chunked/nltk_chunking/model_doc_pvt/chunk_5.txt

Unlike the recently proposed Vision 
Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer 
(PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks.