Fetching the paper…

COMCAT: Towards Efficient Compression and Customization of Attention-Based Vision Models · Around