I am running MD simulations with GROMACS 2018. I am planning to run 500ns but I want my trajectories and other output files in five different sets, i.e., filename XXX_100ns. , XXX_200ns., XXXX_300ns., etc.
Now initially I used and my set of output files corresponding to 100ns were perfectly generated:
gmx grompp -f md.mdp -c XXX.gro -r XXX.gro -t XXX.cpt -p topol.top -o XXX_100ns.tpr
gmx mdrun -s XXX_100ns.tpr -deffnm XXX_200ns
and for continuing the run (My job gets killed every 24hours due to HPC rules)
gmx mdrun -cpi XXX_100ns.cpt -deffnm XXX_100ns
Next in order to continue the run and get outputs as XXX_200ns. , I used:
gmx convert-tpr -s XXX_100ns.tpr -extend 100000 -o XXX_200ns.tpr
gmx mdrun -s XXX_200ns.tpr -cpi XXX_200ns.cpt -deffnm XXX_200ns -noappend
Here, I saw that the outputs I am getting have names: XXX_200ns.part0005.xtc, XXX_200ns.part0005.edr, etc
My question: 1. can I get rid of the ".part0005" from my file name?
2. Since my runs get killed every 24 hr, I need to restart it so now I use:
gmx mdrun -s XXX_200ns.tpr -cpi XXX_200ns.cpt -deffnm XXX_200ns.part0005
But this does not append the outputs to XXX_200ns.part0005. etc files rather creates new files with names XXX_200ns.part0005.part0005. etc.
Any suggestion on how to deal with this issue. I preferred the earlier GROMACS versions as dealing with new names was easier. Also, I should mentions that I need to break my trajectories into different sets because I will be running very long simulations and transferring smaller size trajectories is easier. And if I can choose the specific file names I can handle them easily.
The best way would be to increase the walltime in your submission script on HPC.
If thats no way possible .. the just try the following
http://www.gromacs.org/Documentation/How-tos/Extending_Simulations
gmx convert-tpr -s md_0_1.tpr -extend 9000 -o md_02.tpr (For extended tprs)
gmx mdrun -s md_02.tpr -0 md_0_1.trr -x md_o_1.xtc -cpi md_0_1.cpt -c md_0_2.gro -e md_0_2.edr -g md_0_1.log
This will help you get the next run from the point of last MD end at the same time creating different file of your results.
Hi Kalyani,
I was also guessing the same that is instead of using -deffnm Should I explicitly state all the output and input flags. I will test this now. Thank you and yes I cannot increase wall time in HPC (24hour is max) !!
Hi,
The command
gmx_mpi mdrun -s new.tpr -cpi old.cpt -o new.trr -x new.xtc -e new.edr -g new.log -c new.gro -cpo new.cpt -noappend
also generates files with name new.part0005.log, new.part0005.trr, etc.
So, explicitly stating the file names also does not help
To answer your question, the "part0005" is automatic if you use the -noappend flag. One way to achieve what I think you want is to output the files into different folders. You'll still get the filename problem but you could include in the script to change the filenames (mv *part0005.xtc __.xtc).
I create folders just to prevent overwriting data. In my top directory ($topdir, which I set), I keep my production.mdp ($mdp), production.tpr ($tpr), starting_structure.gro ($struc), topol.top ($topol), index.ndx ($ndx), "toppar/" directory (with my *.itps), and my "restraints/" directory. I break the trajectory into "parts" using a running count ($count) in the bash submission script.
for starting the production (when $count == 1), i use a command like this:
gmx grompp -f $topdir/$mdp -c $topdir/$struc -p $topdir/$topol -n $topdir/$ndx -r $topdir/$struc -o part$count/$tpr
gmx mdrun -s $topdir/part$count/$tpr -deffnm $topdir/part$count/production$count -cpt 100 -cpo $topdir/part$count/part$count.cpt -maxh 23.5
then i'd use a command like this when the count is greater than 1. here $prev is $count -1.
gmx mdrun -s $topdir/$tpr -deffnm $topdir/part$count/production -cpi $topdir/$prevout/part$prev -cpt 100 -cpo $topdir/part$count/part$count.cpt -maxh 23.5 -noappend
It seems you'd just add the -extend flag and be good to go. If you really don't want the part0004 label, you could just copy the contents to a new folder and remove the -noappend flag.
- Badges
- Science topic