Fetching the paper…

Layer-wise Pruning of Transformer Attention Heads for Efficient Language Modeling · Around