Amazon Cloud Announces Deep Learning Containers
The Amazon cloud, hearing that customers wanted to simplify the complicated and time-consuming process of manually using Amazon Web Services to deploy TensorFlow workloads, announced a new managed service to do just that, and more.
Announced March 27, AWS Deep Learning Containers are designed to relieve the drudgery of building and optimizing custom machine learning environments, AWS said.
Deep learning actually extends machine learning functionality, according to AWS's Deep Learning site: "Unlike traditional machine learning, deep learning attempts to simulate the way our brains learn and process information by creating artificial 'neural networks' that can extract complicated concepts and relationships from data."
AWS said building and testing container images for deep learning projects in the cloud is complex and difficult, requiring special expertise as users must deal with different software dependencies, version compatibility issues, cloud optimization and so on.
To address such issues, the service provides Docker images pre-installed with deep learning frameworks and their included libraries, with TensorFlow and Apache MXNet supported now and PyTorch and others planned.
AWS spokesperson Jeff Barr said the service was developed in response to customer feedback who asked the cloud giant to simplify the processes they had to undertake to use Amazon EKS (managed Kubernetes) and ECS (Elastic Container Service) to deploy their TensorFlow workloads to the cloud.
"The images are pre-configured and validated so that you can focus on deep learning, setting up custom environments and workflows on Amazon ECS, Amazon Elastic Container Service for Kubernetes, and Amazon Elastic Compute Cloud (EC2) in minutes!" Barr said in a March 27 post. "You can find them in AWS Marketplace and Elastic Container Registry, and use them at no charge. The images can be used as-is, or can be customized with additional libraries or packages."
AWS said the new service is part of a broader initiative to help its users learn about deep learning and use the technology in applications. "If you know how to ingest large datasets, train existing models, build new models, and to perform inferences, you'll be well-equipped for the future!" Barr said.
David Ramel is an editor and writer for Converge360.