CASP: Compression of Large Multimodal Models Based on Attention Sparsity.

Mohsen Gholami, Mohammad Akbari, Kevin Cannons, Yong Zhang

Browse the full CVPR paper archive.