AWS Marketplace Model Packages

AWS Marketplace model packages are pre-trained machine learning models from AWS and third-party vendors that can be deployed directly into your AWS environment for inference tasks like image recognition, natural language processing, forecasting, and more.

Key Benefits:

Fast Deployment Pre-trained models can be deployed to production in minutes through Amazon SageMaker with just a few clicks. This eliminates the typical 3-6 month timeline for data gathering, model training, testing, and optimisation. Models come ready for inference with optimised configurations, allowing your team to focus on business problems rather than ML infrastructure setup.

Security Models available through AWS Marketplace are vetted and comply with AWS security standards. They integrate with AWS Identity and Access Management (IAM) for access control and support encryption both in transit and at rest. This reduces security risks compared to using unverified third-party models, and helps organisations meet compliance requirements more easily.

Cost Savings Pay-as-you-go pricing eliminates large upfront investments in ML infrastructure, data collection, and training compute resources. You avoid the significant costs of hiring specialized ML teams and the months of salary spent on model development. Flexible pricing models (hourly, monthly, or per-inference) let you scale costs with actual usage rather than maintaining expensive idle infrastructure.

You use a model package to build a deployable model in SageMaker AI. You can use the deployable model for real-time inference by using SageMaker AI hosting services. Or, they can get inferences for an entire dataset by running batch transform jobs.

Amazon SageMaker AI supports the following ways to deploy a model, depending on your use case:

• For persistent, real-time endpoints that make one prediction at a time, use SageMaker AI real-time hosting services. See Real-time inference.

• Workloads that have idle periods between traffic spikes and can tolerate cold starts, use Serverless Inference. See Deploy models with Amazon SageMaker Serverless Inference.

• Requests with large payload sizes up to 1GB, long processing times, and near real-time latency requirements, use Amazon SageMaker Asynchronous Inference. See Asynchronous inference.

• To get predictions for an entire dataset, use SageMaker AI batch transform. See Batch transform for inference with Amazon SageMaker AI.

Available Models

You can deploy the following DS1 model from AWS Marketplace:

• DS1-EN-V1: English text embedding model (512 dimensions, L2 normalized, 512 token context)

Pre-Requisites

Exact IAM Roles and Permissions Required for AWS Marketplace Model Packages:

1. AWS Marketplace Subscription Permissions

You need one of these managed policies to subscribe to marketplace products:

• AWSMarketplaceRead-only - View subscriptions only
• AWSMarketplaceManageSubscriptions - Subscribe and unsubscribe to products
• AWSMarketplaceFullAccess - Complete control of subscriptions

2. SageMaker Execution Role

The primary role needed is a SageMaker execution role with the AmazonSageMakerFullAccess managed policy attached. This gives SageMaker permission to perform operations on your behalf.

The execution role must include permissions for:

• Accessing Amazon S3 buckets (to read model artefacts and data)
• Pulling Docker images from Amazon ECR
• Creating and managing SageMaker endpoints
• Writing logs to CloudWatch

3. Minimum Required Permissions

If you don't want to grant full access with AmazonSageMakerFullAccess, create a custom IAM role with specific permissions based on which API operations you'll invoke (CreateEndpoint, CreateModel, etc.).

4. IAM PassRole Permission

The IAM user or role creating the deployment must have iam:PassRole permission to pass the execution role to SageMaker.

Process to Subscribe to DS1 in AWS Marketplace

1. Find the Product

Open the SageMaker console and choose "Marketplace model packages" then "Find marketplace model packages" to browse AWS Marketplace model packages. Alternatively, you can discover machine learning products in AWS Marketplace by selecting Amazon SageMaker under "Delivery Methods" and searching for the product name.

2. Review Product Details

Select the DS1 product to go to the product detail page, where you can learn more about the product, read usage instructions, and compare pricing.

3. Initiate Subscription

From the product detail page, click on the "Continue to Subscribe" button.

4. Accept the Offer

On the "Subscribe to this software" page, review and click on "Accept Offer" if you and your organization agree with the EULA, pricing, and support terms.

5. Configure and Get Product ARN

Once you click on "Continue to configuration" button and choose a region, you will see a Product ARN displayed. This is the model package ARN that you need to specify when creating a deployable model.

6. Deploy the Product

After subscription, you can deploy the product using:

• Amazon SageMaker console
• Jupyter notebooks
• CloudFormation templates
• AWS CLI command-line interface

AWS Marketplace Model Package Deployment

This section describes how to deploy a model package using Amazon SageMaker Studio and Jupyter Notebooks.

Amazon SageMaker Studio Classic

Amazon SageMaker Studio Classic provides a web-based, integrated development environment for machine learning and AI projects. It features a pre-configured notebook environment that automatically authenticates with your AWS account.

Alternative Options: If you already have a preferred Jupyter notebook environment (such as a local setup) with AWS authentication configured, you can skip the SageMaker Studio Classic setup and use your existing environment instead.

Getting Started: To use SageMaker Studio, refer to the official SageMaker documentation for instructions on launching SageMaker Studio Classic and starting a JupyterLab environment.

Getting Started with the Example Notebook

An example Jupyter notebook is available to help you begin working with the AWS SDK (Boto3) and Amazon SageMaker Python SDK in Python.

Setup Instructions: To access the notebook, clone the DS1 AWS repository using the following command:

git clone https://github.com/takara-ai/aws

Running the Notebook: Once cloned, you can run the notebook in either SageMaker Studio Classic or your preferred Jupyter notebook environment.

Advanced Deployment Options

While the Jupyter notebook described above is designed to help you learn the fundamentals of deploying model packages, there are additional deployment methods available that may better align with your production workflows.

Alternative Deployment Methods:

• CloudFormation - Ideal for declarative infrastructure-as-code specifications, allowing you to define and version your deployment architecture.
• SageMaker Console - Provides an interactive, user interface-driven approach for visual deployment management.
• AWS CLI - Enables programmatic deployment through command-line scripting and automation.

Choose the deployment method that best fits your team's existing processes and infrastructure management practices.

To configure and deploy a model package using one of the above methods.

(N.B. this assumes you have already subscribed, if not please follow the above defined process):

1. Go to the product listing page for the subscribed DS1 model package.

2. Click on Continue to Subscribe in the upper-right corner.

3. On the Subscribe to this software page, you should see that you are Already Subscribed. Click on the Continue to configuration button.

4. On the Configure and launch page, select your preferred deployment method (CloudFormation, SageMaker Console, or AWS CLI). The page will automatically update to display the specific configuration settings, instructions, and resources relevant to your selected method.