Fetching the paper…

Adding Recurrence to Pretrained Transformers for Improved Efficiency and Context Size · Around