RuntimeError: stack expects each tensor to be equal size, but got [1, 128, 224, 224] at entry 0 and [1, 224, 224, 128] at entry 3

[lg2578]

RuntimeError: stack expects each tensor to be equal size, but got [1, 128, 224, 224] at entry 0 and [1, 224, 224, 128] at entry 3
it's strange,who can help me?
my main code:

import monai.transforms as T
train_transforms = T.Compose(
    [
        T.LoadImaged(keys=["img", "seg"]),
        T.AddChanneld(keys=["img", "seg"]),
        T.ScaleIntensityd(keys=["img", "seg"]),
        T.RandCropByPosNegLabeld(
            keys=["img", "seg"],
            label_key="seg",
            spatial_size=[224, 224, 128],
            pos=1,
            neg=1,
            num_samples=4,
        ),
        T.RandRotate90d(keys=["img", "seg"], prob=0.5, spatial_axes=[0, 2]),
        T.EnsureTyped(keys=["img", "seg"]),
    ]
)
val_transforms = T.Compose(
    [
        T.LoadImaged(keys=["img", "seg"]),
        T.AddChanneld(keys=["img", "seg"]),
        T.ScaleIntensityd(keys=["img", "seg"]),
        T.EnsureTyped(keys=["img", "seg"]),
    ]
)
# define dataset, data loader
from monai.data import list_data_collate,  decollate_batch,pad_list_data_collate
check_ds = monai.data.Dataset(data=train_files, transform=train_transforms)
# use batch_size=2 to load images and use RandCropByPosNegLabeld to generate 2 x 4 images for network training
check_loader = monai.data.DataLoader(check_ds, batch_size=2, num_workers=2, collate_fn=pad_list_data_collate)
check_data = monai.utils.misc.first(check_loader)
print(check_data["img"].shape, check_data["seg"].shape)

# create a training data loader
train_ds = monai.data.CacheDataset(data=train_files,cache_rate=1, transform=train_transforms)
# use batch_size=2 to load images and use RandCropByPosNegLabeld to generate 2 x 4 images for network training
train_loader = DataLoader(
    train_ds,
    batch_size=128,
    shuffle=True,
    num_workers=2,
    collate_fn=pad_list_data_collate,
    pin_memory=torch.cuda.is_available(),
)
# create a validation data loader
val_ds = monai.data.CacheDataset(data=val_files,cache_rate=1, transform=val_transforms)
val_loader = monai.data.DataLoader(val_ds, batch_size=1, num_workers=2, collate_fn=pad_list_data_collate)

the errors:

Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7f8249983200>
Traceback (most recent call last):
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1358, in __del__
    self._shutdown_workers()
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1341, in _shutdown_workers
    if w.is_alive():
  File "/opt/conda/lib/python3.7/multiprocessing/process.py", line 151, in is_alive
    assert self._parent_pid == os.getpid(), 'can only test a child process'
AssertionError: can only test a child process
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7f8249983200>
Traceback (most recent call last):
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1358, in __del__
    self._shutdown_workers()
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 1341, in _shutdown_workers
    if w.is_alive():
  File "/opt/conda/lib/python3.7/multiprocessing/process.py", line 151, in is_alive
    assert self._parent_pid == os.getpid(), 'can only test a child process'
AssertionError: can only test a child process
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_33/2866364715.py in <module>
      4 # use batch_size=2 to load images and use RandCropByPosNegLabeld to generate 2 x 4 images for network training
      5 check_loader = monai.data.DataLoader(check_ds, batch_size=2, num_workers=2, collate_fn=pad_list_data_collate)
----> 6 check_data = monai.utils.misc.first(check_loader)
      7 print(check_data["img"].shape, check_data["seg"].shape)
      8 

/kaggle/input/monai-v081/monai/utils/misc.py in first(iterable, default)
     80     Returns the first item in the given iterable or `default` if empty, meaningful mostly with 'for' expressions.
     81     """
---> 82     for i in iterable:
     83         return i
     84     return default

/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in __next__(self)
    528             if self._sampler_iter is None:
    529                 self._reset()
--> 530             data = self._next_data()
    531             self._num_yielded += 1
    532             if self._dataset_kind == _DatasetKind.Iterable and \

/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _next_data(self)
   1222             else:
   1223                 del self._task_info[idx]
-> 1224                 return self._process_data(data)
   1225 
   1226     def _try_put_index(self):

/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _process_data(self, data)
   1248         self._try_put_index()
   1249         if isinstance(data, ExceptionWrapper):
-> 1250             data.reraise()
   1251         return data
   1252 

/opt/conda/lib/python3.7/site-packages/torch/_utils.py in reraise(self)
    455             # instantiate since we don't know how to
    456             raise RuntimeError(msg) from None
--> 457         raise exception
    458 
    459 

RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "../input/monai-v081/monai/data/utils.py", line 356, in list_data_collate
    ret = default_collate(data)
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py", line 157, in default_collate
    return elem_type({key: default_collate([d[key] for d in batch]) for key in elem})
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py", line 157, in <dictcomp>
    return elem_type({key: default_collate([d[key] for d in batch]) for key in elem})
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py", line 138, in default_collate
    return torch.stack(batch, 0, out=out)
RuntimeError: stack expects each tensor to be equal size, but got [1, 128, 224, 224] at entry 0 and [1, 224, 224, 128] at entry 3

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
    data = fetcher.fetch(index)
  File "/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 52, in fetch
    return self.collate_fn(data)
  File "../input/monai-v081/monai/data/utils.py", line 531, in pad_list_data_collate
    return PadListDataCollate(method=method, mode=mode, **np_kwargs)(batch)
  File "../input/monai-v081/monai/transforms/croppad/batch.py", line 110, in __call__
    return list_data_collate(batch)
  File "../input/monai-v081/monai/data/utils.py", line 372, in list_data_collate
    raise RuntimeError(re_str) from re
**RuntimeError: stack expects each tensor to be equal size, but got [1, 128, 224, 224] at entry 0 and [1, 224, 224, 128] at entry 3**

MONAI hint: if your transforms intentionally create images of different shapes, creating your `DataLoader` with `collate_fn=pad_list_data_collate` might solve this problem (check its documentation).

[Nic-Ma]

Hi @lg2578 ,

Thanks for your interest here.
I think that because you executed random spatial transform RandRotate90d(keys=["img", "seg"], prob=0.5, spatial_axes=[0, 2]).
It's prob=0.5, so some images are rotated, some not.

Thanks.

[gssriram]

Hi @lg2578,

Since you have used spatial_axes=[0, 2] in T.RandRotate90d, it would essentially apply the rotation to the S and Y dimensions of your input image (C, S, X, Y). That's why the transformed image has taken the shape of (1, 224, 224, 128) and other non-transformed images (due to probability=0.5) would be untouched with the original dimension of (1, 128, 224, 224).

Due to the mismatch of dimensions between the images, the error has triggered. Using the spatial_axes=[1, 2] would rotate X and Y dimensions and work as expected.