When updating PyTorch to the current latest 1.8.1 and torchvision to the corresponding 0.9.1 calling torchmeta.transforms.Rotation causes the warning:
~/.cache/pypoetry/virtualenvs/retrainer-TPV0pAOM-py3.8/lib/python3.8/site-packages/torchvision/transforms/functional.py:942: UserWarning: Argument interpolation should be of type InterpolationMode instead of int. Please, use InterpolationMode enum.
Pulling a full stack trace of the warning got me the exact method:
File "~/.cache/pypoetry/virtualenvs/retrainer-TPV0pAOM-py3.8/lib/python3.8/site-packages/torchmeta/transforms/augmentations.py", line 34, in __call__
return F.rotate(image, self.angle % 360, self.resample,
It seems this is the new API. Easy fix is to pass generic **kwargs instead of specific ones which I have already tested, although this be needed for other transforms also. This would be a change more robust to future API changes and flexible for users also I think. I have been using this library a lot for some time now, so let me know if you would be interested in a PR for this.
I'm sorry for the late reply. Thank you very much for the notice, if you have a fix already available I'd be happy to get it in, otherwise I can fix it soon.
When updating PyTorch to the current latest
1.8.1
and torchvision to the corresponding0.9.1
callingtorchmeta.transforms.Rotation
causes the warning:Pulling a full stack trace of the warning got me the exact method:
It seems this is the new API. Easy fix is to pass generic
**kwargs
instead of specific ones which I have already tested, although this be needed for other transforms also. This would be a change more robust to future API changes and flexible for users also I think. I have been using this library a lot for some time now, so let me know if you would be interested in a PR for this.