Parameter-Efficient Finetuning - QLoRA and LoRA Methods

Learn about parameter-efficient fine-tuning techniques with a focus on (Q)LoRA in this 48-minute lecture from UofU Data Science. Explore the principles and implementation of Low-Rank Adaptation (LoRA) and its quantized variant (QLoRA), understanding how these methods enable efficient model adaptation while maintaining performance. Dive into practical applications and methodologies for optimizing large language models with minimal computational resources, accompanied by comprehensive slides that illustrate key concepts and technical implementations.