[gmx-users] Re: Continuing a simulation plus another error

[Justin Lemkul]

On 12/24/12 11:54 AM, Ankita naithani wrote:
> Hi again,
>
> Sorry for the repetition in email.
>
> When I ran the mdrun command, I got an error of
>
> "Attempting to read a checkpoint file of version 13 with code of version 12"
>
> Can anyone please help me with this error too?
>

This means you are not using the same version of the code as the previous run.

> On Mon, Dec 24, 2012 at 4:35 PM, Ankita naithani
> <ankitanaithani at gmail.com> wrote:
>> Hi,
>>
>> I am running a protein simulation for 70 ns. I had a MPI run but due
>> to my time constraints on the server, it stopped after 28ns. Now, I
>> want to continue the simulation from the same point  up until I use up
>> my server time.
>>
>> I just wanted to confirm that there are two checkpoint files written,
>> md.cpt and md_prev.cpt. It would be really helpful if anyone could
>> advice as to which file would be better to choose to continue the
>> simulation?
>>

Checkpoint files are written every 15 minutes by default and the names are 
recycled between (prefix).cpt and (prefix)_prev.cpt to indicate the most 
recently saved state and the previous one.  Either can be used for continuing 
the stopped run, but the current state is most advantageous to avoid wasted 
time.  The previous state is saved as a backup in case of a corrupted checkpoint 
file or frame in the trajectory that would require starting from a previous point.

>> Also, I wanted a confirmation that if I use:
>>
>> mdrun -s topol.tpr -cpi md.cpt -append
>>
>> Do I also need to add -deffnm md?
>>

That depends entirely upon the file names present and what the original 
invocation of mdrun was.

>> and if I run mdrun, would it then continue from say 28ns and up until
>> the time specified in the .mdp file? The reason I wanted to confirm
>> this is that before submitting it to the server, I ran it in my local
>> machine and when I see the log file, it shows step 0 and Time 0.0000,
>> does that mean it is starting the simulation from scratch because I
>> had expected it to show me step from wherever it exited last and
>> continue from there on.
>>

The run will start from scratch if for some reason the checkpoint file cannot be 
found or read for whatever reason.  If you are trying to run on a local machine 
with a different number of processors, for instance, the checkpoint state will 
not be the same so the run will start over again.

Note that most of this information is online for quick reference:

http://www.gromacs.org/Documentation/How-tos/Doing_Restarts#Version_4.x

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================