Overview
Explore advanced fine-tuning optimization techniques for large language models in this comprehensive video tutorial. Delve into the intricacies of LoRA (Low-Rank Adaptation) and its improvements, including DoRA (Double-Rank Adaptation), NEFT (Noisy Embeddings for Fine-Tuning), LoRA+, and Unsloth. Learn how these methods work, their advantages, and practical implementations through detailed explanations and notebook walk-throughs. Compare the effectiveness of each technique and gain insights on choosing the best approach for your fine-tuning needs. Access provided resources, including GitHub repositories, slides, and research papers, to further enhance your understanding and application of these cutting-edge optimization strategies.
Syllabus
Improving on LoRA
Video Overview
How does LoRA work?
Understanding DoRA
NEFT - Adding Noise to Embeddings
LoRA Plus
Unsloth for fine-tuning speedups
Comparing LoRA+, Unsloth, DoRA, NEFT
Notebook Setup and LoRA
DoRA Notebook Walk-through
NEFT Notebook Example
LoRA Plus
Unsloth
Final Recommendation
Taught by
Trelis Research