Re: [AMBER] checkpoint on Amber within batch job

From: Ross Walker <ross.rosswalker.co.uk>
Date: Thu, 23 Jan 2014 09:38:04 -0800

Simple solution here that one should ALWAYS do unless there is a specific
reason to keep using the same random seed is to set ig=-1.

I would argue that this should be the default behavior of the random
number generator if it didn't mean we'd have to revamp every single test
case.



On 1/23/14, 6:54 AM, "James Maier" <jimbo.maier.gmail.com> wrote:

>On Thu, Jan 23, 2014 at 8:57 AM, Zhuang, Yu <yu.zhuang.tum.de> wrote:
>
>> Does this means I just need to re-format the sander command for the next
>> run without modification of .in file?
>
>
>This depends on whether you are using a thermostat or barostat. Some
>thermostats and especially barostats can give you artifacts if you restart
>too quickly with the same random seed (ig flag). Setting ig differently
>for each restart helps alleviate that. But if you're running say, 20 ns
>at
>a time, it might not be a big deal.
>
>
>James
>
>P.S.
>Since you are using slurm, you may want to check out the --dependency flag
>(see man sbatch)
>_______________________________________________
>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 Thu Jan 23 2014 - 10:00:02 PST
Custom Search