Bring a query-responding finely-tuned model into Amazon Bedrock as a personalized model | AWS Machine Learning Blog

Introduction to Amazon Bedrock

Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and Amazon through a single API. It provides a broad set of capabilities to build generative AI applications with an emphasis on security, privacy, and responsible AI.

Retrieval Augment Generation Technique

Retrieval Augment Generation (RAG) is a technique used to optimize the output of FMs by providing context around the questions for common generative AI use cases such as chatbots, virtual assistants, conversational search, and agent assistants.

Custom Model Import for Amazon Bedrock

The Custom Model Import feature, currently in preview, allows the importation of customized FMs created in other environments, such as Amazon SageMaker, Amazon Elastic Compute Cloud (Amazon EC2) instances, and on-premises, into Amazon Bedrock.

Fine-Tuning a Mistral Model

This section provides a step-by-step approach to fine-tuning a Mistral model using Amazon SageMaker and importing it into Amazon Bedrock using the Custom Import Model feature.

Model Architecture and Deployment

The article details the architecture patterns for importing fine-tuned FMs into Amazon Bedrock. It highlights the importance of managing model artifacts and data within the selected AWS Region.

Preparing Data and Fine-Tuning

Learn how to preprocess data and fine-tune the Mistral model using optimization techniques such as QLoRA and Parameter-Efficient Fine-Tuning.

Model Import and Testing

Discover how to import the fine-tuned Mistral model into Amazon Bedrock and test it using either the Amazon Bedrock console or the SDK.

Evaluating the Imported Model

This section explains how to evaluate the imported model using the SageMaker FMEval library, focusing on metrics such as F1 Score, Exact Match Score, and others for question answering tasks.

Conclusion

In conclusion, the article showcases the seamless process of fine-tuning and importing custom models into Amazon Bedrock for generative AI applications, providing a secure and scalable solution for various industries.