Gradient-based Intra-attention Pruning on Pre-trained Language Models

Ziqing Yang, Yiming Cui, Xin Yao, Shijin Wang