Add warning if SubjectsLoader is not used in PyTorch >= 2.3 (#1215)

* Add warning if SubjectsLoader is not used

* Improve warning and move to method

Co-authored-by: valabregue <[email protected]>

src/torchio/data/image.py

@@ -36,6 +36,7 @@
 from ..typing import TypeTripletInt
 from ..utils import get_stem
 from ..utils import guess_external_viewer
+from ..utils import in_torch_loader
 from ..utils import is_iterable
 from ..utils import to_tuple
 from .io import check_uint_to_int
@@ -176,6 +177,7 @@ def __init__(
  warnings.warn(message, DeprecationWarning, stacklevel=2)
 
  super().__init__(**kwargs)
+ self._check_data_loader()
  self.path = self._parse_path(path)
 
  self[PATH] = '' if self.path is None else str(self.path)
@@ -234,6 +236,20 @@ def __copy__(self):
  )
  return new_image
 
+ @staticmethod
+ def _check_data_loader() -> None:
+ if torch.__version__ >= '2.3' and in_torch_loader():
+ message = (
+ 'Using TorchIO images without a torchio.SubjectsLoader in PyTorch >='
+ ' 2.3 might have unexpected consequences, e.g., the collated batches'
+ ' will be instances of torchio.Subject with 5D images. Replace'
+ ' your PyTorch DataLoader with a torchio.SubjectsLoader so that'
+ ' the collated batch becomes a dictionary, as expected. See'
+ ' https:/fepegar/torchio/issues/1179 for more'
+ ' context about this issue.'
+ )
+ warnings.warn(message, stacklevel=1)
+
  @property
  def data(self) -> torch.Tensor:
  """Tensor data (same as :class:`Image.tensor`)."""

src/torchio/utils.py

@@ -2,6 +2,7 @@
 
 import ast
 import gzip
+import inspect
 import os
 import shutil
 import sys
@@ -19,6 +20,7 @@
 import numpy as np
 import SimpleITK as sitk
 import torch
+import torch.utils.data.dataloader
 from nibabel.nifti1 import Nifti1Image
 from torch.utils.data import DataLoader
 from torch.utils.data._utils.collate import default_collate
@@ -415,3 +417,24 @@ def is_iterable(object: Any) -> bool:
  return True
  except TypeError:
  return False
+
+
+def in_class(classes) -> bool:
+ classes = to_tuple(classes)
+ stack = inspect.stack()
+ for frame_info in stack:
+ instance = frame_info.frame.f_locals.get('self')
+ if instance is None:
+ continue
+ if instance.__class__ in classes:
+ return True
+ else:
+ return False
+
+
+def in_torch_loader() -> bool:
+ classes = (
+ torch.utils.data.dataloader._SingleProcessDataLoaderIter,
+ torch.utils.data.dataloader._MultiProcessingDataLoaderIter,
+ )
+ return in_class(classes)