RuntimeError: inconsistent tensor sizes at /pytorch/torch/lib/TH/generic/THTensorMath.c:2864

[eromoe]

env:

• pytorch 0.3
• python 3.5
• cuda8.0

inpainting.ipynb got this error at

Visualize : plot_image_grid([img_np, img_mask_np, img_mask_np*img_np], 3,11);

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-e9e29a4716c2> in <module>()
      1 img_mask_var = np_to_var(img_mask_np).type(dtype)
      2 
----> 3 plot_image_grid([img_np, img_mask_np, img_mask_np*img_np], 3,11);

/data/notebooks/deep-image-prior/utils/common_utils.py in plot_image_grid(images_np, nrow, factor, interpolation)
     70         interpolation: interpolation used in plt.imshow
     71     """
---> 72     grid = get_image_grid(images_np, nrow)
     73 
     74     plt.figure(figsize=(len(images_np)+factor,12+factor))

/data/notebooks/deep-image-prior/utils/common_utils.py in get_image_grid(images_np, nrow)
     57     '''Creates a grid from a list of images by concatenating them.'''
     58     images_torch = [torch.from_numpy(x) for x in images_np]
---> 59     torch_grid = torchvision.utils.make_grid(images_torch, nrow)
     60 
     61     return torch_grid.numpy()

/usr/local/lib/python3.5/dist-packages/torchvision/utils.py in make_grid(tensor, nrow, padding, normalize, range, scale_each, pad_value)
     33     # if list of tensors, convert to a 4D mini-batch Tensor
     34     if isinstance(tensor, list):
---> 35         tensor = torch.stack(tensor, dim=0)
     36 
     37     if tensor.dim() == 2:  # single image H x W

/usr/local/lib/python3.5/dist-packages/torch/functional.py in stack(sequence, dim, out)
     62     inputs = [t.unsqueeze(dim) for t in sequence]
     63     if out is None:
---> 64         return torch.cat(inputs, dim)
     65     else:
     66         return torch.cat(inputs, dim, out=out)

RuntimeError: inconsistent tensor sizes at /pytorch/torch/lib/TH/generic/THTensorMath.c:2864

[DmitryUlyanov]

Do you run it on example images or your own?

[eromoe]

On example images. Same notebook without modification.

## Fig 7 (top)
img_path  = 'data/inpainting/lena.png'
mask_path = 'data/inpainting/lena_mask.png'

NET_TYPE = 'skip_depth6' # one of skip_depth4|skip_depth2|UNET|ResNet

[turbo]

Appears to be the same issue as #12

[philtomson]

Also getting this same error on Linux with:

• pytorch 0.2.0.post3

• Python 2.7.14

• cuda 8.0

using the lena example image (didn't edit the original code)

---

RuntimeError Traceback (most recent call last)

in () 1 img_mask_var = np_to_var(img_mask_np).type(dtype) 2 ----> 3 plot_image_grid([img_np, img_mask_np, img_mask_np*img_np], 3,11); /home/phil/devel/ML/deep-image-prior/utils/common_utils.pyc in plot_image_grid(images_np, nrow, factor, interpolation) 70 interpolation: interpolation used in plt.imshow 71 """ ---> 72 grid = get_image_grid(images_np, nrow) 73 74 plt.figure(figsize=(len(images_np)+factor,12+factor)) /home/phil/devel/ML/deep-image-prior/utils/common_utils.pyc in get_image_grid(images_np, nrow) 57 '''Creates a grid from a list of images by concatenating them.''' 58 images_torch = [torch.from_numpy(x) for x in images_np] ---> 59 torch_grid = torchvision.utils.make_grid(images_torch, nrow) 60 61 return torch_grid.numpy() /home/phil/py-virt-env/env/local/lib/python2.7/site-packages/torchvision/utils.pyc in make_grid(tensor, nrow, padding, normalize, range, scale_each, pad_value) 33 # if list of tensors, convert to a 4D mini-batch Tensor 34 if isinstance(tensor, list): ---> 35 tensor = torch.stack(tensor, dim=0) 36 37 if tensor.dim() == 2: # single image H x W /home/phil/py-virt-env/env/local/lib/python2.7/site-packages/torch/functional.pyc in stack(sequence, dim, out) 62 inputs = [t.unsqueeze(dim) for t in sequence] 63 if out is None: ---> 64 return torch.cat(inputs, dim) 65 else: 66 return torch.cat(inputs, dim, out=out) RuntimeError: inconsistent tensor sizes at /pytorch/torch/lib/TH/generic/THTensorMath.c:2709

[DmitryUlyanov]

Should work now: https://github.com/DmitryUlyanov/deep-image-prior/commit/8067fbbae78637e743ce40f362d0bdfdebfcd6ad.

[philtomson]

yes, it works now. Thanks for the quick response.

[eromoe]

So it is due to lena is 3 channels but lena_mask is 1 chennel . Thank you @DmitryUlyanov .