Fine-tuning Large Models on Local Hardware Using PEFT and Quantization

Explore the world of fine-tuning large neural networks like Large Language Models (LLMs) on modest hardware in this 28-minute EuroPython 2024 conference talk. Discover how Parameter-Efficient Fine-Tuning (PEFT) and quantization techniques have made it possible to train big models without excessive hardware requirements. Learn about the challenges associated with fine-tuning large models, the proposed solutions and their mechanisms, and gain practical insights into applying the PEFT library. Understand how the PEFT library and the Hugging Face ecosystem have democratized these advanced techniques, making them accessible to a wider audience of developers and researchers.