Just a note, sometimes if your equilibration isn't working on GPU the
easiest step can be to use a different CUDA executable (you should find
several different precision-levels in $AMBERHOME/bin) or to give the input
file to pmemd. Cuda is optimised for speed and can act wacky at larger
forces than those normally encountered in everyday equilibrium MD.
Another lazy fix is to insert an extra equilibration step before the
problematic one, exactly the same but with a much smaller timestep or a
much larger gamma_ln, or both.
Finally note that TI calculations with ifsc are supposed to be smooth and
stable, but they still aren't as reliably smooth as regular non-perturbed
Hamiltonians. If all else fails you might find for some systems that
certain lambda points unavoidably need a smaller timestep.
Josh Berryman
Uni Luxembourg
On Tue, 3 Dec 2019 at 23:06, Debarati DasGupta <
debarati_dasgupta.hotmail.com> wrote:
> Hi Prof Case,
>
> I am providing details of whatever setup I did for my TI absolute free
> energy calculation.
>
> My lambda windows for absolute free energy calc is as follows:
> 0.01592
> 0.08198
> 0.19331
> 0.33787
> 0.5
> 0.66213
> 0.80669
> 0.91802
> 0.98408
>
> My input files for minimization, heating and equilibration are as follows:
> Min.in
> minimisation
> &cntrl
> imin = 1, ntmin = 2,
> maxcyc = 10000,
> ntpr = 20, ntwe = 20,
> ntb = 1,
> ntr = 1, restraint_wt = 200.00,
> restraintmask='!:WAT & !.H=',
>
> icfe = 1, ifsc = 1, clambda = 0.01592, scalpha = 0.5, scbeta = 12.0,
> logdvdl = 0,
> timask1 = ':C3N', timask2 = '',
> scmask1 = ':C3N', scmask2 = ''
> /
> &ewald
> /
>
> Heat1.in
> heating
> &cntrl
> imin = 0, nstlim = 10000, irest = 0, ntx = 1, dt = 0.001,
> nmropt = 1,
> ntt = 1, temp0 = 100.0, tempi = 50.0, tautp = 1.0,
> ntb = 1, tol = 0.00001,
> ntc = 2, ntf = 1,
> ioutfm = 1, iwrap = 1,
> ntwe = 1000, ntwx = 1000, ntpr = 1000, ntwr = 5000,
>
> ntr = 1, restraint_wt = 200.00,
> restraintmask='!:WAT & !.H=',
>
> icfe = 1, ifsc = 1, clambda = 0.01592, scalpha = 0.5, scbeta = 12.0,
> logdvdl = 0,
> timask1 = ':C3N', timask2 = '',
> scmask1 = ':C3N', scmask2 = ''
> /
> &ewald
> /
>
> &wt
> type='TEMP0',
> istep1 = 0, istep2 = 8000,
> value1 = 50.0, value2 = 100.0
> /
>
> &wt type = 'END'
> /
> Heat2.in
> heating
> &cntrl
> imin = 0, nstlim = 10000, irest = 0, ntx = 1, dt = 0.001,
> nmropt = 1,
> ntt = 1, temp0 = 200.0, tempi = 100.0, tautp = 1.0,
> ntb = 1, tol = 0.00001,
> ntc = 2, ntf = 1,
> ioutfm = 1, iwrap = 1,
> ntwe = 1000, ntwx = 1000, ntpr = 1000, ntwr = 5000,
>
> ntr = 1, restraint_wt = 200.00,
> restraintmask=':287',
>
> icfe = 1, ifsc = 1, clambda = 0.01592, scalpha = 0.5, scbeta = 12.0,
> logdvdl = 0,
> timask1 = ':C3N', timask2 = '',
> scmask1 = ':C3N', scmask2 = ''
> /
> &ewald
> /
>
> &wt
> type='TEMP0',
> istep1 = 0, istep2 = 8000,
> value1 = 100.0, value2 = 200.0
> /
>
> &wt type = 'END'
> /
> Heat3.in
> heating
> &cntrl
> imin = 0, nstlim = 10000, irest = 0, ntx = 1, dt = 0.001,
> nmropt = 1,
> ntt = 1, temp0 = 300.0, tempi = 200.0, tautp = 1.0,
> ntb = 1, tol = 0.00001,
> ntc = 2, ntf = 1,
> ioutfm = 1, iwrap = 1,
> ntwe = 1000, ntwx = 1000, ntpr = 1000, ntwr = 5000,
>
> ntr = 1, restraint_wt = 200.00,
> restraintmask=':287',
>
> icfe = 1, ifsc = 1, clambda = 0.01592, scalpha = 0.5, scbeta = 12.0,
> logdvdl = 0,
> timask1 = ':C3N', timask2 = '',
> scmask1 = ':C3N', scmask2 = ''
> /
> &ewald
> /
>
> &wt
> type='TEMP0',
> istep1 = 0, istep2 = 8000,
> value1 = 200.0, value2 = 300.0
> /
>
> &wt type = 'END'
> /
> Equil1A.in
> &cntrl
> imin = 0, nstlim = 1000000, dt = 0.001,
> irest = 1, ntx = 5, ig = -1,
> tempi = 300.0, temp0 = 300.0,
> ntc = 2, ntf = 1, tol = 0.00001,
> taup = 0.1,
> ntwx = 10000, ntwe = 0, ntwr = 1000, ntpr = 1000,
> cut = 10.0, iwrap = 1,
> ntt =3, ntb = 2, ntp = 1, gamma_ln=1.0,
> nscm = 0,
> ntr=1, restraintmask=':287', restraint_wt=100,
> icfe = 1, clambda=0.01592, ifsc=1,
> timask1=':C3N',timask2='',
> scmask1 = ':C3N', scmask2 = ''
> /
> Equil1B.in
> &cntrl
> imin = 0, nstlim = 1000000, dt = 0.002,
> irest = 1, ntx = 5, ig = -1,
> tempi = 300.0, temp0 = 300.0,
> ntc = 2, ntf = 1, tol = 0.00001,
> taup = 0.1,
> ntwx = 10, ntwe = 0, ntwr = 10, ntpr = 10,
> cut = 10.0, iwrap = 1,
> ntt =3, ntb = 2, ntp = 1, gamma_ln=1.0,
> nscm = 0,
> ntr=1, restraintmask=':287', restraint_wt=100,
> icfe = 1, clambda=0.01592, ifsc=1,
> timask1 = ':C3N',timask2 = '',
> scmask1 = ':C3N', scmask2 = ''
> /
>
> My protein is a 160 residue kinase+ my ligand is a small molecule( 10atoms)
>
> Everytime I do run the last step of equlibriation equil1B.in I get this
> error,
> Hostname = grape1
> Date = Mon Dec 2 17:06:33 EST 2019
> SLURM_GPUFILE ...
> gpuNum=
> cudaMemcpy GpuBuffer::Download failed an illegal memory access was
> encountered
>
>
> Any help will be really useful.
> I have provided all the steps I did, this is only for my lambda 0.01592
> I need to sort it out before running other lambdas as I need the rst files
> from 0.01592 to run at higher lambdas.
> Any help ?
> Thanks
>
> ________________________________________________________________________________________________________________________________________________________________________
> From: David Case<mailto:david.case.rutgers.edu>
> Sent: Friday, November 29, 2019 9:10 PM
> To: AMBER Mailing List<mailto:amber.ambermd.org>
> Subject: Re: [AMBER] help on this error
>
> On Fri, Nov 29, 2019, Debarati DasGupta wrote:
>
> >vlimit exceeded for step 0; vmax = 36.5121
>
> The fact that you get this error on step 0 suggests that something has
> changed from the previous run, that is changing the energy and
> gradients.
>
> Try a "short" (10 step run) with ntpr=1. Compare the energies on step 1
> with the final energies from the previous runs. Try to figure out what
> is changing.
>
> >In my equlibriation steps in TI, in the final step I always am getting
> >this error.
>
> The problem is that we don't know what you mean by the "final step".
> (There is no single way to carry out equilibration.)
> You don't have to tell us: try to figure out what is different between
> the steps that "work" and the one that gives the error you cite above.
>
> ...good luck...dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Dec 06 2019 - 03:30:01 PST
