03. AWS SageMaker

A. What is AWS Sagemaker?

AWS (or Amazon) SageMaker is a fully managed service that provides the ability to build, train, tune, deploy, and manage large-scale machine learning (ML) models quickly. Sagemaker provides tools to make each of the following steps simpler:

  1. Explore and process data
  • Retrieve
  • Clean and explore
  • Prepare and transform
  1. Modeling
  • Develop and train the model
  • Validate and evaluate the model
  1. Deployment
  • Deploy to production
  • Monitor, and update model & data

The Amazon Sagemaker provides the following tools:

  • Ground Truth - To label the jobs, datasets, and workforces
  • Notebook - To create Jupyter notebook instances, configure the lifecycle of the notebooks, and attache Git repositories
  • Training - To choose an ML algorithm, define the training jobs, and tune the hyperparameter
  • Inference - To compile and configure the trained models, and endpoints for deployments

The snapshot of the Sagemaker Dashboard below shows the tools mentioned above.

IMPORTANT NOTICE: This is the current AWS UI as of April 6th, 2020. The AWS UI is subject to change on a regular basis. We advise students to refer to AWS documentation for the above process.

A.1. Why is SageMaker a "fully managed" service?

SageMaker helps to reduce the complexity of building, training and deploying your ML models by offering all these steps on a single platform.
SageMaker supports building the ML models with modularity , which means you can reuse a model that you have already built earlier in other projects.

A.2. SageMaker Instances - Important to Read

SageMaker instances are the dedicated VMs that are optimized to fit different machine learning (ML) use cases. The supported instance types, names, and pricing in SageMaker are different than that of EC2. Refer the following links to have better insight:

A.3. Supported Instance Types and Availability Zones

Amazon SageMaker offers a variety of instance types. Interestingly, the type of SageMaker instances that are supported varies with AWS Regions and Availability Zones .

A.4. Instances Required for Deep Learning

The table below describes the three types of SageMaker instances that you would use in this course:

SageMaker Instance vCPU GPU Mem (GiB) GPU Mem (GiB) Network Performance Usage Default Quota (Limit)
ml.t2.medium 2 - 4 - Low to Moderate Run notebooks 0 - 20
ml.m4.xlarge 4 - 16 - High
  • Train and batch transform XGBOOST models;
  • Deploy all models preceding the first project
0 - 20
ml.p2.xlarge 4 1xK80 61 12 High Train and batch transform GPU accelerated Pytorch models for the first project 0 -1

In this course, the ml.m4.xlarge is needed at an early stage, while ml.p2.xlarge is needed only when working on the for the first project: Deploying a Sentiment Analysis Model.

Note

Sagemaker quotas, also referred to as limits, are very tricky. Every AWS user does not get the default quotas for SageMaker instances, which is why the last column shows a range, e.g., 0 - 20. The Default Quota depends on the instance type, the task you want to run (see table above), and also the region in which the Sagemaker service is requested. Refer this document having a caveat that new accounts may not always get the default limits.

B. Shut Down SageMaker Instances, if not in use

Note: We recommend you shut down every resource (e.g., SageMaker instances, or any other hosted service) on the AWS cloud immediately after the usage; otherwise, you will be billed even if the resources are not in actual use.

Even if you are in the middle of the project and need to step away, PLEASE SHUT DOWN YOUR SAGEMAKER INSTANCE . You can re-instantiate later.

AWS Sagemaker FAQs

Next Concept