Simplify Generative AI with Amazon Bedrock and Custom Model Import

Introduction

Building and deploying generative AI applications requires flexibility, scalability, and cost efficiency. Amazon Bedrock simplifies this process with its fully managed, on-demand API.

With its capabilities, Amazon Bedrock enables organizations to create AI applications while ensuring security, privacy, and responsible AI practices.

Previously, using custom fine-tuned models with Amazon Bedrock involved managing your inference infrastructure in Amazon SageMaker or relying on costly provisioned throughput. Now, with Amazon Bedrock Custom Model Import, you can integrate models trained or fine-tuned in SageMaker, including those created using Amazon SageMaker JumpStart, into Amazon Bedrock. This innovation allows you to harness the power of these models on demand through Amazon Bedrock’s fully managed API, streamlining the deployment process and reducing infrastructure complexity.

Solution Overview

Amazon Bedrock supports importing custom models from the following architectures:

• Mistral
• Flan
• Meta Llama 2, Llama 3, Llama3.1, Llama3.2, and Llama 3.3

For this, we use a Hugging Face Flan-T5 Base model.

Prerequisites

Ensure you have an AWS account with access to Amazon Bedrock and Amazon SageMaker Studio before you start.

See Launch Amazon SageMaker Studio for details on how to set up an instance of Amazon SageMaker Studio if you don’t already have one.

Steps to Train a model in Amazon SageMaker JumpStart

1. Open the AWS Management Console and go to Amazon SageMaker Studio.
2. In Amazon SageMaker Studio, select JumpStartin the navigation pane.
3. Search for and choose the Hugging Face Flan-T5 Base.

4. Select Train to begin fine-tuning the model on your training data.

Use the default options when creating the training job. The defaults add suggested settings to the training job.

5. A prepopulated example dataset is used in this post’s example. Ensure your data satisfies the format criteria before entering its location in the Data section.

6. Configure the security settings such as AWS Identity and Access Management (IAM) role, virtual private cloud (VPC), and encryption.

7. Note the value for Output artifact location (S3 URI) to use later.

8. To begin training, submit the job.

You may monitor your work by choosing Training from the Jobs dropdown menu. The training task is considered completed when its status is displayed as Completed. Training takes roughly ten minutes with the default settings.

Steps to Import the model into Amazon Bedrock

1. Choose Imported models under Foundation models in the navigation pane on the Amazon Bedrock console.
2. Select Import model.
3. Enter a memorable name for your model in the Model name
4. Under Model import settings, select Amazon SageMaker model and select the radio button next to your model.

5. Under Service access, select Create, use a new service role, and enter a name for the role.

6. Select Import model.

7. The model import will be completed in about 15 minutes.

8. Under Playgrounds in the navigation pane, select

9. Choose Select model.

10. For Category, choose Imported models.

11. For Model, choose flan-t5-fine-tuned.

12. For Throughput, choose On-demand.

13. Choose Apply.

Your personalized model is now available for interaction. The following screenshot summarises Amazon Bedrock using our sample custom model.

Clean up

1. You should delete your Amazon SageMaker domain if you want to stop using it.
2. Delete the Amazon Simple Storage Service (Amazon S3) bucket containing your model artifacts if you no longer wish to keep them.
3. Select your model on the Imported Models page in the Amazon Bedrock console, then click the options menu (three dots) and choose Delete to remove your imported model.

Conclusion

In conclusion, the Custom Model Import feature in Amazon Bedrock revolutionizes how you deploy and utilize custom-trained or fine-tuned models, offering a seamless, cost-effective solution for on-demand inference. By combining the training capabilities of Amazon SageMaker with the fully managed, scalable infrastructure of Amazon Bedrock, businesses can focus on creating innovative applications without the complexities of managing machine learning infrastructure.

Whether leveraging the intuitive Amazon SageMaker Studio console or the flexibility of Amazon SageMaker notebooks, you can train, fine-tune, and import your models effortlessly into Amazon Bedrock. This integration empowers you to deliver tailored, high-value experiences to your customers while benefiting from the scalability, efficiency, and simplicity of a fully managed API.

As the field of large language models continues to advance, the ability to integrate custom models into your workflows becomes indispensable. Amazon Bedrock Custom Model Import ensures that you can maximize the potential of your specialized models, transforming insights into impactful, real-world solutions.

Drop a query if you have any questions regarding Amazon Bedrock Custom Model and we will get back to you quickly.

About CloudThat

CloudThat is a leading provider of Cloud Training and Consulting services with a global presence in India, the USA, Asia, Europe, and Africa. Specializing in AWS, Microsoft Azure, GCP, VMware, Databricks, and more, the company serves mid-market and enterprise clients, offering comprehensive expertise in Cloud Migration, Data Platforms, DevOps, IoT, AI/ML, and more.

CloudThat is the first Indian Company to win the prestigious Microsoft Partner 2024 Award and is recognized as a top-tier partner with AWS and Microsoft, including the prestigious ‘Think Big’ partner award from AWS and the Microsoft Superstars FY 2023 award in Asia & India. Having trained 650k+ professionals in 500+ cloud certifications and completed 300+ consulting projects globally, CloudThat is an official AWS Advanced Consulting Partner, Microsoft Gold Partner, AWS Training Partner, AWS Migration Partner, AWS Data and Analytics Partner, AWS DevOps Competency Partner, AWS GenAI Competency Partner, Amazon QuickSight Service Delivery Partner, Amazon EKS Service Delivery Partner,  AWS Microsoft Workload Partners, Amazon EC2 Service Delivery Partner, Amazon ECS Service Delivery Partner, AWS Glue Service Delivery Partner, Amazon Redshift Service Delivery Partner, AWS Control Tower Service Delivery Partner, AWS WAF Service Delivery Partner, Amazon CloudFront, Amazon OpenSearch, AWS DMS and many more.