Code Monkey home page Code Monkey logo

amazon-sagemaker-from-idea-to-production-vvi's Introduction

Amazon SageMaker MLOps: from idea to production in six steps

This repository contains a sequence of simple notebooks demonstrating how to move from an ML idea to production by using Amazon SageMaker.

The notebooks make use of SageMaker processing and training jobs, and SageMaker MLOps features such as SageMaker Pipelines, SageMaker Feature Store, SageMaker Model Registry, SageMaker Experiments, and SageMaker Model Monitor.

You start with a simple notebook with basic ML code for data preprocessing, feature engineering, and model training, all local to the notebook. Each subsequent notebook builds on top of the previous and introduces one or several SageMaker MLOps features:

Each notebook also provides links to useful hands-on resources and proposes real-world ideas for additional development.

You follow along the six notebooks and develop your ML idea from an experimental notebook to a production-ready solution following the recommended MLOps practices:

Additional topics

There are also additional hands-on examples of other SageMaker features and ML topics, like A/B testing, custom processing, training and inference containers, debugging and profiling, security, multi-model and multi-container endpoints, and serial inference pipelines. Explore the notebooks in the folder additional-topics to test out these features.

Getting started

For the full version of the instructions and detailed setup of the account refer to the public AWS workshop Amazon SageMaker MLOps: from idea to production in six steps.

Prerequisites

You need an AWS account. If you don't already have an account, follow the Setting Up Your AWS Environment getting started guide for a quick overview.

AWS Instructor-led workshop

If you participating in an AWS Immersion Day or a similar instructor-led event and would like to use a provided AWS account, please follow this instructions how to claim your AWS account via Event Engine and how to start SageMaker Studio.

❗ Skip the following steps Set up Amazon SageMaker domain and Deploy CloudFormation template if you use an AWS-provisioned account.

Set up Amazon SageMaker domain

To run the notebooks you must use SageMaker Studio which requires a SageMaker domain.

Existing SageMaker domain

If you already have a SageMaker domain and would like to use it to run the workshop, follow the SageMaker Studio setup guide to attach the required AWS IAM policies to the IAM execution role used by your Studio user profile. For this workshop you must attach the following managed IAM policies to the IAM execution role of the user profile you use to run the workshop:

  • AmazonSageMakerFullAccess
  • AWSCloudFormationFullAccess
  • AWSCodePipeline_FullAccess
  • AmazonSageMakerPipelinesIntegrations

You can also create a new user profile with a dedicated IAM execution role to use for this workshop.

Provision a new SageMaker domain

If you don't have a SageMaker domain or would like to use a dedicated domain for the workshop, you must create a new domain.

❗ If you have more than one domain in your account, consider the limit of the active domains in a Region in an account.

To create a new domain, you can follow the onboarding instructions in the Developer Guide or use the provided AWS CloudFormation template that creates a SageMaker domain, a user profile, and adds the IAM roles required for executing the provided notebooks.

❗ If you create a new domain via AWS Console, make sure you attach the following policies to the IAM execution role of the user profile:

  • AmazonSageMakerFullAccess
  • AWSCloudFormationFullAccess
  • AWSCodePipeline_FullAccess
  • AmazonSageMakerPipelinesIntegrations

❗ If you use the provided CloudFormation template for domain creation, the template creates an IAM execution role with the following policies attached:

  • AmazonSageMakerFullAccess
  • AmazonS3FullAccess
  • AWSCloudFormationFullAccess
  • AWSCodePipeline_FullAccess
  • AmazonSageMakerPipelinesIntegrations

Download the sagemaker-domain.yaml CloudFormation template.

This template creates a new SageMaker domain and a user profile named studio-user-<UUID>. It also creates the required IAM execution role for the domain.

❗ This stack assumes that you already have a public VPC set up in your account. If you do not have a public VPC, see VPC with a single public subnet to learn how to create a public VPC.

❗ The template supports only us-east-1, us-west-2, and eu-central-1 Regions. Select one of those regions for deployment.

Open AWS CloudFormation console. The link opens the AWS CloudFormation console in your AWS account. Check the selected region and change it if needed.

  • Select Upload a template file and upload the downloaded CloudFormation template, click Next
  • Enter the stack name, for example sagemaker-from-idea-to-prod, click Next
  • Leave all defaults on this pane, click Next
  • Select I acknowledge that AWS CloudFormation might create IAM resources, click Submit

On the CloudFormation pane, choose Stacks. It takes about 15 minutes for the stack to be created. When the stack is created, the status of the stack changes from CREATE_IN_PROGRESS to CREATE_COMPLETE.

Start SageMaker Studio

After signing into the AWS account, follow Launch Amazon SageMaker Studio Classic instructions to open Studio.

If you deployed the provided CloudFormation template or are participating in an instructor-led event, use the section Launch Studio Classic if Studio is your default experience and select studio-user-<UUID> user profile to launch Studio. If you use your own AWS Account and Studio Classic, follow the section Launch Studio Classic if Studio Classic is your default experience.

The following instructions and screenshots show how to start Studio Classic if you are in an AWS-led workshop event.

First, launch Studio via your user profile:

SageMaker Studio UX opens in a new browser window and you're redirected to that window.

Second, select Studio Classic from Studio application pane and click Open in Studio Classic pane on the right:

If there is no Open button in the Action cell, you'll see Run action instead. Click Run and wait 2-5 minutes until you see Open in the cell for Studio Classic app.

A new browser window with Studio Classic opens and you're redirected to that window.

Start the workshop

If you're participating in an AWS-led workshop or used the provided CloudFormation template, the workshop content is cloned on the Studio EFS volume automatically, no action required from you. If you use your own domain and user profile or created a domain via AWS Console UX, follow the instructions in the section Download notebooks into your Studio environment to clone the content.

The public GitHub repository Amazon SageMaker MLOps: from idea to production in six steps contains all source code.

Open and execute a setup notebook

Navigate to the Studio Classic file browser inside the folder amazon-sagemaker-from-idea-to-production. Open 00-start-here.ipynb notebook and follow the instructions:

If you don't have the workshop folder in the file browser, follow the instructions in the section Download notebooks into your Studio environment to clone the content.

Select Data Science 3.0 image in Set up environment for "00-start-here-ipynb":

Wait until notebook kernel finishes starting.

You must run the 00-start-here.ipynb notebook to setup the environment and be able to execute other notebooks.

After executing the 00-start-here.ipynb notebook, you can move to the first notebook 01-idea-development.

Download notebooks into your Studio environment

You need to clone the source code for the workshop only if you don't have it on the Studio EFS volume.

After you open Studio Classic select File in the top menu, then New, then Terminal:

Run the following command in the terminal:

git clone https://github.com/aws-samples/amazon-sagemaker-from-idea-to-production.git

The code repository will be downloaded and saved in your home directory on the EFS in Studio.

How to use this workshop

You can do this workshop in two ways:

  • Go through the provided notebooks, execute code cells sequentially, and follow the instructions and execution flow
  • Write your own code with hands-on assignments and exercises

The following diagram shows the possible flows of the workshop:

Execution mode

Use this mode if you're not familiar with Python programming and new to Jupyter notebooks. You follow each notebook 00-..., 01-..., ..., 06-...and execute all code cells with Shift + Enter. The given instructions explain what code is doing and why. You need about two and half hours to run through all code cells in all notebooks. All notebooks and all code cells are idempotent. Make sure you run all code cells sequentially, top to bottom.

Assignment mode

Use this mode if you have experience working with Jupyter notebooks and would like to write own code to have a deeper hands-on understanding of SageMaker features and SageMaker Python SDK. Each foundational instruction notebook 00-..., 01-..., ..., 06-... in the workshop root folder has a corresponding "assignment" notebook with exercises in the assignments folder. First, go through the instructions in the root folder notebook and then complete the exercises in the corresponding assignment notebook. The notebooks are mapped as follows:

  • 00-start-here > ./assignments/00-assignment-setup
  • 01-idea-development > ./assignments/01-assignment-local-development
  • 02-sagemaker-containers > ./assignments/02-assignment-sagemaker-containers
  • 03-sagemaker-pipeline > ./assignments/03-assignment-sagemaker-pipeline
  • 04-sagemaker-projects > ./assignments/04-assignment-sagemaker-project
  • 05-deploy > ./assignments/05-assignment-deploy
  • 06-monitoring > ./assignments/06-assignment-monitoring

Clean-up

❗ You don't need to perform a clean-up if you run an AWS-instructor led workshop.

To avoid charges, you must remove all project-provisioned and generated resources from your AWS account.

First, run all steps in the provided clean-up notebook. Second, if you used the AWS Console to provision a domain for this workshop, and don't need the domain, you can delete the domain by following this instructions.

If you provisioned a domain use a CloudFormation template, you can delete the CloudFormation stack in the AWS console.

Delete EFS

❗ Delete the SageMaker EFS only if you provisioned a new SageMaker domain in your account. Do not delete your own existing EFS!

The deployment of Studio creates a new EFS in your account. This EFS is shared with all users of Studio and contains home directories for Studio users and may contain your data. When you delete the data science environment stack, the domain, user profile and Apps are also deleted. However, the EFS is not deleted and kept "as is" in your account. Additional resources are created by Studio and retained upon deletion together with the EFS:

  • EFS mounting points in each private subnet of your VPC
  • ENI for each mounting point
  • Security groups for EFS inbound and outbound traffic

❗ To delete the EFS and EFS-related resources in your AWS account created by the deployment of this solution, do the following steps after deleting the CloudFormation stack.

This is a destructive action. All data on the EFS will be deleted (SageMaker home directories). You may want to backup the EFS before deletion.

From AWS console
Got to the EFS console and delete the SageMaker EFS. You may want to backup the EFS before deletion.

To find the SageMaker EFS, click on the file system ID and then on the Tags tab. You see a tag with the Tag Key ManagedByAmazonSageMakerResource. Its Tag Value contains the SageMaker domain ID: efs-tags

❗ If you have multiple EFS, double check that you selected the correct domain ID.

Click on the Delete button to delete this EFS.

If you provisioned a new VPC for the domain, go to the VPC console and delete the provisioned VPC.

Dataset

This example uses the direct marketing dataset from UCI's ML Repository:

[Moro et al., 2014] S. Moro, P. Cortez and P. Rita. A Data-Driven Approach to Predict the Success of Bank Telemarketing. Decision Support Systems, Elsevier, 62:22-31, June 2014

Resources

The following list presents some useful hands-on resources to help you to get started with ML development on Amazon SageMaker.


QR code for this repository

Use the following QR code to link this repository.

https://bit.ly/3KkhzYW

QR code for the workshop

Use the following QR code to link the public AWS workshop.

https://bit.ly/3zjk07S


Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. SPDX-License-Identifier: MIT-0

amazon-sagemaker-from-idea-to-production-vvi's People

Contributors

pdifranc avatar yevgeniyilyin avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.