Our next meeting we will explore the fine-tuning of open-source Large Language Models (LLMs). In an example project, we will create a model that summarizes medical notes, using a 8B Mistral model. This session will highlight the benefits of fine-tuning small LLMs and their potential to match the performance of larger models like ChatGPT in narrow tasks, while maintaining data security and enhancing compute efficiency. We will also discuss other expected use cases of smaller LLMs on edge devices.
Agenda:
1. Introduction
- Welcome and opening remarks.
2. Overview of Fine-Tuning Small LLMs
- Discussion of the project focused on fine-tuning a small open-source LLM to extract key information from medical notes
- Key advantages of fine-tuning smaller models
- Maintaining data integrity in closed systems
- Improved compute efficiency compared to larger models
3. Technical Insights
- Comparison between the fine-tuned small LLM and larger models like ChatGPT on narrow tasks
- Examination of how models 1% the size of ChatGPT can perform similarly on specific tasks with targeted fine-tuning
4. Use Cases and Practical Applications
- Exploring real-world applications of fine-tuned models in healthcare and other industries
- Benefits of using open-source models in closed environments