Tweeted By @evolvingstuff
Residual Shuffle-Exchange Networks for Fast Processing of Long Sequences
— Thomas (@evolvingstuff) June 5, 2020
"It has O(n log n) complexity and enables processing of sequences up to length 2 million where standard methods, like attention, fail."https://t.co/LoAAzImzVAhttps://t.co/RAKPB6Aa8i pic.twitter.com/x81Sn7NkXK