[gmx-users] mdrun with append option
Mark Abraham
Mark.Abraham at anu.edu.au
Wed Feb 2 00:48:50 CET 2011
On 2/02/2011 3:38 AM, Sai Pooja wrote:
> From the website:
> "If you change the integrator or ensemble, you should pass the
> checkpoint file to tpbconv only, not to mdrun, since the state might
> change and thus output files can not be appended."
Where was that? It could use clarification.
Mark
> So now instead of supplying the checkpoint file to mdrun I supply it
> to tpbconv... does this assure that the simulations start from the
> coordinates/velocities specified by the .cpt file?
> Pooja
>
> On Tue, Feb 1, 2011 at 11:20 AM, Sai Pooja <saipooja at gmail.com
> <mailto:saipooja at gmail.com>> wrote:
>
> Thanks Mark. So if the simulation doesn't start from the
> checkpoint file, from where are the initial coordinates velocities
> etc. taken from?... the trajectory files?
> Also, I could not find the environment variable... and I am not
> sure how to use one.
> Pooja
> On Tue, Feb 1, 2011 at 3:03 AM, Mark Abraham
> <Mark.Abraham at anu.edu.au <mailto:Mark.Abraham at anu.edu.au>> wrote:
>
> On 1/02/2011 7:50 AM, Sai Pooja wrote:
>> I think I have figured out the reason. It is because I am
>> carrying out replica exchange (manual) after every mdrun. If
>> the exchange occurs, I exchange the checkpoint files, extend
>> the simulation by 500 steps and continue. The new simulation
>> starts from exchanged cptfile. It seems that whenever the
>> exchange occurs, the earlier log,traj files are not appended.
>> They are instead overwritten. the obv solution is to save and
>> index these files with the relevant replicas everytime an
>> exchange occurs.
>
> This would have been good to know earlier. If replica-exchange
> leads to the ensemble of the .tpr not matching the ensemble of
> the .cpt, then IIRC 4.5.3 mdrun will refuse to start from the
> .cpt, which means the subsequent mdrun will start from the
> .tpr only. Certainly a non-appending mdrun prints a warning
> (or error, I forget which) message to the log file, but
> perhaps the use of -append (erroneously) doesn't do that.
> Please have a look and see if that is the issue. There is an
> environment variable that can be set to tell mdrun that you
> (think you) know what you are doing mismatching .tpr and .cpt.
>
> Mark
>
>
>> However, i have a more general question. Since mdrun still
>> runs with the exchanged checkpoint files and starts from the
>> point where the previous run ended, can I be assured that an
>> exchange has been affected - since tpr files correspond to
>> the "replica-box" and cpi to the most recent exchanged replicas?
>> Pooja
>>
>> On Mon, Jan 31, 2011 at 2:33 PM, Sai Pooja
>> <saipooja at gmail.com <mailto:saipooja at gmail.com>> wrote:
>>
>>
>> I manually index checkpoint files after every mdrun. What
>> troubles me is the randomness with which -append fails/works.
>> For eg, I have a simulation which runs from 3ns, 1ps in 1
>> mdrun. Now oddly enough, the logfile starts from
>> 1184ps(in the end, I do remember the one starting from 0
>> but that was overwritten it seems) and the rest is
>> appended uptil the 3000ps step. Why would append work
>> from 1184ps to 3000ps but not for the previous ones?Could
>> it have anything to do with the network/cluster? If that
>> is the case is it safer to create a new file everytime
>> and then concatenate them after say every 100ps?
>> Pooja
>> On Sat, Jan 29, 2011 at 6:52 PM, Mark Abraham
>> <Mark.Abraham at anu.edu.au
>> <mailto:Mark.Abraham at anu.edu.au>> wrote:
>>
>> On 30/01/2011 10:39 AM, Sai Pooja wrote:
>>> I would be happy to supply more information.. if
>>> someone could please look into this.. otherwise I
>>> will have to switch to storing every file and then
>>> just concatenating them which seems like a rather
>>> roundabout way of doing it.
>>
>> As I suggested a few emails ago, are you sure that
>> -cpi file exists? If your numerical suffixes are
>> indexing restarts, then unless you've done some
>> manual copying that you haven't told us about, it
>> won't. Your filename scheme seems a bit contorted -
>> like you're trying to do the work that GROMACS 4.5.x
>> will just do for you if you let it.
>>
>> Otherwise, you'll have to do some detective work with
>> gmxcheck on the -cpi to see what might be the issue.
>>
>> In your case, an initial
>>
>> mdrun -deffnm rex_3
>>
>> (perhaps save some copies while you're experimenting)
>> and subsequently
>>
>> tpbconv -extend <blah> -f rex_3 -o rex_3
>> mdrun -deffnm rex_3 -append
>>
>> will work and be much simpler than whatever you're
>> trying to do with filenames :-)
>>
>> Mark
>>
>>
>>>
>>> On Fri, Jan 28, 2011 at 4:37 PM, Sai Pooja
>>> <saipooja at gmail.com <mailto:saipooja at gmail.com>> wrote:
>>>
>>> This is the command:
>>> nbs submit -command
>>> "(/usr/local/gromacs/4.5.1/bin/mdrun_mpi -s
>>> rex_3.tpr -e rex_3 -c after_rex_3 -cpi restart3
>>> -cpo restart3 -ap
>>> pend -g rexlog3 -x rextraj3);" -nproc 1 -name
>>> "GENHAM-DIHEDRAL-3" -mail start end
>>> Pooja
>>>
>>> On Fri, Jan 28, 2011 at 4:20 PM, Mark Abraham
>>> <Mark.Abraham at anu.edu.au
>>> <mailto:Mark.Abraham at anu.edu.au>> wrote:
>>>
>>> On 29/01/2011 3:56 AM, Sai Pooja wrote:
>>>
>>> Hi,
>>> I am using tpbconv and mdrun to extend a
>>> simulation. I use it with the append
>>> option but the files still get
>>> overwritten or erased. Can someone help
>>> me in this regard?
>>> Pooja
>>> Commands (in python)
>>> cmd = '(%s/tpbconv -extend %f -s
>>> rex_%d.tpr -o rex_%d.tpr)'
>>> %(GROMPATH,dtstep,i,i)
>>> os.system(cmd)
>>> time.sleep(1)
>>> cmd = 'nbs submit -command "'
>>> cmd += '(%s/mdrun_mpi -noh
>>> -noversion -s rex_%d.tpr -e rex_%d -c
>>> after_rex_%d -cpi restart%d -cpo
>>> restart%d -append -g rexlog%d -x
>>> rextraj%d >/dev/null); '
>>> %(GROMPATH,i,i,i,i,i,i,i)
>>> cmd += '" '
>>> cmd += '-nproc 1 '
>>> cmd += '-name "GENHAM-DIHEDRAL-%d"
>>> '%(i)
>>> cmd += '-mail start end '
>>> cmd += '-elapsed_limit 16h >> rexid'
>>> os.system(cmd)
>>>
>>>
>>> More useful for diagnostic and
>>> record-preservation purposes is to construct
>>> the cmd string and print it to stdout (or
>>> something).
>>>
>>> At the moment it is far from clear that your
>>> -cpi file exists for the new run.
>>>
>>> Mark
>>> --
>>> gmx-users mailing list gmx-users at gromacs.org
>>> <mailto:gmx-users at gromacs.org>
>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>> Please search the archive at
>>> http://www.gromacs.org/Support/Mailing_Lists/Search
>>> before posting!
>>> Please don't post (un)subscribe requests to
>>> the list. Use the www interface or send it
>>> to gmx-users-request at gromacs.org
>>> <mailto:gmx-users-request at gromacs.org>.
>>> Can't post? Read
>>> http://www.gromacs.org/Support/Mailing_Lists
>>>
>>>
>>>
>>>
>>> --
>>> Quaerendo Invenietis-Seek and you shall discover.
>>>
>>>
>>>
>>>
>>> --
>>> Quaerendo Invenietis-Seek and you shall discover.
>>
>>
>> --
>>
>> gmx-users mailing list gmx-users at gromacs.org
>> <mailto:gmx-users at gromacs.org>
>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/Search
>> before posting!
>> Please don't post (un)subscribe requests to the list.
>> Use the
>> www interface or send it to
>> gmx-users-request at gromacs.org
>> <mailto:gmx-users-request at gromacs.org>.
>> Can't post? Read
>> http://www.gromacs.org/Support/Mailing_Lists
>>
>>
>>
>>
>> --
>> Quaerendo Invenietis-Seek and you shall discover.
>>
>>
>>
>>
>> --
>> Quaerendo Invenietis-Seek and you shall discover.
>
>
> --
>
> gmx-users mailing list gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org>
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before
> posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org
> <mailto:gmx-users-request at gromacs.org>.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
>
>
>
> --
> Quaerendo Invenietis-Seek and you shall discover.
>
>
>
>
> --
> Quaerendo Invenietis-Seek and you shall discover.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110202/9720cfdf/attachment.html>
More information about the gromacs.org_gmx-users
mailing list