Fetching the paper…

ERNIE-SPARSE: Learning Hierarchical Efficient Transformer Through Regularized Self-Attention · Around