Fetching the paper…

Various Lengths, Constant Speed: Efficient Language Modeling with Lightning Attention · Around