Re: [AMBER] Failure to restart H-REMD

From: Jason Swails <jason.swails.gmail.com>
Date: Wed, 10 Dec 2014 07:11:09 -0500

On Wed, Dec 10, 2014 at 4:57 AM, Carlo Guardiani <
carlo.guardiani.dsf.unica.it> wrote:

> Dear Amber Developers,
> I am running H-REMD simulations of a peptide using the
> CUDA version of Amber14. I had no problems running the
> first 100 ns of the simulation, but when I tried to
> restart it, the run aborted with the error message:
>
> | ERROR: I could not understand line 14935
> 0.0061866 0.2322136 -0.1654300 0.5206336 0.6339461 0.2485021
>

​I suspect the real problem is elsewhere in your inpcrd file. Can you look
through and see if there are any ****s in your restart file that would
indicate overflowing coordinates?

You didn't set iwrap=1, which suggests that this is a strong possibility.



>
> Since in my system I have 29864 atoms and the restart file
> shows the coordinates of two atoms per line, line 14935 is
> the line where the velocities begin (considering also the
> two lines before the coordinates). The program thus behaves
> as if it doesn't understabd that it must perform a restart,
> so I am wondering if there is some special instruction to
> restart H-REMD.
>
> Here is the content of my input file (for example file
> hremd.mdin.001):
>
> H-REMD
> &cntrl
> irest=1, ntx=5,
> nstlim=1250, dt=0.002,
> ntt=3, gamma_ln=1.0,
> tempi=300.0, temp0=300.0,
> ig=578,
> ntc=2, ntf=2, nscm=1000,
> ntp=0, ntb=1, cut=8.0,
> ntpr=1250, ntwx=1250, ntwr=1250,
> numexchg=40000,
> /
>

​A few unsolicited suggest​ions regarding your input file: set ig=-1 to
seed the random generator from the system clock. This ensures that you
*always* have a different random seed for each replica for each restart.
If you have the same seed for different replicas and different restarts
with Langevin dynamics, your trajectories will begin to synchronize (which
is a bad thing).

Also, set ntxo=2 to write NetCDF restart files -- this is not vulnerable to
the errors you (probably) got above. You should also (always!) set
ioutfm=1 to write NetCDF trajectories -- they are *much* higher precision
(full single precision for each coordinate), *much* faster to write, and
the resulting files are *much* smaller than if you use the standard ASCII
trajectory files.

For more complicated REMD simulations (e.g., multi-dimensional replica
exchange), NetCDF restarts and trajectories are actually required.

HTH,
Jason

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Dec 10 2014 - 04:30:03 PST
Custom Search