Error if CropOrPad and RescaleIntensity are used together

🐛Bug

I have an error when using ‘CropOrPad’ and ‘RescaleIntensity’ as part of the dataset transformations. Note that there is no error if the two transformations are switched.

To reproduce

from torchvision.transforms import Compose
from torch.utils.data import DataLoader

import torchio
from torchio.transforms import (
    RescaleIntensity,
    ToCanonical,
    CropOrPad,
)
from torchio import SubjectsDataset, DATA


transformations = (
        ToCanonical(),
        CropOrPad((192,192,128)),
        RescaleIntensity((-1,1),percentiles=(0.05, 99.5))
        )
transformations = Compose(transformations)

subject = torchio.datasets.Colin27()
dataset = SubjectsDataset([subject], transform=transformations)
loader = DataLoader(dataset, batch_size=1)
for batch in loader:
    print(batch['brain']['stem'])

  File "<stdin>", line 1, in <module>                                                                                                                                                                                                          
  File "/opt/conda/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 345, in __next__                                                                                                                                          
    data = self._next_data()                                                                                                                                                                                                                   
  File "/opt/conda/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 385, in _next_data                                                                                                                                        
    data = self._dataset_fetcher.fetch(index)  # may raise StopIteration                                                                                                                                                                       
  File "/opt/conda/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch                                                                                                                                            
    data = [self.dataset[idx] for idx in possibly_batched_index]                                                                                                                                                                               
  File "/opt/conda/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>                                                                                                                                       
    data = [self.dataset[idx] for idx in possibly_batched_index]                                                                                                                                                                               
  File "/workspace/home/local4/.local/lib/python3.6/site-packages/torchio/data/dataset.py", line 87, in __getitem__                                                                                                                            
    sample = self._transform(sample)                                                                                                                                                                                                           
  File "/opt/conda/lib/python3.6/site-packages/torchvision/transforms/transforms.py", line 70, in __call__                                                                                                                                     
    img = t(img)                                                                                                                                                                                                                               
  File "/workspace/home/local4/.local/lib/python3.6/site-packages/torchio/transforms/transform.py", line 95, in __call__                                                                                                                       
    transformed = self.apply_transform(sample)                                                                                                                                                                                                 
  File "/workspace/home/local4/.local/lib/python3.6/site-packages/torchio/transforms/preprocessing/intensity/normalization_transform.py", line 69, in apply_transform                                                                          
    self.apply_normalization(sample, image_name, mask)                                                                                                                                                                                         
  File "/workspace/home/local4/.local/lib/python3.6/site-packages/torchio/transforms/preprocessing/intensity/rescale.py", line 55, in apply_normalization                                                                                      
    image_dict[DATA] = self.rescale(image_dict[DATA], mask, image_name)                                                                                                                                                                        
  File "/workspace/home/local4/.local/lib/python3.6/site-packages/torchio/transforms/preprocessing/intensity/rescale.py", line 63, in rescale                                                                                                  
    array = tensor.clone().numpy()                                                                                                                                                                                                             
RuntimeError: unsupported operation: more than one element of the written-to tensor refers to a single memory location. Please clone() the tensor before performing the operation.

TorchIO et co version numpy 1.18.1 torch 1.5.0a0+8f84ded torchio 0.17.42 SimpleITK 1.2.2

[ReubenDo]

Upgrading torch solved the problem. For the record, I was using the NVIDIA docker container: nvcr.io/nvidia/pytorch:20.03-py.

Thanks for your help guys!

[fepegar]

Can’t reproduce with your setup either (except I installed torch 1.5.0 instead of that specific revision). Can you upgrade torch and try again? If it doesn’t work I’ll try to reproduce on Linux next week.