Hi Ludo,
This seems to be a hardware issue. It's strange that 3D goes well:
usually that takes more of your hardware... How many 2D classes are you
trying to refine? Using more threads and only one mpi process per node
may help. I would NOT use --dont_combine_weights_via_disc in your case.
If you MPI installation does not like large messages, you actually do
want to write data to disc for exchange with other nodes.
Anyway, did your binned reconstruction's resolution approach the Nyquist
frequency of the binned data? If not, don't bother unbinning your data.
Relion does a good job at interpolation until ~95% of Nyquist. So
unbinning only makes sense if your binned reconstruction's resolution is
close to 2x the binned pixel size. Otherwise, it is not worth your time.
HTH,
S
PS: I CC your message to the CCP-EM mailing list. Please ask future
questions there.
On 12/13/2013 11:20 AM, Ludovic Renault wrote:
> Hi Sjors,
>
> I have a question.
> I processed my binned data on our cluster without much problem.
> Now I was trying to process the unbinned data and the 2d classification
> job crashes at the end of the first iteration.
> Always something like:
> PBS: job killed: node 5 (node102) requested job terminate, 'EOF'
> (code 1099) - received
> SISTER_EOF attempting to communicate with sister MOM's
> mpirun: killing job...
>
> Our cluster's admin says it's I/O error and that the cluster wasn't design
> for large MPI jobs.
> I tried to increase threads number up to 4, and the dont_write_on_disc
> option but it still crashes.
> Surprsingly, I started a 3d refinement on the unbinned data just to try
> and it seems to work fine (requires less Memory I see).
> Would you have any suggestion on how to proceed?
> It's really annoying as the binned data processing worked so well and we
> have a really nice 3d map. We just want to get the best possible
> resolution now.
> Thanks so much for developping RELION is a really great software. I can't
> wait for 1.3 and an integrated gui.
> Best,
> Ludo
>
>
>
> On 12/11/2013 16:36, "Sjors Scheres" <[log in to unmask]> wrote:
>
>> Great! Thanks a lot. The curve are actually rather similar. I might post
>> these on the relion Wiki.
>> S
>>
>> On 11/12/2013 04:25 PM, Ludovic Renault wrote:
>>> Hi,
>>> Rishi from NeCen got back to me in the end.
>>> I attach the file in case it's any different than yours.
>>> Ludo
>>>
>>>
>>> NOTICE AND DISCLAIMER
>>> This e-mail (including any attachments) is intended for the above-named
>>> person(s). If you are not the intended recipient, notify the sender
>>> immediately, delete this email from your system and do not disclose or
>>> use for any purpose.
>>>
>>> We may monitor all incoming and outgoing emails in line with current
>>> legislation. We have taken steps to ensure that this email and
>>> attachments are free from any virus, but it remains your responsibility
>>> to ensure that viruses do not adversely affect you.
>>> Cancer Research UK
>>> Registered charity in England and Wales (1089464), Scotland (SC041666)
>>> and the Isle of Man (1103)
>>> A company limited by guarantee. Registered company in England and
>>> Wales (4325234) and the Isle of Man (5713F).
>>> Registered Office Address: Angel Building, 407 St John Street, London
>>> EC1V 4AD.
>> --
>> Sjors Scheres
>> MRC Laboratory of Molecular Biology
>> Francis Crick Avenue, Cambridge Biomedical Campus
>> Cambridge CB2 0QH, U.K.
>> tel: +44 (0)1223 267061
>> http://www2.mrc-lmb.cam.ac.uk/groups/scheres
>>
> NOTICE AND DISCLAIMER
> This e-mail (including any attachments) is intended for the above-named person(s). If you are not the intended recipient, notify the sender immediately, delete this email from your system and do not disclose or use for any purpose.
>
> We may monitor all incoming and outgoing emails in line with current legislation. We have taken steps to ensure that this email and attachments are free from any virus, but it remains your responsibility to ensure that viruses do not adversely affect you.
> Cancer Research UK
> Registered charity in England and Wales (1089464), Scotland (SC041666) and the Isle of Man (1103)
> A company limited by guarantee. Registered company in England and Wales (4325234) and the Isle of Man (5713F).
> Registered Office Address: Angel Building, 407 St John Street, London EC1V 4AD.
--
Sjors Scheres
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge Biomedical Campus
Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 267061
http://www2.mrc-lmb.cam.ac.uk/groups/scheres
|