this Issue


Page 7 of 50

Amazon Web Services Deep Learning on AWS Page 3 Table 1: Deep learning landscape diagram descriptions Label Description 1 Six steps required to execute deep learning projects. The six steps involved are discussed in Deep Learning Process for Build, Train, and Deploy. 2 The different layers required to support a deep learning environment for build, train, and deploy tasks. The layers in the figure extend from infrastructure to tools required for deep learning projects. 3 The do-it-yourself (DIY) option where the customer is responsible for building and managing components and features required for deep learning using AWS compute, storage, and network technology building blocks. 4 Amazon SageMaker is a fully-managed service that covers the entire deep learning workflow to label and prepare your data, choose an algorithm, train the model, tune and optimize it for deployment, make predictions, and take action. Your models get to production faster with much less effort and lower cost. 5 A measure of infrastructure experience required to set up the deep learning environment in the context of ease-of-use and the shared responsibility model between customer and AWS. Fully managed is easy to use because AWS manages the major part of the stack. The do-it-yourself (DIY) option is more challenging because customers manage most of the stack Note: Between the fully managed (4) and do-it-yourself (DIY) (3) options, there is a partially managed approach where you use a fully managed container service and a self-managed deep learning workflow service like Kubeflow. This partially managed approach is relevant for organizations that have decided to standardize their infrastructure on top of Kubernetes. For more details, see DIY Partially Managed Solution: Use Kubernetes with Kubeflow on AWS. Using this Guide This guide is organized into seven sections, as described below. 1. Deep Learning Process for Build, Train, and Deploy 2. Challenges with Deep Learning Projects 3. Highly Optimized AWS Technology Building Blocks for Deep Learning 4. Code, Data, and Model Versioning 5. Automation of Deep Learning Process for Retrain and Redeploy

Articles in this issue

Links on this page

view archives of NORDICS20 - Assets - Deep Learning on AWS