.h5 data load in mri_3d example

[jonbmartin]

When running examples/demo_3d.ipynb on Ubuntu system with GPU, CUDA 11.8, Python 3.10, x86_64 architecture I run into an issue using h5py to load the .h5 data for the example. I don't encounter on my local system when running things on CPU. May relate to h5py/h5py#817.

---

ValueError Traceback (most recent call last)
Cell In[15], line 12
10 device0 = torch.device('cuda:0')
11 # Load Non-cartesian k-space trajectory, [nbatch (1), ndimension, nshot, nreadout]
---> 12 ktraj = torch.tensor(hf['ktraj'][()], dtype=torch.double).unsqueeze(0).to(device = device0, dtype = torch.float)
13 print('traj shape', ktraj.shape)
14 # Load k-space, [nbatch (1), ncoil, nshot, nreadout]

File h5py/_objects.pyx:54, in h5py._objects.with_phil.wrapper()

File h5py/_objects.pyx:55, in h5py._objects.with_phil.wrapper()

File ~/python_venvs/sheeprl/lib/python3.10/site-packages/h5py/_hl/dataset.py:781, in Dataset.getitem(self, args, new_dtype)
779 if self._fast_read_ok and (new_dtype is None):
780 try:
--> 781 return self._fast_reader.read(args)
782 except TypeError:
783 pass # Fall back to Python read pathway below

File ~/python_venvs/sheeprl/lib/python3.10/site-packages/h5py/_hl/dataset.py:551, in Dataset._fast_reader(self)
548 if '_fast_reader' in self._cache_props:
549 return self._cache_props['_fast_reader']
--> 551 rdr = _selector.Reader(self.id)
553 # If the file is read-only, cache the reader to speed up future uses.
554 # This cache is invalidated by .refresh() when using SWMR.
555 if self._readonly:

File h5py/_selector.pyx:320, in h5py._selector.Reader.cinit()

File h5py/h5t.pyx:1087, in h5py.h5t.TypeFloatID.py_dtype()

ValueError: Insufficient precision in available types to represent (63, 52, 11, 0, 52)