Rearchitecting LLMs: Structural techniques for efficient models

By default, general purpose LLMs are not optimized for specific domains and business goals. Using techniques like specialized fine-tuning, pruning unnecessary neural components, and knowledge distillation, you can rearchitect your models to cost less, run faster, and deliver more accurate results.

This book turns research from the latest AI papers into production-ready practices for domain-specific model optimization. As you work through this practical book, you’ll perform hands-on surgery on popular open-source models like Llama-3, Gemma, and Qwen to create cost-effective local small language models (SLMs). Along the way, you’ll learn how to combine behavioral analysis with structural modifications, identifying and removing parts that don’t contribute to your model’s goals, and even use “fair pruning” to reduce model bias at the neuron level.

What’s inside

• Universal techniques for customizing model architecture
• End-to-end pipelines for model rearchitecting
• Improve bias and explainability with model “cleanup”
• Replacing external LLMs with local SLMs

About the reader

For practicing AI, ML, and data engineers who know Python.

About the author

Pere Martra is an Applied AI Engineer, the creator of the Optipfair model efficiency library, an international AI speaker, and maintainer of widely-used LLM courses and popular open-source tools. He is the author of Large Language Models Projects.