[QE-users] Problem with Parallelization of Hp.x

Iurii TIMROV iurii.timrov at epfl.ch
Fri Dec 2 19:23:43 CET 2022 

> In regards to the DFT+U+V I was following this line from the paper "for the sake of activating the Hubbard-related machinery (in the DFT+U+V case, there is no need to initialize V, instead initialize U for O(2p)).” Does this refer to something else?

Actually this is not correct, sorry! One needs to specify at least one V to activate the DFT+U+V machinery.

> If I output each HP runs to different output directories, how would I recombine the results at the end?

You need to copy all files back to the same temporary folder and run the last step with compute_hp=.true. (Please check the exemples in the HP folder of Quantum ESPRESSO).

HTH

Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: Jamie Holber <holber at umich.edu>
Sent: Friday, December 2, 2022 5:58:39 PM
To: Iurii TIMROV
Cc: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] Problem with Parallelization of Hp.x

Hi Iurii,

Thanks for the quick resolution!

In regards to the DFT+U+V I was following this line from the paper "for the sake of activating the Hubbard-related machinery (in the DFT+U+V case, there is no need to initialize V, instead initialize U for O(2p)).” Does this refer to something else?

If I output each HP runs to different output directories, how would I recombine the results at the end?

Best,
Jamie

On Dec 2, 2022, at 11:44 AM, Iurii TIMROV <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Jamie,

First of all, please note that you are not performing the DFT+U+V calculation but DFT+U because you specified this:

HUBBARD {ortho-atomic}
U Fe1-3d 1e-10
U Fe2-3d 1e-10
U Mn1-3d 1e-10
U Mn2-3d 1e-10
U O-2p 1e-10

The pw.x code will perform DFT+U+V only if it detects at least one V parameter in the input. So you should specify e.g. this:

HUBBARD {ortho-atomic}
U Fe1-3d 1e-10
U Fe2-3d 1e-10
U Mn1-3d 1e-10
U Mn2-3d 1e-10
V Fe1-3d O-2p 1 5 1e-10

I specified some random couple: 1st and 5th atom. The HP code will consider all O atoms as Hubbard atoms and will perturb non-equivalent O atoms.

Now, coming to your problem. Your calculations crash because you are using the same temporary directory for all calculations: outdir='outdir/'. So independent HP calculations write and read to/from the same folder and this creates a mess. Instead, for each independent calculation you need to use a separate temporary directory, e.g. outdir1, outdir2, etc. So once you performed  the ground-state calculation using pw.x, just copy outdir to outdir1, outdir2, ... and run independent HP calculations simultaneously:
HP run 1 : perturb_only_atom(5) = .true. -> outdir1
HP run 2 : perturb_only_atom(6) = .true. -> outdir2
...

HTH

Iurii


--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: Jamie Holber <holber at umich.edu<mailto:holber at umich.edu>>
Sent: Friday, December 2, 2022 5:04:00 PM
To: Iurii TIMROV
Cc: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] Problem with Parallelization of Hp.x

Just uploaded.

Thanks,
Jamie

On Dec 2, 2022, at 10:56 AM, Iurii TIMROV <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Thanks! Could you upload please also the input and output files of the pw.x calculations?

Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________

From: Jamie Holber <holber at umich.edu<mailto:holber at umich.edu>>
Sent: Friday, December 2, 2022 4:45:13 PM
To: Iurii TIMROV; Quantum ESPRESSO users Forum
Subject: Re: [QE-users] Problem with Parallelization of Hp.x

Hi Iurii,

Thanks for the response. I don’t believe it is a problem with disc space because when I run the hp calculations in serial or I run hp.x calculations at the same time with different outdir they run successfully.

I have uploaded the hp files to the google drive linked. I attempted to run hp_atom5.in and hp_atom6.in at the same time.
https://drive.google.com/drive/folders/1gRVALU0B8CLsaKBZpXzvEFAHcb5dMXZ_?usp=share_link

Thank you,
Jamie

On Dec 2, 2022, at 6:04 AM, Iurii TIMROV via users <users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>> wrote:

Dear Jamie,

The problem seems to occur when writing or reading files on a disc. Are you sure that you have enough free disc space?

Could you provide please more details about your calculations? Please share your input and output files for all pw.x and hp.x calculations that fail (e.g. using Google Drive).

Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Jamie Holber <holber at umich.edu<mailto:holber at umich.edu>>
Sent: Thursday, December 1, 2022 9:03:19 PM
To: users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
Subject: [QE-users] Problem with Parallelization of Hp.x

Hello everyone,

I am trying to replicate the U+V calculations for LiMnFePo4 as described in Timrov, Iurii, Nicola Marzari, and Matteo Cococcioni. "HP--A code for the calculation of Hubbard parameters using density-functional perturbation theory." arXiv  preprint arXiv:2203.15684 (2022).<https://arxiv.org/pdf/2203.15684.pdf> The hp files run fine when I run them one at a time. However, when I try to parallelize over different perturbed atoms by running them simultaneously as described in section 4.2 I receive errors and at least one run fails.  I have tried it on two different computing cluster and they both failed, but with different error messages. I’ve included the input files/errors below. Does anyone know of a way to solve this issue?

Input file 1:
&inputhp
    prefix = 'olivine', outdir='outdir/',
    nq1 = 1, nq2 = 2, nq3 = 3,
    conv_thr_chi = 1.0d-7,
    niter_max=250,
    dist_thr = 5.D-3
    perturb_only_atom(5) = .true.
/

Input file 2:
&inputhp
    prefix = 'olivine', outdir='outdir/',
    nq1 = 1, nq2 = 2, nq3 = 3,
    conv_thr_chi = 1.0d-7,
    niter_max=250,
    dist_thr = 5.D-3
    perturb_only_atom(6) = .true.
/

Errors from computing cluster 1

Error termination. Backtrace:
At line 700 of file buffers.f90 (unit = 20, file = 'outdir/olivine.wfc1')
Fortran runtime error: File cannot be deleted

Error termination. Backtrace:
#0  0x479784 in __buffers_MOD_close_buffer
at /home/jholber/LFP/QE_studies/copy/qe-7.1/PW/src/buffers.f90:700
#1  0x405a27 in hp_close_q_
at /home/jholber/LFP/QE_studies/copy/qe-7.1/HP/src/hp_close_q.f90:28
#2  0x4057c1 in hp_main
at /home/jholber/LFP/QE_studies/copy/qe-7.1/HP/src/hp_main.f90:143
#3  0x404fac in main
at /home/jholber/LFP/QE_studies/copy/qe-7.1/HP/src/hp_main.f90:14

Errors from Computer clusters 2


 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine davcio (22):
     error writing file "/home/holber/q-e/HP/examples/LFP/../../tempdir/HP/LFP.dwfc11"
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


Thank you,
Jamie Holber
University of Michigan
_______________________________________________
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu<http://www.max-centre.eu/>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221202/b7e49247/attachment.html>

More information about the users mailing list