Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers.
Yingyu Liang, Heshan Liu, Zhenmei Shi, Zhao Song, Zhuoyan Xu, Jiale Zhao, Zhen Zhuang
Browse the full EMNLP paper archive.
Yingyu Liang, Heshan Liu, Zhenmei Shi, Zhao Song, Zhuoyan Xu, Jiale Zhao, Zhen Zhuang
Browse the full EMNLP paper archive.