AMLab-Amsterdam/AttentionDeepMIL

Question about batch size (not bag size)

QingboKang opened this issue · 1 comments

Hi, I am just wondering whether the model supports batch size > 1?

I tried batch size > 1 on mnist dataloader and my own data loader, it always said " Expected 4-dimensional input for 4-dimensional weight [20, 1, 5, 5], but got 5-dimensional input of size [2, 10, 1, 28, 28] instead "

thanks!

FYI, I used fixed bag size, so I think it should be supported for batch size > 1

@qdsclove Hi, i meet the same problem with you. May I ask do you have any idea to make batch size > 1, since each bag have different dimension. It's hard to make bags>1(batch size>1) to train the network.