Question about the "output_padding" in nn.COnvTrasnpose2d

1

Hello everyone,

I’ve got one question about the “output_padding” parameter in nn.ConvTranspose2d,
while what does this parameter means? or where could I find the source realization file,
like c/c++, or anything else so I could figure out the operation.

Thank you very very much ~ ~

2

Have you looked at the documentation for this module? In particular, the second note under nn.ConvTranspose2d has a lot of information on output_padding.

https://pytorch.org/docs/stable/nn.html#torch.nn.ConvTranspose2d

3

There’s lots of information about everything. I do not think it explained the parameter clearly. So what does it mean by “it controls the amount of implicit zero-paddings on both sides of the output for output_padding number of points for each dimension.” “Note that output_padding is only used to find output shape, but does not actually add zero-padding to output.”

1 Like
4

Ignore this. It is inaccurate. See the note and the description below that.

1 Like
6

I still can’t understand this “Note that output_padding is only used to find output shape, but does not actually add zero-padding to output.” even with the note. Could you explain it? Thx!