Hi All,
I am looking to perform calculations using IBRION=6 in VASP to extract the elastic properties of a system. Due to the wall time limit of the supercomputer that I use, the job gets killed while the calculations are not done. Is there any solution for restarting the job?
Thank you.
Hi Sam,
unfortunately, as far as I am aware, the elastic properties calculation needs to be done in a single "sitting", as the individual displacements are performed in sequence and, in contrast to optimization runs, no updates to the CONTCAR are written; neither is the current displacement step recorded in any kind of "log" file that would allow for resumption of the interrupted process. The displacement series, to my knowledge, behaves as if forming a part of one SCF cycle, which cannot be restarted.
You can try contacting your HPC provider to increase the wall time limit, if possible. As IBRION = 6 already by design is faster than IBRION = 5 by considering only symmetry-inequivalent displacements, there is no real way to speed the procedure up. You may try reducing the k-mesh, but that may of course lead to convergence problems and inaccurate results.
Cheers,
Dimitri
Hello Sam,
indeed, there is no way. In the VASP forum there was recently a hint to PHONOPY: https://phonopy.github.io/phono3py/
However, it's not dealing with pertubation theory.
Hth,
Alex
Hi Sam,
I agree with Dimitri. I don't think you can solve it by restarting the job. Instead, you can definitely request HPC cluster-admin in your university to increase the wall time and I hope they will help you out. It is very common everywhere.
I faced same problem while doing LOPTICS calculations with HSE. I want to know can I restart the job or shall I rerun it
Hi Neha Kapila Sharma ,
I am also facing the same issue for LOPTICS calculations. Could you please guide me know if you have solved this already?
Thanks.
Hi Vivek, it's actually due to lack of memory. I run that job on bigger cluster and the problem was solved.
- Badges
- Science topic
- Similar topics
- Filing