Problem with VASP6

Problems running VASP: crashes, internal errors, "wrong" results.

Moderators: Global Moderator, Moderator

Post Reply
Message
Author
suyeon_ju
Newbie
Newbie
Posts: 2
Joined: Tue May 12, 2020 12:32 pm

Problem with VASP6

#1 Post by suyeon_ju » Tue Jun 02, 2020 12:24 pm

Hello

I've read in vaspwiki that on-the-fly machine learning force fields are available from VASP6.0 or higher.
(wiki/index.php/Machine_learning_force_field:_Theory)
I followed the exact tutorial from this site: wiki/index.php/Liquid_Si_-_MLFF.
However, when I run the MD simulation with vasp.6.1.0, it says
ERROR: ML_INTERFACE: Machine learning feature (ML_FF_LMLFF=.TRUE.)
is not supported in this version!

Is vasp.6.1.0 currently not supporting the machine learning part? or are there any ways to solve this?

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

ferenc_karsai
Global Moderator
Global Moderator
Posts: 126
Joined: Mon Nov 04, 2019 12:44 pm

Re: Problem with VASP6

#2 Post by ferenc_karsai » Wed Jun 03, 2020 6:19 am

Currently the machine learning code is extensively rewritten for robustness and speed.
We will distribute the updated version most likely late this year or early next year.

suyeon_ju
Newbie
Newbie
Posts: 2
Joined: Tue May 12, 2020 12:32 pm

Re: Problem with VASP6

#3 Post by suyeon_ju » Wed Jun 03, 2020 11:53 am

Thank you

alex
Hero Member
Hero Member
Posts: 516
Joined: Tue Nov 16, 2004 2:21 pm
License Nr.: 5-67
Location: Germany

Re: Problem with VASP6

#4 Post by alex » Fri Oct 16, 2020 8:41 am

Hi there,

do you have a timeline for the re-release available? I just upgraded my V5 license because of that feature...

Thanks and best regards,

alex

ferenc_karsai
Global Moderator
Global Moderator
Posts: 126
Joined: Mon Nov 04, 2019 12:44 pm

Re: Problem with VASP6

#5 Post by ferenc_karsai » Wed Oct 21, 2020 8:38 am

We still don't have a definite timeline, since we are still heavily debugging and working on making the code very robust.

We can only assure that the machine learning code will be available in future realeses.

alex
Hero Member
Hero Member
Posts: 516
Joined: Tue Nov 16, 2004 2:21 pm
License Nr.: 5-67
Location: Germany

Re: Problem with VASP6

#6 Post by alex » Thu Oct 22, 2020 7:08 am

Thank you Ferenc.

Could anybody point me to some working tools for the generation/development of forcefields from e.g. MD simulations, please? I'm interested in diffusion in transition metal oxides.

Thanks again,

alex

john_low1
Newbie
Newbie
Posts: 17
Joined: Wed Nov 06, 2019 3:12 pm

Re: Problem with VASP6

#7 Post by john_low1 » Wed Oct 28, 2020 12:44 pm

Alex,

We have been getting good results for alloys with DeepMDkit. http://www.deepmd.org/

I can't say that it is easy to build or use DeepMDkit, but it works once it is installed correctly. It has a nice ML algorithm for fitting.

If you try it, let me know if it works for you. It would be good to compare notes.

John

pjerabek
Newbie
Newbie
Posts: 2
Joined: Mon Jan 06, 2020 8:19 am

Re: Problem with VASP6

#8 Post by pjerabek » Wed Apr 14, 2021 1:10 pm

Dear VASP Team,

I have recently installed VASP 6.2.0 on our cluster and found that ML is not yet integrated.

Could you give an update regarding machine-learning integration in VASP? (weeks, months?) :)

Thank you and best wishes,

--- Paul

ferenc_karsai
Global Moderator
Global Moderator
Posts: 126
Joined: Mon Nov 04, 2019 12:44 pm

Re: Problem with VASP6

#9 Post by ferenc_karsai » Thu Apr 15, 2021 6:07 am

Indeed machine learning is still not included in the latest VASP. There will definitely be a clearly noticable announcement when it will be included.

We aim to include it within this year and it looks very positive that it will happen.

pjerabek
Newbie
Newbie
Posts: 2
Joined: Mon Jan 06, 2020 8:19 am

Re: Problem with VASP6

#10 Post by pjerabek » Thu Apr 15, 2021 3:03 pm

Thank you very much for the update, Ferenc!

I'll keep an eye out then. :)

Best,

-- Paul

Post Reply