'Segmentation fault (core dumped)' in image_harmonization about libcom HOT 5 CLOSED

(1) The released code has been validated and did not raise any errors. (2) The provided information is unable to pinpoint the error. It is suggested to run the code step by step to identify the bug.

from libcom.

(1) The released code has been validated and did not raise any errors. (2) The provided information is unable to pinpoint the error. It is suggested to run the code step by step to identify the bug.

During debugging, I found the error raised from './libcom/image_harmonization/source/functions.py' line 489

class TrilinearInterpolationFunction(torch.autograd.Function):
    @staticmethod
    def forward(ctx, lut, x):
        x = x.contiguous()

        output = x.new(x.size())
        dim = lut.size()[-1]
        shift = dim ** 3
        binsize = 1.000001 / (dim-1)
        W = x.size(2)
        H = x.size(3)
        # batch = x.size(0)
        batch = x.size(1)
        
        assert 1 == trilinear.forward(lut, 
                                      x, 
                                      output,
                                      dim, 
                                      shift, 
                                      binsize, 
                                      W, 
                                      H, 
                                      batch)  # <======= Bug came from here

        int_package = torch.IntTensor([dim, shift, W, H, batch])

I followed the solution from HuiZeng/Image-Adaptive-3DLUT#70 , but still no work.
Still needing your help!

from libcom.

If your PyTorch version is not 1.10.1, please try installing this specific version of PyTorch and then reinstall the libcom library by running python setup.py install.

from libcom.

If your PyTorch version is not 1.10.1, please try installing this specific version of PyTorch and then reinstall the libcom library by running python setup.py install.

Thx for reply. I did exactly follow the installation instruction . And the "torch-related" libraries are:

open-clip-torch           2.7.0
pytorch-lightning         1.9.0
torch                     1.10.1
torchaudio                0.10.1
torchmetrics              1.3.0.post0
torchvision               0.11.2

But still no work.

from libcom.

The latest code has fixed the trilinear bugs for PyTorch version 1.11.0 or later. Please consider upgrading PyTorch and rebuilding the library. Hope this can help you. Alternatively, you can try the source code for image harmonization, the link to which is provided in our documentation.

from libcom.