A Training-Free Sub-quadratic Cost Transformer Model Serving Framework with Hierarchically Pruned Attention.

Heejun Lee, Geon Park, Youngwan Lee, Jaduk Suh, Jina Kim, Wonyong Jeong, Bumsik Kim, Hyemin Lee, Myeongjae Jeon, Sung Ju Hwang

Browse the full ICLR paper archive.