Efficient Low-rank Backpropagation for Vision Transformer Adaptation

The increasing scale of vision transformers (ViT) has made the efficient fine-tuning of these large models for specific needs a significant challenge in various applications.