CI-NEB out of memory

[Hasan Al-Mahayni]

Hello All,

This is the first time I am running BAND calculation for Climbing image NEB. I attach the my input file and my two xyz files. I am getting a huge number of output files (639!). I don't think this is normal but I don't see what I did wrong in my input file. I only requested 5 image replicas. I gave the job a lot of memory/CPUS but it fails after 1 minute telling me its out of memory.  Can someone help me with this?

Thanks in advance,

Hasan.

[Krack Matthias (PSI)]

Hi Hasan

 

You are running a CI-NEB band calculation with 4x4x1 k points which might explain the large number of output files. I am not sure, if that combination of features has been sufficiently tested or even validated at all, since all NEB test files are Gamma point only runs.

 

Best regards

 

Matthias

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/a4ed1e09-cfb2-42ea-b3c6-0b233e08f8b3n%40googlegroups.com.

[Hasan Al-Mahayni]

Hello Matthias,

Thank you for your reply. I commented out the kpoints, I still get the out of memory error. Do I have to remove something else in my input file other than kpoints?

Hasan Al-Mahayni

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/3837FFEA-DD77-42DB-B39F-22B1D9F0D5E9%40psi.ch.

[Krack Matthias (PSI)]

Hi Hasan

 

Your system is too large to run each of the 5 frames only on one CPU core. Increase NPROC_REP from 1 to a larger value depending on the CPU resources available (e.g. 16 or better more like 64 which requires 5*64=320 cores total).

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAECRhQyRmnGV9qx-3W%3DgCkgh4yaKGRXTanxM3J8Mo3tjYM-DUw%40mail.gmail.com.

[Hasan Al-Mahayni]

Oh! Ok i see how the NPROC_REP works now. So I can put This value to 16 of there are enough cpus?

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/B5E57BA7-F487-4CAA-BE2A-586DA162E51A%40psi.ch.

--

Hasan Al-Mahayni

[Hasan Al-Mahayni]

or 64

--

Hasan Al-Mahayni

[Hasan Al-Mahayni]

Hello Matthias,

I would like to first thank you as my NEB is now working. I have two small questions now that I am running it :) 

1) The energies of my first and last image (my inputs) are different in the NEB simulation than my previous normal GEO_OPT simulations. How do people usually deal with that? Do they consider only energy differences of the NEB simulation? Because my thermodynamic energy difference between initial and final state is 0.18eV while in NEB it is 0.43 eV! 

2) One of my NEB simulations ran for 24 hours at 512 cpus and it doesn't seem to add any iterations to my overall RESULTS file. This does not happen to my other NEB simulations... So I am wondering how I can figure out if it is stuck or just taking a long time (which I doubt). For this question I attach the xyz files, input file and a screenshot of the end of the output file after 24 hours.

 

I deeply appreciate your help!

Hasan Al-Mahayni

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/B5E57BA7-F487-4CAA-BE2A-586DA162E51A%40psi.ch.

[Krack Matthias (PSI)]

Hi Hasan

 

Did you make sure that your runs, GEO_OPT and NEB., converged to the same state?

 

Best

 

Matthias

 

From: "cp...@googlegroups.com" <cp...@googlegroups.com> on behalf of Hasan Al-Mahayni <hasanal...@gmail.com>
Reply to: "cp...@googlegroups.com" <cp...@googlegroups.com>
Date: Tuesday, 16 August 2022 at 00:07
To: "cp...@googlegroups.com" <cp...@googlegroups.com>
Subject: Re: [CP2K:17489] CI-NEB out of memory

 

Hello Matthias,

 

I would like to first thank you as my NEB is now working. I have two small questions now that I am running it :) 

 

1) The energies of my first and last image (my inputs) are different in the NEB simulation than my previous normal GEO_OPT simulations. How do people usually deal with that? Do they consider only energy differences of the NEB simulation? Because my thermodynamic energy difference between initial and final state is 0.18eV while in NEB it is 0.43 eV! 

 

2) One of my NEB simulations ran for 24 hours at 512 cpus and it doesn't seem to add any iterations to my overall RESULTS file. This does not happen to my other NEB simulations... So I am wondering how I can figure out if it is stuck or just taking a long time (which I doubt). For this question I attach the xyz files, input file and a screenshot of the end of the output file after 24 hours.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAECRhQzfeQZ89gAgeFfah1uYQN8x9exfWes-0o9xmQaFWt0rTw%40mail.gmail.com.

[Hasan Al-Mahayni]

Well if I Specify the last image and it is not getting optimized during NEB I don't see how they would not converge to the same state?

Hasan Al-Mahayni

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/EC084387-8563-4344-885F-705D4E5231C2%40psi.ch.

[Krack Matthias (PSI)]

There might be other things which could interfere, e.g. the frames are aligned at the beginning by default. You have to check/compare the detailed outputs of the runs, if the energies do not agree.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAECRhQwfoZ%2Bos8ZRZXNbQGh%3D2ryBrYry%2BDFoh7Ci-z45pCfovQ%40mail.gmail.com.

[Hasan Al-Mahayni]

Sorry for the many messages,

but any idea on my point 2) above ^?

Hasan Al-Mahayni

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/EC084387-8563-4344-885F-705D4E5231C2%40psi.ch.