NotImplementedError: auto_pad=VALID functionality not implemented.

[boy-be-ambitious]

When using latest onnx2pytorch.ConvertModel to import onnx model, it raise this error: NotImplementedError: auto_pad=VALID functionality not implemented.

[mlizhardy]

I'm having a similar issue with NotImplementedError: auto_pad=SAME_UPPER functionality not implemented

[mlizhardy]

looks like from https://github.com/onnx/onnx/blob/master/docs/Operators.md that VALID is padding=0, and SAME_UPPER and SAME_LOWER are padding = ciel(kernel_size/2)

[boy-be-ambitious]

looks like from https://github.com/onnx/onnx/blob/master/docs/Operators.md that VALID is padding=0, and SAME_UPPER and SAME_LOWER are padding = ciel(kernel_size/2)

So, what can I do to solve this issue? I just have a model, no code. Must I wait the author to solve this issue?

[calvinmccarter-at-lightmatter]

Currently it's not implemented, so you're probably out of luck without code. But auto_pad is getting deprecated across ONNX, so if you're able to get access to a newer model, you'll be less likely to have this issue.

[Arka161]

Adding to the issue, we can't do a SAME_UPPER too, and a manual workaround is to add a pass in the else clause, though that of course doesn't replicate the ONNX file into PyTorch 1:1, so that's a huge con.

[Talmaj]

Fixed in v0.5.2