`RandKSpaceSpikeNoise` got an unexpected keyword argument 'dim'

[mibaumgartner]

Describe the bug
RandKSpaceSpikeNoise does not work with NumPy array if intensity_range is not set manually.

In case a NumPy array is provided, the linked part will call np.mean(.., dim=...) but NumPy does not support dim arguments (should be axis). The implementation works fine for torch.Tensors.

MONAI/monai/transforms/intensity/array.py

Lines 1816 to 1818 in d6327a7

	mod = torch if isinstance(k, torch.Tensor) else np
	log_abs = mod.log(mod.absolute(k) + 1e-10)
	shifted_means = mod.mean(log_abs, dim=tuple(range(-n_dims, 0))) * 2.5

To Reproduce
Call transform on NumPy array with default values (specifically intensity_range =None).

Expected behavior
No crash.

Screenshots
If applicable, add screenshots to help explain your problem.

Environment

Ensuring you use the relevant python executable, please paste the output of:

python -c 'import monai; monai.config.print_debug_info()'

Additional context
Add any other context about the problem here.

[Nic-Ma]

Hi @mibaumgartner ,

Thanks for your feedback!
@yiheng-wang-nv , could you please help confirm and fix it?
I think changing it to mod.mean(log_abs, tuple(range(-n_dims, 0))) * 2.5 may help.

Thanks in advance.