Restart Problem

[bruceli999]

Dear SSAGES developers,

I have used the example SSAGES/Examples/User/ABF/Qbox-CH2Cl-CH2Cl/ for restart calculation test, but there is an error that terminates the simulation, as partially shown in the figure.Is it something wrong with the restart files or other something else? BTW, the restart calculations both in previous folder and a new folder were tested, and the problems are the same. Thanks!

Error messages

For input.json, "restart" is changed to "true"

For md.i, I added "load restart_0.xml"

Script command mpirun qb -server ssages_in_0 ssages_out_0 & mpirun /home/**/apps/SSAGES/build/ssages input.json & wait

[shijiale0609]

If you want to restart, you need to copy or move the previous ABF output files which are in the folder of [Sample_Outputs] to the location where your input.json is. They should be in the same location.

[shijiale0609]

I created a folder called [Sample_Output] to contain those ABF example output files. The reason is that the simulation will overwrite those example putout files if they are in the same location.

[bruceli999]

If you want to restart, you need to copy or move the previous ABF output files which are in the folder of [Sample_Outputs] to the location where your input.json is. They should be in the same location.

Thanks for your reply.

For restart, I have copyed the required files to a new foder, including F_out, Fworld_cv0, Nworld, Input.json (restart "true" turns on), md.i (including "load restart_0.xml" ), restart_0.xml. But there is a error shown in the above picture, which would remain the .lock file all the time. I don't how to fix this problem.

I also noted that "Iteration" in F_out could not automatically accumulate the previous value in the restart folder. So a very larger value of 3418000 iteration in example output is obtained for one simulaion without restart? If I want to get this long iteration with restart process, it means that i should multiply the number of restart by the iteration of each restart by myself?

[bruceli999]

Dear SSAGES developers,

I have summarized some questions about the restart issue of ABF method when using SSAGES coupled with Qbox. Thanks!

1. Does the number of steps that push the system to do biased MD (i.e., the calculated total time of biased MD) depend on "Run xx" in the Qbox input file or "md_iterations: xx" in the SSAGES input file. If it is the latter one. Therefore, for restart calculation, is it reasonable when the MD step in qbox input file is only set to 1? that is, “run 1 20 5”.

2. For the restart, I have copied the required files from the first calculation to a new folder, including F_out, Fworld_cv0, Nworld, Input.json (restart "true" turns on), md.i (including "load restart_0.xml" ), restart_0.xml. Is it right? Do F_out and restart_0.xml from the previous calculation have to be copied to the new restart folder? Moreover, if we don't concern about the overwriting problem of files, can we restart them in one folder during all the simulation time until obtaining the final F_out. After all, F_out file is the most important one for calculating the free energy.

3. When I restarted the calculation, I noted that the "Iteration" in F_out did not automatically accumulate the previous value. For example, I set md_iterations to 3000 in the JSON file, and the Iteration in F_out was still 3000 instead of 6000 after a restart. So the 3418,000 steps in the example are computed all at once? If I want to get this number of steps by restart process, does it mean that I should use the number of restarts multiplied by the md_iteration of each restart by myself?

4. What determines the number of ssages_out_0_run_xx.xml? The document said that it identifies the number of runs, but there are still many such files when I don't do the restart calculation. And the number of iterations seems to be independent of the value of md_iterations in input.json, because when I changed the value of md_Iteration, the number of ssages_out_0_run_xx.xml is not changed.

5. Using the example SSAGES/Examples/User/ABF/Qbox-CH2Cl-CH2Cl/, the first calculation is fine, but problems occur when carrying out the restart calculation. Based on the files shown in Q2 in a new folder, restart calculation sometimes won't run at all, or sometimes will stop after running a few steps, and then the lock file could not be automatically deleted, making the cluster idle, without output files. The error message is shown above.

[bruceli999]

Two other questions.

6. In the first calculation, Qbox can read PBE in md.i, when restart, could the Qbox still read the PBE functional, from “set xc PBE” or “load restart_0.xml”? Because I didn’t see any information about the PBE functional in the output file after restarting.

7. For the example SSAGES/Examples/User/ABF/Qbox-CH2Cl-CH2Cl/, I noted that in input.json file, the CVs is "periodic": true,” and “"CV_isperiodic": [false],” for methods, are they contradictory? If I use a periodic system with “CV type: ParticleSeparation”, how do I set these two parameters?

[shijiale0609]

''' ·1. Does the number of steps that push the system to do biased MD (i.e., the calculated total time of biased MD) depend on "Run xx" in the Qbox input file or "md_iterations: xx" in the SSAGES input file. If it is the latter one. Therefore, for restart calculation, is it reasonable when the MD step in qbox input file is only set to 1? that is, “run 1 20 5”. ''' Answer: "run 500 20 5" in the qbox input file is the unbiased MD steps.

"md_iterations: xx" in the SSAGES input file is biased MD steps.

Yes. it is OK to set “run 1 20 5” in md.i for restart. And the MD step in Qbox only will not update the bias potential. ABF method is very robust. It won't change the result.

[zheng001a]

Dear SSAGES developers, I was expecting to use the SSAGES/Examples/User/ForwardFlux/LAMMPS/Langevin for restart calculation,but I didn't find a way to restart,I wonder if it also has a restart command.