RandSpatialCropSamplesd Error

[ove-hansen]

Hi everyone,
I am trying to extract multiple random 2D slices from a 3D volume and am using RandSpatialCropSamplesd to achieve this with an ROI size of [1,128,128] and num_samples=32, which is equivalent to the number of CPUs used in my case. Up to that point I am only using deterministic transforms and the Cached Dataset. However after a few epochs somewhat randomly I get the following error:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/multiprocessing/queues.py", line 239, in _feed
    obj = _ForkingPickler.dumps(obj)
  File "/usr/local/lib/python3.8/multiprocessing/reduction.py", line 51, in dumps
    cls(buf, protocol).dump(obj)
  File "/net/server/home/user/ptlp38/lib/python3.8/site-packages/torch/multiprocessing/reductions.py", line 312, in reduce_storage
    metadata = storage._share_filename_()
RuntimeError: unable to open shared memory object </torch_3595217_3624873813> in read-write mode

For num_samples=16 the error appears less frequently and I have not observed the error for num_samples=8. Does anyone have an idea how this error could be fixed or what might cause it?
Best,
Ove