Minor bugfix for type A shortcut

[lim0606]

The original one gave me an error when shorcut = type A and nOutputPlane ~= nInpuPlane x 2 (when opt.dataset = imagenet).

This might not be the optimal solution, but I think it could be a palliative for now.

Best regards,

Jaehyun

[ghost]

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have you on file. In order for us to review and merge your code, please sign up at https://code.facebook.com/cla - and if you have received this in error or have any questions, please drop us a line at cla@fb.com. Thanks!

[ghost]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Facebook open source project. Thanks!

[colesbury]

Thanks for the PR and the bug description. This looks like it works, but I'm not thrilled with the idea of using the convolution to add the zero-valued feature planes. We should probably just add a sub-class nn.Module to do the zero-padding.

[lim0606]

@colesbury
Exactly! However, I couldn't think of a better idea for broaching this bug rather than adding (temporaray) solution :)

Best regards,

Jaehyun

[xichangzun]

it did work in my code, thanks a lot