Unkown type name 'torch.device'

🐛 Bug

I ran the one-liner to see if the library works python -c "import kornia; print(kornia.__version__)"

But I got an error

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/kornia/__init__.py", line 18, in <module>
    from kornia import jit
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/kornia/jit/__init__.py", line 8, in <module>
    spatial_soft_argmax2d = torch.jit.script(K.geometry.spatial_soft_argmax2d)
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/torch/jit/__init__.py", line 1281, in script
    fn = torch._C._jit_script_compile(qualified_name, ast, _rcb, get_default_args(obj))
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/torch/jit/_recursive.py", line 555, in try_compile_fn
    return torch.jit.script(fn, _rcb=rcb)
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/torch/jit/__init__.py", line 1281, in script
    fn = torch._C._jit_script_compile(qualified_name, ast, _rcb, get_default_args(obj))
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/torch/jit/_recursive.py", line 555, in try_compile_fn
    return torch.jit.script(fn, _rcb=rcb)
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/torch/jit/__init__.py", line 1281, in script
    fn = torch._C._jit_script_compile(qualified_name, ast, _rcb, get_default_args(obj))
RuntimeError: 
Unknown type name 'torch.device':
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/kornia/utils/grid.py", line 10
        width: int,
        normalized_coordinates: bool = True,
        device: Optional[torch.device] = torch.device('cpu')) -> torch.Tensor:
                         ~~~~~~~~~~~~ <--- HERE
    """Generates a coordinate grid for an image.
'create_meshgrid' is being compiled since it was called from 'spatial_expectation2d'
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/kornia/geometry/dsnt.py", line 89

    # Create coordinates grid.
    grid: torch.Tensor = create_meshgrid(height, width, normalized_coordinates, input.device)
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
    grid = grid.to(input.dtype)
'spatial_expectation2d' is being compiled since it was called from 'spatial_soft_argmax2d'
  File "/home/sandro/anaconda3/envs/supermariopy/lib/python3.7/site-packages/kornia/geometry/spatial_soft_argmax.py", line 515
    """
    input_soft: torch.Tensor = dsnt.spatial_softmax2d(input, temperature)
    output: torch.Tensor = dsnt.spatial_expectation2d(input_soft, normalized_coordinates)
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
    return output

Environment

• PyTorch Version: torch==1.5.0 torchvision==0.5.0 torchviz==0.0.1
• OS: Linux
• How you installed PyTorch: pip
• Python version: python 3.7
• CUDA/cuDNN version: none (CPU)

I guess I am missing something trivial. It would be great if someone could help me figure out my problem. Thanks

[edgarriba]

@theRealSuperMario I will add it somewhere to the readme. This is actually a problem since pip doesn’t solve well dependencies. The package was explicitly built with torch 1.5.0 as install dependency but apparently pip doesn’t throw any error in case you have a different version during installation. It only uses version in case you don’t have any torch in your system, then will download by default minimum required version.

[theRealSuperMario]

I guess it would be greate to note this somewhere on the readme, or did I just not find this piece of information because I was unable to search properly?