×
Only 5% attention is all you need: Efficient long-range document-level neural machine translation. Zihan Liu, Zewei Sun, Shanbo Cheng, Shujian Huang, Mingxuan ...
Sep 25, 2023 · Only 5\% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation. Authors:Zihan Liu, Zewei Sun, Shanbo Cheng, ...
Nov 4, 2023 · Experimen- tal results show that our method could achieve up to 95% sparsity (only 5% tokens attended) approximately, and save 93% computation ...
This work keeps the translation performance while gaining 20% speed up by introducing extra selection layer based on lightweight attention that selects a ...
Dec 31, 2022 · Only 5% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation. Download PDF · Open Webpage · Zihan Liu, ...
... 5% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation | Find, read and cite all the research you need on ResearchGate.
Only 5% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation ... Translation Dataset for Multimodal Machine Translation ...
People also ask
Jun 12, 2017 · We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Only 5\% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation ... Semantics-aware Attention Improves Neural Machine ...
Only 5% Attention Is All You Need: Efficient Long-range Document-level Neural Machine Translation. Conference Paper. Jan 2023. Zihan Liu · Zewei Sun ...