Amazon Web Services Inc. (AWS), has unveiled a new service that simplifies the process of deploying deep-learning workloads to the cloud.
AWS Deep Learning Containers are reportedly able to help developers and users set up custom environments and workflows such as machine learning environments to the cloud.
Containers are Docker images that have deep learning frameworks and the associated libraries pre-installed. TensorFlow, Apache MXNet, PyTorch, and others are supported now.
Deep learning, which uses technology such as machine-learning, goes a bit further. According to AWS’s Deep Learning website: “Unlike traditional machine intelligence, deep learning attempts simulate how our brains learn and process information. Deep learning creates artificial neural networks that can extract complex concepts and relationships from data.
Jeff Barr, an AWS spokesperson, said that the service was created in response to customer feedback. He asked Amazon EKS (managed Kubernetes), and ECS (Elastic Container Service), to help them deploy their TensorFlow workloads into the cloud.
AWS stated that testing container images for deep-learning projects in the cloud can be difficult and time-consuming. Users must deal with different software dependencies and version compatibility issues.
“The images have been pre-configured and validated so you can focus on deeplearning, setting up custom environments and workflows for Amazon ECS, Amazon Elastic Container Service (Kubernetes), and Amazon Elastic Compute Cloud(EC2) in minutes!” Barr stated this in a March 27 blog post. You can find them in AWS Marketplace or Elastic Container Registry and you can use them for free. Images can be used as is or customized with additional libraries and packages.
