Fetching the paper…

LSG Attention: Extrapolation of pretrained Transformers to long sequences · Around