aws-samples/amazon-sagemaker-ab-testing-pipeline

Amazon SageMaker MLOps deployment pipeline for A/B Testing of machine learning models.

Amazon SageMaker A/B Testing Pipeline

This sample demonstrates how to setup an Amazon SageMaker MLOps deployment pipeline for A/B Testing of machine learning models.

The following are the high-level steps to deploy this solution:

1. Publish the SageMaker MLOps Project template in the AWS Service Catalog
2. Deploy the Amazon API Gateway and Testing Infrastructure
3. Create a new Project in Amazon SageMaker Studio

Once complete, you can Train and Deploy machine learning models for A/B Testing in the sample notebook provided.

Get Started

To get started first, clone this repository.

git clone https://github.com/aws-samples/amazon-sagemaker-ab-testing-pipeline.git
cd amazon-sagemaker-ab-testing-pipeline

Prerequisites

This project uses the AWS Cloud Development Kit CDK. To get started with AWS CDK you need Node.js 10.13.0 or later.

Install the AWS CDK

Install the AWS CDK Toolkit globally using the following Node Package Manager command.

npm install -g aws-cdk

Run the following command to verify correct installation and print the version number of the AWS CDK.

cdk --version

Setup Python Environment for CDK

This project uses AWS CDK with python bindings to deploy resources to your AWS account.

The cdk.json file tells the CDK Toolkit how to execute your app.

This project is set up like a standard Python project. The initialization process also creates a virtualenv within this project, stored under the .venv directory. To create the virtualenv it assumes that there is a python3 (or python for Windows) executable in your path with access to the venv package. If for any reason the automatic creation of the virtualenv fails, you can create the virtualenv manually.

To manually create a virtualenv on MacOS and Linux:

python3 -m venv .venv

After the init process completes and the virtualenv is created, you can use the following step to activate your virtualenv.

source .venv/bin/activate

If you are a Windows platform, you would activate the virtualenv like this:

.venv\Scripts\activate.bat

Once the virtualenv is activated, you can install the required dependencies.

pip install -r requirements.txt

Install Python Libraries for Lambda Layer

In order to support AWS X-RAY as part of our python function we require additional python libraries.

Run the following command to pip install the AWS X-Ray SDK for Python into the layers folder.

sh install_layers.sh

This will enabling sample request to visualize the access patterns and drill into any specific errors.

Add Permissions for CDK

AWS CDK requires permissions create AWS CloudFormation Stacks and the associated resources for your current execution role. If you have cloned this notebook into SageMaker Studio, you will need to add an inline policy to your SageMaker Studio execution role. You can find your user's role by browsing to the Studio dashboard.

Browse to the IAM section in the console, and find this role.

Then, click the Add inline policy link, switch to to the JSON tab, and paste the following inline policy:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "apigateway:*"
            ],
            "Resource": "arn:aws:apigateway:*::/*"
        },
        {
            "Action": [
                "dynamodb:*"
            ],
            "Effect": "Allow",
            "Resource": "arn:aws:dynamodb:*:*:table/ab-testing-*"
        },
        {
            "Action": [
                "lambda:*"
            ],
            "Effect": "Allow",
            "Resource": [
              "arn:aws:lambda:*:*:function:ab-testing-api-*",
              "arn:aws:lambda:*:*:layer:*"
            ]
        },
        {
            "Action": [
                "firehose:*"
            ],
            "Effect": "Allow",
            "Resource": "arn:aws:firehose:*:*:deliverystream/ab-testing-*"
        },
        {
            "Action": [
                "s3:*"
            ],
            "Effect": "Allow",
            "Resource": [
                "arn:aws:s3:::cdktoolkit-*",
                "arn:aws:s3:::ab-testing-api-*"
            ]
        },
        {
            "Action": [
                "cloudformation:*",
                "servicecatalog:*",
                "events:*"
            ],
            "Effect": "Allow",
            "Resource": "*"
        },
        {
            "Effect": "Allow",
            "Action": [
                "logs:*"
            ],
            "Resource": "arn:aws:logs:**:*:log-group:ab-testing-api-*"
        },
        {
            "Effect": "Allow",
            "Action": [
                "iam:CreateRole",
                "iam:DeleteRole"
            ],
            "Resource": "arn:aws:iam::*:role/ab-testing-api-*"
        },
        {
            "Effect": "Allow",
            "Action": [
                "iam:GetRole",
                "iam:PassRole",
                "iam:GetRolePolicy",
                "iam:AttachRolePolicy",
                "iam:PutRolePolicy",
                "iam:DetachRolePolicy",
                "iam:DeleteRolePolicy"
            ],
            "Resource": [
              "arn:aws:iam::*:role/ab-testing-api-*",
              "arn:aws:iam::*:role/service-role/AmazonSageMaker*"
            ]
        }
    ]
}

Click Review policy and provide the name CDK-DeployPolicy then click Create policy

Bootstrap the CDK

If this is the first time you have run the CDK, you may need to Bootstrap your account. If you have multiple deployment targets see also Specifying up your environment in the CDK documentation.

cdk bootstrap

You should now be able to list the stacks by running:

cdk list

Which will return the following stacks:

• ab-testing-api
• ab-testing-pipeline
• ab-testing-service-catalog

Publish the API and AWS Service Catalog template

In this section you will publish the AWS Service Catalog template and Deploy the API and Testing infrastructure.

Publish the SageMaker MLOps Project template

In this step you will create a Portfolio and Product to provision a custom SageMaker MLOps Project template in the AWS Service Catalog and configure it so you can launch the project from within your SageMaker Studio domain. See more information on customizing the template, or import the template manually into the AWS Service Catalog.

Resources include:

• AWS CodeCommit seeded with the source from the deployment_pipeline.
• AWS CodeBuild to produce AWS CloudFormation for deploying the Amazon SageMaker Endpoint.
• Amazon CloudWatch Event to trigger the AWS CodePipeline for endpoint deployment.

Run the following command to deploy the MLOps project template, passing the required ExecutionRoleArn parameter. You can copy this from your SageMaker Studio dashboard as show above.

export EXECUTION_ROLE_ARN=<<sagemaker-studio-execution-role>>
cdk deploy ab-testing-service-catalog \
    --parameters ExecutionRoleArn=$EXECUTION_ROLE_ARN \
    --parameters PortfolioName="SageMaker Organization Templates" \
    --parameters PortfolioOwner="administrator" \
    --parameters ProductVersion=1.0

This stack will output the CodeCommitSeedBucket and CodeCommitSeedKey which you will need when creating the Amazon SageMaker Studio project.

NOTE: If you are seeing errors running the above command ensure you have Enabled SageMaker project templates for Studio users to grant access to these resources in Amazon S3.

Deploy the API and Testing infrastructure

In this step you will deploy an Amazon API Gateway and supporting resources to enable dynamic A/B Testing of any Amazon SageMaker endpoint that has multiple production variants.

Resources include:

• Amazon API Gateway and AWS Lambda functions for invocation.
• Amazon DynamoDB table for user variant assignment.
• Amazon DynamoDB table for variant metrics.
• Amazon Kinesis Firehose, Amazon S3 Bucket and AWS Lambda for processing events.
• Amazon CloudWatch Event and AWS Lambda to register in service Amazon SageMaker Endpoints.

Run the following command to deploy the API and testing infrastructure with optional configuration.

cdk deploy ab-testing-api

This stack will output the ApiEndpoint which you will provide to the A/B Testing sample notebook.

You’re done! Now it’s time to create a project using this template.

Creating a new Project in Amazon SageMaker Studio

Once your MLOps project template is registered in AWS Service Catalog you can create a project using your new template.

1. Switch back to the Launcher
2. Click New Project from the ML tasks and components section.

On the Create project page, SageMaker templates is chosen by default. This option lists the built-in templates. However, you want to use the template you published for the A/B Testing Deployment Pipeline.

6. Choose Organization templates.
7. Choose A/B Testing Deployment Pipeline.
8. Choose Select project template.

9. In the Project details section, for Name, enter ab-testing-pipeline.

• The project name must have 32 characters or fewer.

10. In the Project template parameters

• For StageName, enter dev
• For CodeCommitSeedBucket, enter the CodeCommitSeedBucket output from the ab-testing-service-catalog stack
• For CodeCommitSeedKey, enter the CodeCommitSeedKey output from the ab-testing-service-catalog stack

11. Choose Create project.

NOTE: If you have recently updated your AWS Service Catalog Project, you may need to refresh SageMaker Studio to ensure it picks up the latest version of your template.

Train and Deploy machine learning models for A/B Testing

In the following sections, you will learn how to Train, Deploy and Simulate a test against our A/B Testing Pipeline.

Training a Model

Now that your project is ready, it’s time to train, register and approve a model.

1. Download the Sample Notebook to use for this walk-through.
2. Choose the Upload file button
3. Choose the Jupyter notebook you downloaded and upload it.
4. Choose the notebook to open a new tab.

This notebook will step you through the process of

1. Download a dataset
2. Create and Run an Amazon SageMaker Pipeline
3. Approve the model.
4. Create a Amazon SageMaker Tuning Job.
5. Select the best model, register and approve the second model.

Deploying the Multi-Variant Pipeline.

Once the second model has been approved, the MLOps deployment pipeline will run.

See the Deployment Pipeline for more information on the stages to run.

Running an A/B Testing simulation

With the Deployment Pipeline complete, you will be able to continue with the next stage:

1. Test the multi-variant endpoint
2. Evaluate the accuracy of the models, and visualize the confusion matrix and ROC Curves
3. Test the API by simulating a series of invocation, and recording reward conversion.
4. Plot the cumulative reward, and reward rate.
5. Plot the beta distributions of the course of the test.
6. Calculate the statistical significance of the test.

Running Costs

This section outlines cost considerations for running the A/B Testing Pipeline. Completing the pipeline will deploy an endpoint with 2 production variants which will cost less than $6 per day. Further cost breakdowns are below.

• CodeBuild – Charges per minute used. First 100 minutes each month come at no charge. For information on pricing beyond the first 100 minutes, see AWS CodeBuild Pricing.
• CodeCommit – $1/month if you didn't opt to use your own GitHub repository.
• CodePipeline – CodePipeline costs $1 per active pipeline* per month. Pipelines are free for the first 30 days after creation. More can be found at AWS CodePipeline Pricing.
• SageMaker – Prices vary based on EC2 instance usage for the Notebook Instances, Model Hosting, Model Training and Model Monitoring; each charged per hour of use. For more information, see Amazon SageMaker Pricing.
  • The ten ml.c5.4xlarge training jobs run for approx 4 minutes at $0.81 an hour, and cost less than $1.
  • The two ml.t2.large instances for production hosting endpoint costs 2 x $0.111 per hour, or $5.33 per day.
• S3 – Low cost, prices will vary depending on the size of the models/artifacts stored. The first 50 TB each month will cost only $0.023 per GB stored. For more information, see Amazon S3 Pricing.
• API Gateway - Low cost, $1.29 for first 300 million requests. For more info see Amazon API Gateway pricing
• Lambda - Low cost, $0.20 per 1 million request see AWS Lambda Pricing.

Cleaning Up

Once you have cleaned up the SageMaker Endpoints and Project as described in the Sample Notebook, complete the clean up by deleting the Service Catalog and API resources with the AWS CDK:

1. Delete the Service Catalog Portfolio and Project Template

cdk destroy ab-testing-service-catalog

2. Delete the API and testing infrastructure

Before destroying the API stack, is is recommend you empty and delete the S3 Bucket that contains the S3 logs persisted by the Kinesis Firehose.

cdk destroy ab-testing-api

Want to know more?

The FAQ page has some answers to questions on the design principals of this sample.

See also the OPERATIONS page for information on configuring experiments, and the API interface.

Security

See CONTRIBUTING for more information.

License

This library is licensed under the MIT-0 License. See the LICENSE file.