Restart HSE-hybrid calculation

Queries about input and output files, running specific calculations, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
User avatar
SKM
Full Member
Full Member
Posts: 125
Joined: Wed Oct 30, 2019 5:39 am
License Nr.: 5-516

Restart HSE-hybrid calculation

#1 Post by SKM » Thu Sep 03, 2020 4:20 am

Hi Admin,

can the HSE Run be continued form where it stopped after completing previous wall-time? if so how to do?
Regards
Regards
SKM

henrique_miranda
Global Moderator
Global Moderator
Posts: 502
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Restart HSE-hybrid calculation

#2 Post by henrique_miranda » Fri Oct 02, 2020 10:10 am

No, if the calcualation hit the walltime them most likely the WAVECAR was not written so you cannot restart from where the calculations stoped.
You will have to start from scratch.
This applies also to DFT or scGW calculations.

User avatar
SKM
Full Member
Full Member
Posts: 125
Joined: Wed Oct 30, 2019 5:39 am
License Nr.: 5-516

Re: Restart HSE-hybrid calculation

#3 Post by SKM » Mon Oct 05, 2020 9:48 am

Hi
In such case how can we proceed with the HSE run? because as you suggested even if i start again, it would not finish within the aloowed wall time limit of 48 hrs in our systems.

So what is the way to do it?

Regards
Regards
SKM

henrique_miranda
Global Moderator
Global Moderator
Posts: 502
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Restart HSE-hybrid calculation

#4 Post by henrique_miranda » Mon Oct 05, 2020 10:22 am

You can try to set NELM to limit the number of electronic steps.
Once VASP executes those number of electronic steps it will stop the calculation and write the WAVECAR.
Then you can restart the calculation starting from this WAVECAR.

Note that this is more a fix that an adequate way of running your calculations.

I would recommend that you:
1. try to fine-tune your calculation to save computational time: check that you are using an adequate paralelization for your system, that you don't have a too large energy cutoff, too many k-points or a very strict convergence criteria.
2. check with your HPC center for alternatives to run longer jobs.

Post Reply