Hands-On Large Language Models

Chapter 7. Advanced Text Generation Techniques and Tools

In the previous chapter, we saw how prompt engineering can do wonders for the accuracy of your text-generation large language model (LLM). With just a few small tweaks, these LLMs are guided toward more purposeful and accurate answers. This showed how much there is to gain using techniques that do not fine-tune the LLM but instead use the LLM more efficiently, such as the relatively straightforward prompt engineering.

In this chapter, we will continue this train of thought. What can we do to further enhance the experience and output that we get from the LLM without needing to fine-tune the model itself?

Fortunately, a great deal of methods and techniques allow us to further improve what we started with in the previous chapter. These more advanced techniques lie at the foundation of numerous LLM-focused systems and are, arguably, one of the first things users implement when designing such systems.

In this chapter, we will explore several such methods and concepts for improving the quality of the generated text:

Model I/O

Loading and working with LLMs

Memory

Helping LLMs to remember

Agents

Combining complex behavior with external tools

Chains

Connecting methods and modules

These methods are all integrated with the LangChain framework that will help us easily use these advanced techniques throughout this chapter. LangChain is one of the earlier frameworks that simplify working with LLMs through useful abstractions. Newer ...