From: Aron Broom (broomsday_at_gmail.com)
Date: Mon Oct 01 2012 - 15:50:29 CDT
Hi Giacomo,
Since a water dimer is pretty quick to simulate, I ended up running a
1000ns simulation dropping a hill every 1000fs, so I guess in total that
was 1 million hills. When I first did it there was a slowdown in the
simulation over time because some of the hills were close to the boundary
and being written to the state file rather than the grids. After taking
your advice and expanding the colvar boundary well beyond the colvar wall,
that no longer happened and everything was great.
As I say, the shape and depth of the PMF matched with hillWidths of 0.1 up
to 1.0 (accounting for the lack of smoothness in those others) and also
matched perfectly with Umbrella sampling.
~Aron
On Mon, Oct 1, 2012 at 4:44 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com>wrote:
> Hello Aron, no, it is not much of an issue in the end, I also ran some
> tests for different grid spacings and found no detectable errors, other
> than those originating from the finite size of the hills themselves:
> whether they are analytically derived or interpolated on a grid it doesn't
> matter.
>
> By curiosity, how many hills did you accumulate with hillWidth = 0.05?
>
>
> On Mon, Oct 1, 2012 at 4:39 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>
>> Glad to hear it worked. The grids certainly help over the course of a
>> long simulation with many hills also. The one concern here is that the
>> manual mentions that with useGrids on the colvar width should ideally not
>> be less then 1.0 to avoid problems discretizing the potential onto the
>> grids. Since your width is 0.1 that might be an issue. That being said,
>> I've done some simple test cases with a water dimer and a distance colvar
>> with a width of 0.05 and the results are identical in the long to a larger
>> width (though obviously smoother) so perhaps it isn't an issue afterall?
>>
>> ~Aron
>>
>>
>> On Mon, Oct 1, 2012 at 4:24 PM, Jeff Wereszczynski <
>> jmweresz_at_mccammon.ucsd.edu> wrote:
>>
>>> Well that IS funny! I had turned off the usegrids option because I
>>> had assumed it wouldn't work based on this previous mailing list
>>> entry:
>>>
>>> http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l/11810.html
>>>
>>> However, if I put it back on then it works fine. Apparently the use
>>> of grids in multiple replica metadynamics has been enabled? Still, it
>>> probably shouldn't crash if you don't use grids.
>>>
>>> Thanks!
>>> Jeff
>>>
>>> On Mon, Oct 1, 2012 at 1:13 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>>> > That is strange, I have multiple replicas MetaDynamics working in 2.9.
>>> I've
>>> > never tried without useGrids, however. I assume the useGrids is off
>>> because
>>> > of the smaller increment size of the colvar?
>>> >
>>> > ~Aron
>>> >
>>> >
>>> > On Mon, Oct 1, 2012 at 3:52 PM, Jeff Wereszczynski
>>> > <jmweresz_at_mccammon.ucsd.edu> wrote:
>>> >>
>>> >> Hi Giacomo,
>>> >>
>>> >> I have tried leaving out that flag all together, in fact thats how I
>>> >> originally had it set up, and I still get that error.
>>> >>
>>> >> Also, I am using version 2.9. I considered trying the CVS, but a
>>> >> quick look at the online change logs made it seem as if nothing in the
>>> >> colvars routines had been touched since the 2.9 release. I could
>>> >> always give that a try though...
>>> >>
>>> >> Thanks,
>>> >> Jeff
>>> >>
>>> >> On Mon, Oct 1, 2012 at 12:21 PM, Giacomo Fiorin
>>> >> <giacomo.fiorin_at_gmail.com> wrote:
>>> >> > Hello Jeff, and thanks first of all for researching the causes of
>>> the
>>> >> > error
>>> >> > plenty!
>>> >> >
>>> >> > expandBoundaries is off by default: have you tried running without
>>> the
>>> >> > corresponding line altogether?
>>> >> >
>>> >> > Most importantly: which version of NAMD are you using? For this
>>> >> > particular
>>> >> > feature I strongly recommend 2.9.
>>> >> >
>>> >> > Giacomo
>>> >> >
>>> >> >
>>> >> > On Mon, Oct 1, 2012 at 2:52 PM, Jeff Wereszczynski
>>> >> > <jmweresz_at_mccammon.ucsd.edu> wrote:
>>> >> >>
>>> >> >> Hello,
>>> >> >>
>>> >> >> I'm trying to run multiple walker metadynamics, however I keep
>>> getting
>>> >> >> the
>>> >> >> following error:
>>> >> >>
>>> >> >> colvars: Error: expandBoundaries is not supported when using more
>>> >> >> than
>>> >> >> one replicas; please allocate wide enough boundaries for each
>>> >> >> colvarahead of
>>> >> >> time.
>>> >> >>
>>> >> >> While this sounds straightforward, I've explicitly turned off both
>>> >> >> expandBoundaries and useGrids in my colvars file. Here is the
>>> colvars
>>> >> >> file:
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >>
>>> -----------------------------------------------------------------------------------
>>> >> >> colvar {
>>> >> >> name rgyr
>>> >> >> expandBoundaries off
>>> >> >> width .1
>>> >> >> lowerboundary 0
>>> >> >> upperboundary 15
>>> >> >> lowerwallconstant 1
>>> >> >> upperwallconstant 1
>>> >> >> gyration {
>>> >> >> atoms {
>>> >> >> atomsFile rgyr.pdb
>>> >> >> atomsCol O
>>> >> >> }
>>> >> >> }
>>> >> >> }
>>> >> >> metadynamics {
>>> >> >> name metad
>>> >> >> colvars rgyr
>>> >> >> hillWeight .2
>>> >> >> hillWidth 10
>>> >> >> newHillFrequency 100
>>> >> >> usegrids false
>>> >> >>
>>> >> >> multipleReplicas on
>>> >> >> replicaID 01
>>> >> >> replicasRegistry replicas
>>> >> >>
>>> >> >> }
>>> >> >>
>>> >> >>
>>> >> >>
>>> -----------------------------------------------------------------------------------
>>> >> >>
>>> >> >> And here is the relevant output in the NAMD log file:
>>> >> >>
>>> >> >> colvars:
>>> >> >>
>>> ----------------------------------------------------------------------
>>> >> >> colvars: Initializing a new collective variable.
>>> >> >> colvars: # name = rgyr
>>> >> >> colvars: Initializing a new "gyration" component.
>>> >> >> colvars: # componentCoeff = 1 [default]
>>> >> >> colvars: # componentExp = 1 [default]
>>> >> >> colvars: # period = 0 [default]
>>> >> >> colvars: # wrapAround = 0 [default]
>>> >> >> colvars: Initializing atom group "atoms".
>>> >> >> colvars: Atom group "atoms" defined, 21 initialized: total
>>> mass =
>>> >> >> 264.19.
>>> >> >> colvars: All components initialized.
>>> >> >> colvars: # width = 0.1
>>> >> >> colvars: # lowerBoundary = 0
>>> >> >> colvars: # lowerWallConstant = 1
>>> >> >> colvars: # lowerWall = 0 [default]
>>> >> >> colvars: # upperBoundary = 15
>>> >> >> colvars: # upperWallConstant = 1
>>> >> >> colvars: # upperWall = 15 [default]
>>> >> >> colvars: # expandBoundaries = off
>>> >> >> colvars: # extendedLagrangian = off [default]
>>> >> >> colvars: # outputValue = on [default]
>>> >> >> colvars: # outputVelocity = off [default]
>>> >> >> colvars: # outputSystemForce = off [default]
>>> >> >> colvars: # outputAppliedForce = off [default]
>>> >> >> colvars:
>>> >> >>
>>> ----------------------------------------------------------------------
>>> >> >> colvars: Collective variables initialized, 1 in total.
>>> >> >> colvars:
>>> >> >>
>>> ----------------------------------------------------------------------
>>> >> >> colvars: Initializing a new "metadynamics" instance.
>>> >> >> colvars: # name = metad
>>> >> >> colvars: # colvars = { rgyr }
>>> >> >> colvars: # hillWeight = 0.2
>>> >> >> colvars: # newHillFrequency = 100
>>> >> >> colvars: # hillWidth = 10
>>> >> >> colvars: # useGrids = off
>>> >> >> colvars: # multipleReplicas = on
>>> >> >> colvars: Error: expandBoundaries is not supported when using more
>>> >> >> than
>>> >> >> one replicas; please allocate wide enough boundaries for each
>>> >> >> colvarahead of
>>> >> >> time.
>>> >> >> colvars: If this error message is unclear, try recompiling with
>>> >> >> -DCOLVARS_DEBUG.
>>> >> >> FATAL ERROR: Error in the collective variables module: exiting.
>>> >> >>
>>> >> >>
>>> >> >> For clarity, I've bolded the lines that say expandBoundaries is
>>> off and
>>> >> >> the error message telling me it isn't. I've tried recompiling NAMD
>>> >> >> with the
>>> >> >> DCOLVARS_DEBUG flag on, but that didn't provide any more helpful
>>> >> >> information. I also went through the source and found that this
>>> error
>>> >> >> is
>>> >> >> triggered by the expand_grids flag, but I can't figure out where
>>> that
>>> >> >> is
>>> >> >> being set in the code. Any suggestions on this would be greatly
>>> >> >> appreciated!
>>> >> >>
>>> >> >> Also, while I'm asking questions, just for my own sanity have the
>>> >> >> multiple
>>> >> >> walker + well tempered metadynamics options been tested together?
>>> I
>>> >> >> assume
>>> >> >> there are no major issues I should look out for?
>>> >> >>
>>> >> >> Thanks,
>>> >> >> --
>>> >> >> Jeff Wereszczynski
>>> >> >> Postdoctoral Scholar
>>> >> >> University of California, San Diego
>>> >> >> http://mccammon.ucsd.edu/~jwereszc
>>> >> >
>>> >> >
>>> >>
>>> >>
>>> >>
>>> >> --
>>> >> Jeff Wereszczynski
>>> >> Postdoctoral Scholar
>>> >> University of California, San Diego
>>> >> http://mccammon.ucsd.edu/~jwereszc
>>> >>
>>> >
>>> >
>>> >
>>> > --
>>> > Aron Broom M.Sc
>>> > PhD Student
>>> > Department of Chemistry
>>> > University of Waterloo
>>> >
>>>
>>>
>>>
>>> --
>>> Jeff Wereszczynski
>>> Postdoctoral Scholar
>>> University of California, San Diego
>>> http://mccammon.ucsd.edu/~jwereszc
>>>
>>
>>
>>
>> --
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
>>
>>
>
-- Aron Broom M.Sc PhD Student Department of Chemistry University of Waterloo
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:38 CST