Images are in torch.cuda.DoubleTensor while the weights are in torch.cuda.FloatTensor.

[kdexd]

Images are in torch.cuda.DoubleTensor while the weights are in torch.cuda.FloatTensor.

[ginxy]

Hi

I had the same problem. The solution, at least the one I have found, is to cast the tensor for the input data to a double type. I've done that by simply adding .double() to the script "extract_features.py" for the class ClevrImagesDataset(Dataset) which loads, transforms and returns a tensor for feature extraction. The function getitem will look like this:

def __getitem__(self, index):
    image = Image.open(self._image_paths[index]).convert("RGB")
    image_tensor = self._transform(image)
    return image_tensor.double()

This should get rid of the dtype issue.

[Hemant-Parihar]

Hi
I still got the same error. When I changed the ClevrImagesDataset(Dataset) class in extract_features.py.

Traceback (most recent call last): File "scripts/preprocess/extract_features.py", line 142, in <module> main(args) File "scripts/preprocess/extract_features.py", line 129, in main output_features = feature_extractor(batch) File "/home/badri/anaconda3/envs/probnmn/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__ result = self.forward(*input, **kwargs) File "/home/badri/anaconda3/envs/probnmn/lib/python3.6/site-packages/torch/nn/modules/container.py", line 92, in forward input = module(input) File "/home/badri/anaconda3/envs/probnmn/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__ result = self.forward(*input, **kwargs) File "/home/badri/anaconda3/envs/probnmn/lib/python3.6/site-packages/torch/nn/modules/conv.py", line 320, in forward self.padding, self.dilation, self.groups) RuntimeError: Input type (torch.cuda.DoubleTensor) and weight type (torch.cuda.FloatTensor) should be the same

[kdexd]

Please pull from latest master, this issue should be fixed. Feel free to re-open!