NORDICS20 - Assets

Deep Learning on AWS

Amazon Web Services Resources EMEA

Issue link: https://emea-resources.awscloud.com/i/1242450

Contents of this Issue

Navigation

Page 48 of 50

Amazon Web Services Deep Learning on AWS Page 44 AWS Guidance In this guide, we described the deep learning stack and the deep learning process. We discussed AWS services that can be used to address the deep and broad needs of deep learning engineers and scientists. We also described design patterns that deep learning engineers and scientists can leverage to adopt and scale deep learning in their organization. We discussed features that are available out of the box and are ready to use in Amazon SageMaker, a fully managed service for machine learning. We also discussed cases where you would want to build deep learning environments on your own using other AWS services such as Amazon EC2 and Amazon EKS. With AWS, you get the flexibility to choose the approach that works best for you. Below are some of the popular scenarios in which customers are leveraging different options offered by the AWS AI/ML stack. If you are a start-up, you want to spend less on buying high-performance GPU compute and managing the infrastructure. Most likely, you have a small team of developers and data scientists and your focus would be to roll out new deep learning capabilities with small teams. Amazon SageMaker is an ideal choice for you. If you are working as a research scientist in an organization developing a new product feature, using deep learning capabilities, you may need more autonomy, more isolation, and more control. It is possible your organization may not have a directive or a policy to use a standard platform for deep learning. You can continue to use Amazon EC2 with AWS DL AMI as your deep learning desktop, but you may also want to consider Amazon SageMaker for training with automatic hyperparameter tuning to scale experiments. If you are a product team working to keep your deep learning model performing well under changing customer preferences, you can implement an end-to-end automated deep learning pipeline to retrain and deploy your models using AWS Step Functions and Amazon SageMaker. If you are the technology leader in your organization who has been tasked to accelerate deep learning adoption in the organization, you can use Amazon SageMaker as a fully managed service for building, training, and deploying deep learning models in your organization. Amazon SageMaker allows you to achieve more with smaller deep learning teams. The fully managed service helps keep operations lean, eliminates compute infrastructure waste, improves productivity of deep learning engineers and scientists, allows cost efficient experimentation, and a shorter time to market.

Articles in this issue

Links on this page

view archives of NORDICS20 - Assets - Deep Learning on AWS