2D inference from 3D volume with Dataloader (Trying to resize storage that is not resizable)

[pauljhin]

I'm using this example below as a base for my code. My code loads the NIfTI images with DataSet and DataLoader and performs the following transformation operations on the Volume 3D.

https://github.com/Project-MONAI/tutorials/blob/main/modules/2d_inference_3d_volume.ipynb

val_transforms = Compose(
    [
        LoadImaged(keys=["image", "label"]),
        EnsureChannelFirstd(keys=["image", "label"]),
        ScaleIntensityRanged(
                        keys=["image"],\
                        a_min=800,\
                        a_max=500,\
                        b_min=0.0,\
                        b_max=1.0,\
                        clip=True,
        ),
        Spacingd(keys=["image", "label"], pixdim=(
            sp_a, sp_b, sp_c)),
    ])

The problem is when making the inference (in the following code snippet).

       slice_inferer = SliceInferer(roi_size=(512, 512), sw_batch_size=1, cval=-1, progress=True)
       val_outputs = slice_inferer(val_inputs, model)

The following error occurs:

RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/opt/conda/lib/python3.10/site-packages/monai/data/utils.py", line 468, in list_data_collate
    ret[key] = collate_meta_tensor(data_for_batch)
  File "/opt/conda/lib/python3.10/site-packages/monai/data/utils.py", line 433, in collate_meta_tensor
    collated = default_collate(batch)
  File "/opt/conda/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 265, in default_collate
    return collate(batch, collate_fn_map=default_collate_fn_map)
  File "/opt/conda/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 124, in collate
    return collate_fn_map[collate_type](batch, collate_fn_map=collate_fn_map)
  File "/opt/conda/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 162, in collate_tensor_fn
    out = elem.new(storage).resize_(len(batch), *list(elem.size()))
  File "/opt/conda/lib/python3.10/site-packages/monai/data/meta_tensor.py", line 268, in __torch_function__
    ret = super().__torch_function__(func, types, args, kwargs)
  File "/opt/conda/lib/python3.10/site-packages/torch/_tensor.py", line 1279, in __torch_function__
    ret = func(*args, **kwargs)
RuntimeError: Trying to resize storage that is not resizable

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

Traceback (most recent call last):
  File "/opt/conda/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py", line 302, in _worker_loop
    data = fetcher.fetch(index)
  File "/opt/conda/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 61, in fetch
    return self.collate_fn(data)
  File "/opt/conda/lib/python3.10/site-packages/monai/data/utils.py", line 483, in list_data_collate
    raise RuntimeError(re_str) from re
RuntimeError: Trying to resize storage that is not resizable

Does anyone know how I can solve it?

Thanks in advance

[KumoLiu]

Hi @pauljhin, what's your "val_inputs"? I think it should be 3D tensor.
https://github.com/Project-MONAI/MONAI/blob/14fcf72a6b733d7b65888d67181fd66d8ebcf755/monai/inferers/inferer.py#L710
Thanks!

[pauljhin]

Hi @KumoLiu

Thank you for your reply. Apologies for my delay.

Yes they are 3D tensor.

The problem has different dimensions. When I use batch size greater than 1 this problem occurs. How do I use larger batch sizes with arbitrary volumetric dimensions?

Thanks!

[KumoLiu]

Hi @pauljhin, could you please try pad_list_data_collate in DataLoader?
https://github.com/Project-MONAI/MONAI/blob/f239825f3b3af097bfba29534cfaaf2eff8051cf/monai/data/utils.py#L652
Thanks!