AWS Compute Week, Day 2: Deep Learning Workshop Introduction
16 May 2017 | AWSDeploy a Deep Learning Framework on amazon ECS and EC2 Spot Instances
Introduce MXnet Containers Overview of ECS and ECR Overview of AWS CloudFormation Overview of EC2 Spot Instances ECS and EC2 Spot Instances Together Hands-on, Self paced workshop
What is MX net ?open source deep learning framework Define train and deploy deep neural networks. Highly scalable, multihost, cpu/gpu
Why containers? Increase infrastructure utilization, environment isolation and fidelity, run diverse applications on shared hardware, changes are tracked, easy to deploy. Increase portability, and flexibility, speed and efficiency.
ECR is EC2 container registry that you use to store your docker images It becomes a task definition and then reference an ECR container in the task
AWS Cloudformation Great for infrastructure as Code. USE JSON or Yaml files as templates, make a service and build the stack Stack replication and infrastructure scale out
Workshop: Image classification Lab 1: Setup the workshop environment on AWS Lab 2: Build an MXNet Docker Image Lab 3: Launch MXnet with ECS Lab 4: Image classification Demo Lab 5: Wrap Image Classification in an ECS Task