Generative AI for Cloud Solutions

3

Fine-Tuning – Building Domain-Specific LLM Applications

In developing ChatGPT-based applications, ensuring the model’s precision, relevance, and alignment to its intended purpose is paramount. As we navigate the intricacies of this technology, it becomes evident that a one-size-fits-all approach doesn’t suffice. Hence, customizing the model becomes necessary to adapt to certain specialized domains, such as medicine, biotechnology, legal, and others. This chapter delves deep into model customization for domain-specific applications via fine-tuning and parameter-efficient fine-tuning (PEFT). But how do we evaluate that our refinements truly hit the mark? How do we know that they align with human values? Through rigorous evaluation metrics and ...