Large Models Are Expensive to Fine-Tune on Downstream Tasks

Metadata

• Author: Sourab Mangrulkar
• Full Title: Large Models Are Expensive to Fine-Tune on Downstream Tasks
• URL: https://twitter.com/sourab_m/status/1624090029198020612

Highlights

• With PEFT, you can fit the fine-tuning of very large models within consumer GPUs with support from other techniques like CPU offloading! (View Highlight)