Sigpy.mri.poission with core dumped

[SZUHvern]

When using 'Sigpy.mri.poission' in pytorch's 'Dataloader' to generate random mask(iterative generation), it is easy to occur 'Aborted (core dumped)' in any step and shut down my process. I want to know is there any way to solve this problem. Thanks!

[sidward]

Hi,

Thanks for the report. Could you please post code to reproduce the bug?

[SZUHvern]

Yes, I just do an simply reimplement, like this:

for i in range(1000000): sigpy.mri.poission((320,320)，accel=8, seed=None, tol=0.3) it will running slower and slower and randomly stop at a certain step.

When this program is called into torch.dataloader, an Aborted warning will appear， like this： class test(Dataset): def init(self): super(test, self).init() def len(self): return len(self.examples) def getitem(self, item): mask_sub = sigpy.mri.poisson((320, 320), accel=8, dtype=np.float32, seed=None, tol=0.1) mask_sub = torch.from_numpy(mask_sub).float() return mask_sub

I'm thinking if there is some process that keeps getting called as it iterates until it fills the memory. Is there any way like sigpy.close() to fix this problem?

[sidward]

Ok, thanks for the report. I will look at it when I have time. My gut feeling is that it is related to #102, so if you can try that change and let me know, that would be appreciated.