RElbers/ada-conv-pytorch

Bias before pointwise conv

ariG23498 opened this issue · 2 comments

ariG23498 commented

ada-conv-pytorch/lib/nn/adaconv/adaconv.py

Line 44 in bb3bab5

x = F.conv2d(x, w_spatial, groups=self.n_groups, bias=bias)

In the paper the bias is added after the pointwise cov. Is there a reason why we do it before here?

RElbers commented

I think you are correct and that's a bug in the code. I have been refactoring the code as well. I'll push the new code with this bug fixed later today.

  • 👍1
ariG23498 commented

I think this has been sorted in the latest commit. Feel free to close this!