fork vs spawn on MacOS Python 3.9 error

Hi all,

I'm running macos conda-forge on the M1 architecture and testing the mednist_tutorial.ipynb and other jupyter notebooks. I get the following error

  File "/opt/homebrew/Caskroom/miniforge/base/envs/pytorch_env/lib/python3.9/multiprocessing/spawn.py", line 116, in spawn_main
    exitcode = _main(fd, parent_sentinel)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/pytorch_env/lib/python3.9/multiprocessing/spawn.py", line 126, in _main
    self = reduction.pickle.load(from_parent)
AttributeError: Can't get attribute 'MedNISTDataset' on <module '__main__' (built-in)>

This seems to be caused by spawn which apparently became default by Python 3.8 and later. It is fixed by adding the argument multiprocessing_context="fork" to calls of torch.utils.data.DataLoader, however, I suggest that a deeper fixed is made for non-experts.

Steps to reproduce the behavior
On MacOS/python-forge
2. jupyter lab mednist_tutorial.ipynb
3. press the clear kernel and run all button

Expected behavior
A demo network should be set up and trained

Environment

% python --version
Python 3.9.10
% python -c 'import monai; monai.config.print_debug_info()'

================================
Printing MONAI config...
================================
MONAI version: 0.9.dev2210
Numpy version: 1.22.3
Pytorch version: 1.10.2
MONAI flags: HAS_EXT = False, USE_COMPILED = False
MONAI rev id: 1a660e6a7a50e985af5ff76b559baab44175438c
MONAI __file__: /opt/homebrew/Caskroom/miniforge/base/envs/pytorch_env/lib/python3.9/site-packages/monai/__init__.py

Optional dependencies:
Pytorch Ignite version: NOT INSTALLED or UNKNOWN VERSION.
Nibabel version: NOT INSTALLED or UNKNOWN VERSION.
scikit-image version: NOT INSTALLED or UNKNOWN VERSION.
Pillow version: 9.0.1
Tensorboard version: 2.8.0
gdown version: NOT INSTALLED or UNKNOWN VERSION.
TorchVision version: NOT INSTALLED or UNKNOWN VERSION.
tqdm version: 4.63.0
lmdb version: NOT INSTALLED or UNKNOWN VERSION.
psutil version: NOT INSTALLED or UNKNOWN VERSION.
pandas version: NOT INSTALLED or UNKNOWN VERSION.
einops version: 0.4.1
transformers version: NOT INSTALLED or UNKNOWN VERSION.
mlflow version: NOT INSTALLED or UNKNOWN VERSION.

For details about installing the optional dependencies, please visit:
    https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies


================================
Printing system config...
================================
`psutil` required for `print_system_info`

================================
Printing GPU config...
================================
Num GPUs: 0
Has CUDA: False
cuDNN enabled: False

Thank you for a nice package, Jon

Beginne mit mednist_tutorial.ipynb in Jupyter Lab unter macOS mit Python 3.9, reproduziere den Fehler, indem du den Kernel leerst und alles ausführst, und untersuche anschließend die Aufrufe von torch.utils.data.DataLoader im Notebook. Als erledigt gilt die Aufgabe, wenn das Demo-Netzwerk ohne den Multiprocessing-Fehler von MedNISTDataset eingerichtet und trainiert werden kann.