A litter question about CORblock_S about cornet HOT 1 CLOSED

I think it all depends on how you define your model. In our case, we define the weights of all convolutions once (in the __init__), but then call the convolutional operators several times. So when you look at the parameters file, you'll only see the following convolutional weights for V2 block, for instance:

'module.V2.conv_input.weight',
'module.V2.skip.weight',
'module.V2.conv1.weight',
'module.V2.conv2.weight',
'module.V2.conv3.weight'

For each convolution, there is only a single weight stored.

Perhaps you could share a code example of how you were thinking this could cause different weights to be learned for the same convolutional layer?

from cornet.