Linearly Decomposing and Recomposing Vision Transformers for Diverse-Scale Models

Open in new window