search

lironui / Multi-Attention-Network

The semantic segmentation of remote sensing images
https://lironui.github.io/
GNU Affero General Public License v3.0
67 stars 7 forks source link

kv=torch.einsum("bmn, bcn->bmc", K, V) #7

Open yuanqinzhi opened 2 years ago

yuanqinzhi commented 2 years ago

I read your paper and have some doubts here, shouldn't it be the dot product of K and Q? Why is K and V doing dot product in the code