Introduction
Unit 1 of 7 · Module 5: Optimize generative AI model performance with Microsoft Foundry
Read this lesson
A base model on its own
Welcome to Module 5. A language model straight out of the box is impressive, but on its own it rarely does everything your app needs. Its answers' quality, accuracy, and consistency all hinge on how you set it up, and what you add to it. Say you build software for a travel agency, and you're making a chat app that answers travelers' questions. Out of the box, the model's answers are fine, just fine. But your team wants more than fine. Every reply should sound like the agency, warm and upbeat. Every fact about the agency's hotels has to be right, down to whether breakfast is included. And every answer should be laid out the same way, every single time. So how do you turn a fine model into one that meets that bar?
A range of strategies
You've got a few strategies, and they work well together. At one end are quick, cheap tweaks, like prompt engineering; further along is retrieval augmented generation, which needs a search index, and furthest of all is fine-tuning, which means training the model itself. In this module, you'll try each one, and learn when to use it, on its own, or in combination with the others.
Facts from Microsoft's documentation: Prompt engineering techniques; Retrieval augmented generation (RAG) and indexes in Microsoft Foundry; Customize a model with fine-tuning (CC BY 4.0), rewritten in Loop's own words and examples. Narration, animation, and quiz by Loop. Microsoft, Azure, and Microsoft Foundry are trademarks of Microsoft. Loop is independent and not affiliated with or endorsed by Microsoft.