Memory-efficient Transformers via Top-k Attention.

Ankit Gupta, Guy Dar, Shaya Goodman, David Ciprut, Jonathan Berant

Browse the full EMNLP paper archive.