Customizing Models for Enhanced Performance
When general-purpose models fall short of delivering satisfactory results for your domain-specific use case, customizing FMs becomes crucial. This chapter delves into the process of customizing FMs while using techniques such as fine-tuning and continued pre-training to enhance their performance. We’ll begin by examining the rationale behind customizing the base FM and exploring the mechanics of fine-tuning. Subsequently, we will delve into data preparation techniques to ensure our data is formatted appropriately for creating a custom model using both the AWS console and APIs. We will understand various components within model customization and different customization APIs that you can call from your application.
Furthermore, we will analyze the model’s behavior and perform inference. Finally, we will conclude this chapter by discussing guidelines and best practices for customizing Bedrock models.
By the end of this chapter...
