Run error

[liusheng0620]

When I randomly generate a tensor, It's OK. But when I input a picture, it will report an error. How can I solve it?
Code:
d = 3
generate a tensor
input = torch.randn(2,2,2,2)
input a picture
input = Image.open('/data/lu/ls/data/BDDU/ShadowImages/04056725-14e1b11d-00001.png' ).convert('RGB')
input = torch.from_numpy(np.array(input)).unsqueeze(0).permute(0,3,2,1)
out = spatial_correlation_sample(input, input, kernel_size=d, patch_size=d, stride=1, padding=int(d/2))

Error:
terminate called after throwing an instance of 'c10::Error'
what(): "correlation_forward_cpp" not implemented for 'Byte' (operator() at Correlation_Module/correlation.cpp:119)
frame #0: c10::Error::Error(c10::SourceLocation, std::string const&) + 0x47 (0x7fb1fe957687 in /home/amax/anaconda3/envs/py36tor131/lib/python3.6/site-packages/torch/lib/libc10.so)
frame #1: + 0x35fc3 (0x7fb1fdc76fc3 in /home/amax/anaconda3/envs/py36tor131/lib/python3.6/site-packages/spatial_correlation_sampler-0.3.0-py3.6-linux-x86_64.egg/spatial_correlation_sampler_backend.cpython-36m-x86_64-linux-gnu.so)
frame #2: + 0x1465f (0x7fb1fe48665f in /home/amax/anaconda3/envs/py36tor131/lib/python3.6/site-packages/torch/lib/../../../../libgomp.so.1)
frame #3: + 0x76ba (0x7fb20311f6ba in /lib/x86_64-linux-gnu/libpthread.so.0)
frame #4: clone + 0x6d (0x7fb202e5551d in /lib/x86_64-linux-gnu/libc.so.6)

terminate called recursively
已放弃

[ClementPinard]

Are you sure your image is converted to float ? Looks like you are trying to run the correlation on uint8 tensors, which might not be available.