Overview
Learn about Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning approach for large language models, in this 11-minute technical video. Explore the fundamentals of LoRA from rank decomposition to its practical implementation in transformer models. Discover why LoRA has become a popular choice for budget-friendly transformer model fine-tuning, understand its training and inference processes, and learn how to select appropriate rank parameters. Gain insights from the original research paper and access practical implementations through various frameworks like HuggingFace's PEFT library. Master the technical concepts behind this efficient adaptation technique that's revolutionizing the way we fine-tune large language models.
Syllabus
- Intro
- Adapters
- Twitter https://twitter.com/ai_bites
- What is LoRA
- Rank Decomposition
- Motivation Paper
- LoRA Training
- LoRA Inference
- LoRA in Transformers
- Choosing the rank
- Implementations
Taught by
AI Bites