Resume an interrupted run

[Manuel Blesa Cábez]

Dear Gary,

I'm running NODDI for a lot of subjects in a server and sometimes it crashes. Then if a re run it I find the following:

Resume an interrupted run from 54101

If I'm right this means that the process continues from the las place it was stopped. 

My question is: is it possible if then I start another subject, the process  will start using the steps for the previous subject? Even if I remove the FittedParams.mat?

Thanks in advance.

Best regards,

Manuel

[Hui Zhang]

HI Manuel,

Yes, that is the design.  The interruption is recorded in FittedParams.mat (take a look at the mat file); so if you delete it, the new run will start from the very beginning.  The best strategy is to give each subject a unique FittedParams.mat file name, e.g. by prefixing it with the subject ID.

Gary

--
You received this message because you are subscribed to the Google Groups "NODDI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to noddi+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/77a981aa-36c2-474d-ab18-804584ea0b0d%40googlegroups.com.

[Manuel Blesa Cábez]

Hi Gary,

Thanks for your reply. I'll do what you suggested from now on.

For the maps that I already have, is it safe to inspect it and check that they look correct? or to be sure I should run it again?

Best regards,

Manuel

[Hui Zhang]

Let me know a bit more of your set up.  Do you have all the subjects in a single folder, or they all are in their individual folders?

Gary

--
You received this message because you are subscribed to the Google Groups "NODDI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to noddi+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/f156312b-3589-44dc-adbc-9ef3a1189d11%40googlegroups.com.

[Manuel Blesa Cábez]

Hi Gary,

I have all the subjects innthe same folder, I could easyly change this in the future, but all the subjects I have calculated, they were in the same folder.

Best regards,

Manuel

[Hui Zhang]

Ok.  You must be running one subject at a time in turn, right?  If, when a crash has occurred, you know exactly what subject you are fitting at that moment, and if you restart the fitting from that subject, you should be okay.

Inspecting the output maps can help detect any interrupted run if there are enough voxels missing.  It is conceivable that if it is just a handful of voxels, you may not be able to catch the problem.

I will update the toolbox so that SaveParamsAsNIfTI will not proceed unless FittedParams.mat is complete.

Gary

--
You received this message because you are subscribed to the Google Groups "NODDI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to noddi+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/f500098f-b64d-4195-86e5-b42cd3dc3fd1%40googlegroups.com.

[Manuel Blesa Cábez]

Hi Gary,

Thanks for your reply. I was doing exactly that, and for the old subjects it only crashed a couple of times. The only thing I'm not 100% sure is if I restrted dor the same subject or the previous one, but in that case it should only affect to a couple of subjects, am I right? For now on, I will process it subject in his own folder with his own fittedparameters.mat file.

Best regards,

Manuel

[Hui Zhang]

Ok.  If you can't be sure for which subject a run is interrupted and may have resumed the interrupted run with the previous subject, the outcome is unpredictable I am afraid.  It is entirely possible that this may have corrupted that previous subject.

This is not going to help with what has already happened but I have just released version 1.04 that will prevent SaveParamsAsNIfTI to operate a FittedParams.mat that is saved from an interrupted run.  But this still will not protect you from resuming from the wrong subject.  So it is essential that you either have each subject in its own subfolder or give each subject unique filenames for all the relevant files.

Gary

--
You received this message because you are subscribed to the Google Groups "NODDI" group.
To unsubscribe from this group and stop receiving emails from it, send an email to noddi+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/b673a998-a762-4dc5-baca-fba6a9b96aae%40googlegroups.com.

[zeno Zeng]

Hi Gray, 

So if it is saved in their individual folder: subject_id/FittedParams.mat, the fitting will continue from the last recorded time and there will be no influence on the NODDI_roi.mat and FittedParams.mat either running single batch fitting or the one on 8 core, is it correct?

Thank you so much!

Best, 

Qi

Gary 在 2020年5月24日 星期日下午1:24:01 [UTC-4] 的信中寫道：

[Hui Zhang]

Hi Qi,

With the current implementation, it is the user’s responsibility to make sure the input ROI.mat matches the output FittedParams.mat.  If they are not matched by mistake, the current implementation will proceed as if they were matched.

Gary

To view this discussion on the web visit https://groups.google.com/d/msgid/noddi/18288fc8-26bc-4138-bc0e-803f66cbc787n%40googlegroups.com.