some bugs here , I tried your code ,and ..... about matrix-capsules-pytorch HOT 6 OPEN

activating cuda
Epoch 0
Traceback (most recent call last):
File "train.py", line 96, in
loss = model.loss(out_labels, labels, m)
File "train.py", line 43, in loss
a_t = torch.cat([x[i][target[i]] for i in range(b)]) #b
RuntimeError: zero-dimensional tensor (at position 0) cannot be concatenated

from matrix-capsules-pytorch.

I met this bug too.

from matrix-capsules-pytorch.

Have you solved it?

from matrix-capsules-pytorch.

Change the loss function in train.py to below.

def loss(self, x, target, m): #x:b,10 target:b        
        target = target*x
        zero_tensor = torch.tensor(0.0).cuda()
        loss = torch.max(zero_tensor,m-(target-x))**2
        loss = torch.mean(loss)
        return loss

And rather than passing labels as the 2nd argument to model.loss() pass torch.eye(10).index_select(dim=0,index=labels) this. That is on line 96 the code is
loss = model.loss(out_labels, labels, m) change that to
loss = model.loss(out_labels, torch.eye(10).index_select(dim=0,index=labels), m)

This will work if your use_cuda flag is False. If it is set to True, do the following:
After line 92 define a new variable
labels_zero = torch.eye(10).index_select(dim=0,index=labels).cuda()
and then on line 96
loss = model.loss(out_labels, labels, m) change this to
loss = model.loss(out_labels, labels_zero, m)

from matrix-capsules-pytorch.

Also considering you got an issue at line 43, you may additionally get an error at line 102 acc = pred.eq(labels).cpu().sum().data[0] in that case change it to acc = pred.eq(labels).cpu().sum().item()

from matrix-capsules-pytorch.

Pytorch version problem
a_t = torch.cat([x[i][target[i]].unsqueeze(0) for i in range(b)])

from matrix-capsules-pytorch.