Writing

Fine-Tuning a Large Vision Model Without Touching Most of Its Weights

How LoRA works, why it exists, and how I used it to adapt SAM for medical image segmentation.