More user-friendly crash message when model is not found

[oesteban]

What happened?

nipreps/nifreeze@f0569bc

When the model isn't present, the interface only crashes when the output mask is being "aggregated" (by nipype) and evidently it's not found where expected.

What command did you use?

n/a

What version of the software are you running?

nifreeze

How are you running this software?

Local installation ("bare-metal")

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

No

Please copy and paste any relevant log output.

FileNotFoundError                         Traceback (most recent call last)
Cell In[4], line 19
     14 if not bmask_path.exists():
     15     bmsk_results = SynthStrip(
     16         in_file=str(avg_path),
     17         use_gpu=True,
     18     ).run(cwd=str(WORKDIR))
---> 19     copy(bmsk_results.outputs.out_mask,bmask_path)

(...)

FileNotFoundError: [Errno 2] No such file or directory:
 '/home/runner/tmp/nifreezedev/ismrm25/sub-01_task-mixedgamblestask_run-03_boldref_desc-brain_mask.nii.gz'

Additional information / screenshots

No response

[jhlegarreta]

As I was debugging the NiFreeze PR, after fixing the model weights file issue, it became apparent that the failure message was then due to the fact CUDA drivers were not being found:
nipreps/nifreeze#117 (comment)

In this case, the error message was descriptive; however, I'm wondering whether it would be beneficial to the user to offer the possibility of an automatic fallback to using the CPU in case they are not found. Besides some management in the code, this may require changing the use_gpu flag description and/or name (the latter breaking the API).

[mgxd]

Good catch - I think use_gpu should be a preference - ultimately we should fallback to using the CPU (with a warning) if it is not available.

We'll have to update the logic

synthstrip/nipreps/synthstrip/cli.py

Lines 91 to 102 in dd29a6c

	# configure GPU device
	if args.gpu:
	os.environ['CUDA_VISIBLE_DEVICES'] = '0'
	device = torch.device('cuda')
	device_name = 'GPU'
	else:
	os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
	device = torch.device('cpu')
	device_name = 'CPU'
	if args.num_threads and args.num_threads > 0:
	torch.set_num_threads(args.num_threads)