Extra Global Attention Designation Using Keyword Detection in Sparse Transformer Architectures

Open in new window