Repository navigation
AttributeError in DataLoader when using RandGridDistortiond transform #8604
Description
Description
I have tried to include RandGridDistortiond transform as part of my data augmentation pipeline. However, when I include this transform the following error raises:
Exception has occurred: AttributeError
Caught AttributeError in DataLoader worker process 0.
AttributeError: Caught AttributeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py", line 351, in _worker_loop
data = fetcher.fetch(index) # type: ignore[possibly-undefined]
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 55, in fetch
return self.collate_fn(data)
File "path/lib/python3.10/site-packages/monai/data/utils.py", line 514, in list_data_collate
ret[key] = collate_fn(data_for_batch)
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 398, in default_collate
return collate(batch, collate_fn_map=default_collate_fn_map)
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 155, in collate
return collate_fn_map[elem_type](batch, collate_fn_map=collate_fn_map)
File "path/lib/python3.10/site-packages/monai/data/utils.py", line 458, in collate_meta_tensor_fn
collated = collate_fn(batch) # type: ignore
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 269, in collate_tensor_fn
numel = sum(x.numel() for x in batch)
File "path/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 269, in <genexpr>
numel = sum(x.numel() for x in batch)
AttributeError: 'int' object has no attribute 'numel'
Reproduction details
This is how I defined the transform:
RandGridDistortiond(keys=["image"], prob=0.3, distort_limit=(-0.35, 0.35), padding_mode='zeros'),
Environment details
Python 3.10.16
MONAI version: 1.3.1
Numpy version: 1.24.3
Pytorch version: 2.5.0+cu124
MONAI flags: HAS_EXT = False, USE_COMPILED = False, USE_META_DICT = False
MONAI rev id: 96bfda0
Optional dependencies:
Pytorch Ignite version: 0.4.11
ITK version: 5.3.0
Nibabel version: 5.1.0
scikit-image version: 0.22.0
scipy version: 1.11.1
Pillow version: 9.5.0
Tensorboard version: 2.14.1
gdown version: 4.7.1
TorchVision version: 0.20.0+cu124
tqdm version: 4.65.0
lmdb version: 1.4.1
psutil version: 5.9.0
pandas version: 2.0.3
einops version: 0.7.0
transformers version: 4.49.0
mlflow version: 2.7.1
pynrrd version: 1.0.0
clearml version: 1.13.1
Found solution
I managed to solve this issue in my case by changing the __call__ function of RandGridDistortiond in monai/transforms/spatial/dictionary.py.
Instead of:
if not self._do_transform:
out: dict[Hashable, torch.Tensor] = convert_to_tensor(d, track_meta=get_track_meta())
return out
Use this:
if not self._do_transform:
for key in self.key_iterator(d):
d[key] = convert_to_tensor(d[key], track_meta=get_track_meta())
return d
For me, this change made the DataLoader work properly.
Activity
Fix RandGridDistortiond crash when transform is skipped
Fix RandGridDistortiond crash when transform is skipped
AlexanderSanin commentedon May 18, 2026
Hi @albertojbc , I've submitted a fix for this in #8862. The issue was that
convert_to_tensorwas called on the entire data dictionary when the transform is skipped, converting metadata integers into 0-dim tensors. The fix only converts the target keys viaself.key_iterator, matching how the transform handles keys in the normal path. Happy to adjust if needed.ericspod commentedon May 19, 2026
@HeyangQin we'd like to get this fix in, you had a PR on this first but there were unresolved issues I mentioned when reviewed. If you'd like to come back to it your PR has the priority here. Thanks!
AlexanderSanin commentedon Jun 17, 2026
I've submitted a PR implementing this: #8920. Please review
Fix RandGridDistortiond crash when transform is skipped (Project-MONA…