Errors on Machine learning force field in vasp6

Question on input files/tags, interpreting output, etc.

Please check whether the answer to your question is given in the VASP online manual or has been discussed in this forum previously!

Moderators: Global Moderator, Moderator

Post Reply
Message
Author
exk301
Newbie
Newbie
Posts: 13
Joined: Thu Jan 28, 2021 12:46 am

Errors on Machine learning force field in vasp6

#1 Post by exk301 » Thu Mar 16, 2023 7:33 pm

Hello!

I am a new user, and I need help here: I'm trying to fit the force field with ML using vasp 6.3.0 now.
For some reason, i don't have output files (ML_ABN, ML_AB) generated except for ML_logfile.
I have ML_logfile generated, however it does not give me any RMSE updated per step/force/energy/tensor. It looks like ML did not really work in INCAR file.
My job also ended with the error ' Abort(-9) on node 12 (rank 12 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, -9) - process 12' at the end now.



# machine learning
ML_LMLFF = .TRUE.
ML_ISTART = 0

#elecronic degrees of freedom
ENCUT = 325
PRECT = Normal
ALGO = Fast
LREAL = A
ISMEAR = 0
SIGMA = 0.1
EDIFF = 1E-05

#technical flags, update when changing number of cores
NELM = 100
NELMIN = 6
LCHARG = .FALSE.
LWAVE = .FALSE.
NWRITE = 0

#MD
ISYM = 0
POTIM = 2
IBRION = 0
ISIF = 3
NSW = 20000
TEBEG = 300 # start temperature
TEEND = 390 # Final temperature
SMASS = 0 #Canonical (Nose-Ahoover) thermostat

I'm testing with very simple system (Pd metal surface). I have an attachment here. I would appreciated if anyone give me advice regarding this!

Thank you very much!
You do not have the required permissions to view the files attached to this post.

henrique_miranda
Global Moderator
Global Moderator
Posts: 414
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Errors on Machine learning force field in vasp6

#2 Post by henrique_miranda » Wed Mar 22, 2023 4:53 pm

It looks like something goes wrong with your parallel job but it's hard for me to tell exactly what.
If you are running your calculation on an HPC system are you sure you requested enough time in the scheduler?

Could you try re-running your calculation with a smaller number of ionic steps (something like NSW=20) and see if the calculation completes and the machine learning files are produced?

exk301
Newbie
Newbie
Posts: 13
Joined: Thu Jan 28, 2021 12:46 am

Re: Errors on Machine learning force field in vasp6

#3 Post by exk301 » Fri Mar 24, 2023 2:41 pm

Dear Henrique,

Thanks for the reply. After I've tried several things, the problem has been resolved when I unfixed the atoms.
Earlier, my POSCAR had some fixed atoms (selective dynamics = F) and the error was gone when I did not fix any atom in my POSCAR file. I thought it shouldn't give me any error although I was fixing some atoms. Problem has been resolved, but not sure why this is the case.

henrique_miranda
Global Moderator
Global Moderator
Posts: 414
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Errors on Machine learning force field in vasp6

#4 Post by henrique_miranda » Mon Mar 27, 2023 3:17 pm

Did running the code for a smaller number of ionic steps (NSW=20) work?

exk301
Newbie
Newbie
Posts: 13
Joined: Thu Jan 28, 2021 12:46 am

Re: Errors on Machine learning force field in vasp6

#5 Post by exk301 » Tue Mar 28, 2023 3:33 pm

Hello Henrique,

I ran the vasp calculation with NSW = 20 instead of the NSW = 20000. I do not see the same error in my output file and the job safely ended after 20 steps. However, I do not see any ML files (ex. ML_ABN, ML_FFN, ML_HIS, ML_REG) except for ML_LOGFILE.

Is this because my vasp calculation did not enter training FF yet during only 20 steps, so did not produce trained force field outputs? I'm not sure.
(I have an attachment here in this reply).

I appreciate for your feedback. Thank you!
You do not have the required permissions to view the files attached to this post.

Post Reply