RuntimeError: Cannot re-initialize CUDA in forked subprocess

[bhjolly]

I cloned the repo, added a training las file, ran python train.py and got the following:

(fsct) [jollyb@styx scripts]$ python train.py
Using default number of CPU cores (all of them).
Processing using  128 / 128  CPU cores.
/home/jollyb/code/FSCT/data directory found.
/home/jollyb/code/FSCT/data/train directory found.
/home/jollyb/code/FSCT/data/train/sample_dir directory created.
Preprocessing train_dataset point clouds...
/home/jollyb/code/FSCT/data/train/example.las
Loading file... /home/jollyb/code/FSCT/data/train/example.las
Pre-processing point cloud...
/home/jollyb/code/FSCT/data/train/all_points_train.las
Loading file... /home/jollyb/code/FSCT/data/train/all_points_train.las
Pre-processing point cloud...
/home/jollyb/code/FSCT/data directory found.
/home/jollyb/code/FSCT/data/validation directory found.
/home/jollyb/code/FSCT/data/validation/sample_dir directory created.
Preprocessing train_dataset point clouds...
/home/jollyb/code/FSCT/data/validation/example.las
Loading file... /home/jollyb/code/FSCT/data/validation/example.las
Pre-processing point cloud...
/home/jollyb/code/FSCT/data/validation/all_points_validate.las
Loading file... /home/jollyb/code/FSCT/data/validation/all_points_validate.las
Pre-processing point cloud...
Running deep learning using  1 / 128  CPU cores.
Loading existing model...
File not found, creating new model...
=====================================================================
EPOCH  0
Traceback (most recent call last):
  File "/home/jollyb/code/FSCT/scripts/train.py", line 389, in <module>
    run_training.run_training()
  File "/home/jollyb/code/FSCT/scripts/train.py", line 284, in run_training
    for data in self.train_loader:
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 521, in __next__
    data = self._next_data()
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
    return self._process_data(data)
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
    data.reraise()
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/_utils.py", line 425, in reraise
    raise self.exc_type(msg)
RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
    data = fetcher.fetch(index)
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/home/jollyb/code/FSCT/scripts/train_datasets.py", line 33, in __getitem__
    x = torch.from_numpy(x.copy()).type(torch.float).to(self.device)
  File "/home/jollyb/miniconda3/envs/fsct/lib/python3.9/site-packages/torch/cuda/__init__.py", line 162, in _lazy_init
    raise RuntimeError(
RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method

I've had a quick look at the code but the solution is not obvious to me, any ideas?