Fetching the paper…

EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction · Around