Hi Heidi,
This indeed does not sound normal - are you sure all your input data is OK, e.g. there are no NANs in your data etc.
What you are emailing is all very inconsistent. A previous analysis seems to have failed at the mixture modellling stage, right? In this case there will be output files prior to this. Amongst other things you should have a melodic_IC.nii.gz file (or similar, depending on your output type). Now you report that nothing happens for hours, presumably not even a new output directory is being generated? I'm afraid without you being more acurate in your descriptions and error reporting (e.g. by sending the log file if exists) there is little we can do.
can you do a
ls -ltR | grep melodic_IC\*' in a terminal to see what exists and what doesn't
hth
Christian
On 21 Feb 2013, at 09:16, "Jacobs H (NP)" <[log in to unmask]> wrote:
> Hi Matthew,
>
> Thanks for your reply. I started melodic again and will let it run.
> Is it normal that no files are produced during this time? I was checking the output folder of melodic, but no files were added and nothing was added to the log file.
>
> I will let you know what happened.
>
> Thanks!
> Heidi
> ================================================
> Dr. Heidi Jacobs
> Postdoc researcher
> Faculty of Health, Medicine and Life Sciences
> School for Mental Health and Neurosciences
> Division Cognitive Neuropsychiatry and Clinical Neurosciences
> Alzheimer Center Limburg
> [log in to unmask]
> www.maastrichtuniversity.nl
> www.heidijacobs.nl
>
> Dr. Tanslaan 12, 6229 ET Maastricht
> P.O. Box 616, 6200 MD Maastricht, The Netherlands
> T +31 43 38 84 090 F +31 43 38 84 092
> ================================================
> ________________________________________
> Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Matthew Webster [[log in to unmask]]
> Verzonden: woensdag 20 februari 2013 17:03
> Aan: [log in to unmask]
> Onderwerp: Re: [FSL] Melodic crashes from gui and command line
>
> Hello Heidi,
> For a data set as large as yours a 35 hour+ runtime is not implausible - can you rerun the command line call that did _not_ use --report and leave it running? It might also be worth monitoring the jobs memory usage ( e.g. with top ) to make sure it does not exceed 12Gb.
>
> Kind Regards
>
> Matthew
>> Dear Christian,
>>
>> Thanks for your feedback. I tried to run melodic from the command line as you suggested.
>> As no report was generated by firefox, I ran it without the -report.
>> I aborted the computer after 35 hours, as nothing was happening (even though the CPU was working for about 96%, seems like a crash). Melodic generated no file.
>> To make sure it is not the code generating the figures, as you thought, I ran 4 subjects (2 by 2) and that was no problem. In less than 3 hours, Melodic finished successfully.
>> I did analyses on 80 out of these 120 people, and that was neither a problem. Could it be that 120 people is too much for melodic?
>> I don't think my computer is not powerful enough (intel i7, 3.8Ghz, 12GB memory, quad core) and I did run everything within Centos.
>>
>> Is it normal that it would take this long? Or is something else going on? I am desperate now, I have no idea what I could do?
>> Any help or advice is welcome.
>> Thanks!
>> Heidi
>> ________________________________________
>> Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Christian F. Beckmann [[log in to unmask]]
>> Verzonden: dinsdag 12 februari 2013 22:57
>> Aan: [log in to unmask]
>> Onderwerp: Re: [FSL] 120 subjects Melodic: only one component?
>>
>> Hi Heidi,
>>
>> I think you're reporting a whole set of unrelated things here. It looks like Melodic finishes and the mixture modelling does crash on your data. Can you do the following:
>> (i) run melodic from the command line for the mixture modelling only:
>>
>> melodic -i bgimage --ICs=melodic_IC --mix=melodic_mix -v --report --Ostats
>>
>> and report on what you're getting.
>>
>> If that crashes then please run the same without the --report - if that works then the problem is in the code generating the figures, not anything to do with data size or number of subjects.
>>
>> hth
>> Christian
>>
>>
>>
>>
>>
>>
>> On 12 Feb 2013, at 22:46, "Jacobs H (NP)" <[log in to unmask]> wrote:
>>
>>> Hi all,
>>>
>>> I have been trying to run melodic on 120 subjects. However, melodic seems to stop after giving me one component in the firefox output (mixture modelling). The melodic_IC.nii.gz is complete, showing all 25 components. This was done on the latest version of FSL (5.2) on a Linux machine (Centos 6.3). FSL ran for about 16 hours and then stopped.
>>> I thought I might have been the BBR registration (as I never used that before), but after a 7DOF registration, I got the same problem.
>>>
>>> So, then I tried to run the data on my Macbook pro with FSL version 4.1.9. FSL stopped during the PCA (total run time 25 hours).
>>>
>>> In earlier analyses, I was able to do the melodic on my Macbook pro (80 and 40 subjects) without problems.
>>>
>>> Could it be that 120 subjects it too much for melodic? Any help is appreciated as I have no idea what to do now.
>>>
>>> Any help is appreciated!
>>> Best
>>> Heidi
>>>
>>>
>>>
>>> ================================================
>>> Dr. Heidi Jacobs
>>> Postdoc researcher
>>> Faculty of Health, Medicine and Life Sciences
>>> School for Mental Health and Neurosciences
>>> Division Cognitive Neuropsychiatry and Clinical Neurosciences
>>> Alzheimer Center Limburg
>>> [log in to unmask]
>>> www.maastrichtuniversity.nl
>>> www.heidijacobs.nl
>>>
>>> Dr. Tanslaan 12, 6229 ET Maastricht
>>> P.O. Box 616, 6200 MD Maastricht, The Netherlands
>>> T +31 43 38 84 090 F +31 43 38 84 092
>>> ================================================
>>
|