Discover ways to automate immediate engineering and unlock vital efficiency enhancements in your LLM workload
Automated Immediate Engineering (APE) is a method to automate the method of producing and refining prompts for a Giant Language Mannequin (LLM) to enhance the mannequin’s efficiency on a selected job. It makes use of the thought of immediate engineering which entails manually crafting and testing numerous prompts and automates all the course of. As we are going to see it’s very much like automated hyperparameter optimisation in conventional supervised machine studying.
On this tutorial we are going to dive deep into APE: we are going to first have a look at the way it works in precept, a few of the methods that can be utilized to generate prompts, and different associated strategies corresponding to exemplar choice. Then we are going to transition into the hands-on part and write an APE program from scratch, i.e. we received’t use any libraries like DSPy that may do it for us. By doing that we are going to get a a lot better understanding of how the ideas of APE work and are a lot better geared up to leverage the frameworks that may supply this performance out of the field.