Cannot employ a Lambda transform into SubjectsDatasets for Queue

🐛Bug

Hi guys, thanks for the library is quite cool!

I want to employ torchio in a 2D architecture so I am employing the Queue dataset in order to extract patches that represent one slice. Everything works fine when I do not apply transforms or I employ your built-in transforms. However, since I am using a pre-pre-trained model with a different slice orientation I am trying to apply an easy Lambda which just employs torch.transpose. In this case, I get the error showed at traces paragraph.

Just in case let me add that the Lambda transform alone works OK

To reproduce

transpose_transform = tio.Lambda(lambda x: torch.transpose(x, 1,2))
transformadas = tio.Compose([transpose_transform]) #Without or  with built-in transforms works properly
dataset_transform = tio.SubjectsDataset(subjects, transform=transformadas) # I tried with just the transpose (instead a composition) transform but I got the same result

slices_queue = tio.Queue(
    subjects_dataset=dataset_transform,
    max_length=24,
    samples_per_volume=4, # For this test I employ a datsset just with 6 3D-images 
    sampler=tio.sampler.UniformSampler((500,400,1)),
    num_workers=6,
    shuffle_subjects=False,
    shuffle_patches=True)

Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed
Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed
Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed
Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed
Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <function <lambda> at 0x7f3d89c6fb70>: attribute lookup <lambda> on __main__ failed

Expected behavior

Work as with the built-in transforms

“Can’t pickle”

TorchIO version 0.18.0

[PeterMcGor]

Yeah!

[fepegar]

Hi, @Geeks-Sid. Could you please open a new issue and share a minimal, reproducible example that I can run on my Linux computer? That will help me solve your issue.