Proper error description for float resize dims in tv.transforms.Resize

[MovsisyanM]

🚀 The feature, motivation and pitch

I came across an obscure error message when trying to resize an image:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-55-e056f4a6fffd> in <module>
      2     img = tio.read_image(image_paths.location[i])
      3     resizer = tv.transforms.Resize(((fixed_h), int(fixed_w)))
----> 4     img = resizer(img)
      5     #img = tv.transforms.Resize((int(fixed_h), int(fixed_w)))(img.float())
      6     shape = img.shape

/usr/lib/python3/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
   1108         if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
   1109                 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1110             return forward_call(*input, **kwargs)
   1111         # Do not call functions when jit is used
   1112         full_backward_hooks, non_full_backward_hooks = [], []

/usr/lib/python3/dist-packages/torchvision/transforms/transforms.py in forward(self, img)
    347             PIL Image or Tensor: Rescaled image.
    348         """
--> 349         return F.resize(img, self.size, self.interpolation, self.max_size, self.antialias)
    350 
    351     def __repr__(self) -> str:

/usr/lib/python3/dist-packages/torchvision/transforms/functional.py in resize(img, size, interpolation, max_size, antialias)
    436         return F_pil.resize(img, size=size, interpolation=pil_interpolation, max_size=max_size)
    437 
--> 438     return F_t.resize(img, size=size, interpolation=interpolation.value, max_size=max_size, antialias=antialias)
    439 
    440 

/usr/lib/python3/dist-packages/torchvision/transforms/functional_tensor.py in resize(img, size, interpolation, max_size, antialias)
    488             img = torch.ops.torchvision._interpolate_bicubic2d_aa(img, [new_h, new_w], align_corners=False)
    489     else:
--> 490         img = interpolate(img, size=[new_h, new_w], mode=interpolation, align_corners=align_corners)
    491 
    492     if interpolation == "bicubic" and out_dtype == torch.uint8:

/usr/lib/python3/dist-packages/torch/nn/functional.py in interpolate(input, size, scale_factor, mode, align_corners, recompute_scale_factor, antialias)
   3917         if antialias:
   3918             return torch._C._nn._upsample_bilinear2d_aa(input, output_size, align_corners, scale_factors)
-> 3919         return torch._C._nn.upsample_bilinear2d(input, output_size, align_corners, scale_factors)
   3920     if input.dim() == 5 and mode == "trilinear":
   3921         assert align_corners is not None

TypeError: upsample_bilinear2d() received an invalid combination of arguments - got (Tensor, list, bool, NoneType), but expected one of:
 * (Tensor input, tuple of ints output_size, bool align_corners, tuple of floats scale_factors)
      didn't match because some of the arguments have invalid types: (Tensor, !list!, bool, !NoneType!)
 * (Tensor input, tuple of ints output_size, bool align_corners, float scales_h, float scales_w, *, Tensor out)

From this, I couldn't tell that the shape I wanted to resize to was provided using floats instead of ints. This shouldn't be a hard thing to implement. Thanks for your time

Alternatives

No response

Additional context

No response

cc @albanD @mruberry @jbschlosser @walterddr @mikaylagawarecki

[ezyang]

see #87601 for some improvements, but I guess it's not enough? (Actually I feel this one should have been addressed on master)

[pbanavara]

Looking into this as it is tagged as a good first issue to contribute.
The ideal message should indicate that the second argument - output_size should be tuple of floats instead of tuple of ints ?

[ezyang]

yes, we want to print what kind of list we were given

[MovsisyanM]

Looking into this as it is tagged as a good first issue to contribute. The ideal message should indicate that the second argument - output_size should be tuple of floats instead of tuple of ints ?

It should be a tuple of ints, however a tuple of floats were given

[sambhavnoobcoder]

just confirming, is this issue still open or has it been resolved? i would like to give it a go as well. I am new to the Pytorch and open-source community in general and it would be appreciable if someone pointed me in the right direction.

[mikaylagawarecki]

@sambhavnoobcoder It looks like #99243 was opened to address this issue though it hasn't been landed yet. You can take a look at other issues marked with the label actionable to find other actionable issues if you would like to tackle them