What is LoRA?

Fine-Tuning

LoRA (Low-Rank Adaptation) — A parameter-efficient fine-tuning technique that freezes the original model weights and trains small rank-decomposition matrices. Reduces trainable parameters by 10,000x while maintaining quality.

FAQ

What is LoRA?

A technique to fine-tune LLMs cheaply by only training small adapter matrices instead of all model weights.

LoRA vs full fine-tuning?

LoRA uses 10,000x fewer parameters, trains in hours instead of days, and produces models nearly as good as full fine-tuning.

What is QLoRA?

LoRA applied to a 4-bit quantized base model. Even more memory efficient — fine-tune a 65B model on a single GPU.

Related Terms

Learn LoRA in depth

Free hands-on course with code examples and Google Colab notebooks.

Start Course →