Got problems of atomd and ksum

65 views
Skip to first unread message

Yuan Yin

unread,
Jul 7, 2022, 10:32:59 AM7/7/22
to DMFTwDFT
pic.jpgDear Developers,

I always got problems when running RUNDMFT.py even though I have tried many times to re-compile the codes. The followings are errors:
# for atom_d.error, here shows: "/bin/sh: /data/home/yinyuan/DMFTwDFT-old/binatom_d.py: No such file or directory"
# for ksum_error_XHF), here shows "HYDU_create_process (../../utils/launch/launch.c:622): execvp error on file /data/home/yinyuan/DMFTwDFT-old/binXHF0.py (No such file or directory)"

"[yinyuan@login01 DMFT]$ RUNDMFT.py >> out.dat
cp: cannot stat ‘UC.dat’: No such file or directory
Traceback (most recent call last):
  File "/data/home/yinyuan/DMFTwDFT-old/bin/RUNDMFT.py", line 398, in <module>
    DMFT.EKIN0 = float(lines[-1].split()[4])
ValueError: could not convert string to float: Nd[-1]"

I can not fix these errors. My compile environment is: 1) intel/compiler_2016;   2) intel/impi_5.1.3;  3) gcc (GCC) 4.8.5 20150623;    4) Python 2.7.14   scipy (0.19.1)    numpy (1.13.3)    weave (0.17.0).

Hope someone can give me a hand if you have experienced the same issues. Thanks so much!!!

pic.jpg

Uthpala Herath

unread,
Jul 7, 2022, 10:52:48 AM7/7/22
to Yuan Yin, DMFTwDFT
Hello Yuan, 

In your INPUT.py add a / after bin in your directory path. 
That way it would be /data/home/yinyuan/DMFTwDFT-old/bin/atom_d.py instead of /data/home/yinyuan/DMFTwDFT-old/binatom_d.py. 

Best,
Uthpala

--
You received this message because you are subscribed to the Google Groups "DMFTwDFT" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dmftwdft+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dmftwdft/c9c68cf1-037e-48a8-8262-f2ee055d3da7n%40googlegroups.com.


--
Uthpala Herath
Postdoctoral Associate
Department of Mechanical Engineering and Materials Science
Duke University
Durham, NC

Uthpala Herath

unread,
Jul 8, 2022, 10:02:24 AM7/8/22
to Yuan Yin, DMFTwDFT
Can you try to see if you can import mpi4py from a compute node?


On Fri, Jul 8, 2022 at 6:07 AM Yuan Yin <13030...@gmail.com> wrote:
Thank you, sir,
I modified the path in the file INPUT.py, and it works. And then, I run an example of LaNiO on my computer, the dmft cycle runs well. But an error occurred when I tried to submit a task to another node in my servers. In brief, I can successfully run a dmft calculation on my login node but can not use a PBS script to run it on a specified node on my servers.
Here are the errors:

# in ksum_error_XHF0, here shows:"Traceback (most recent call last):
  File "/data/home/yinyuan/DMFTwDFT-old/bin/XHF0.py", line 22, in <module>
    from mpi4py import MPI
ImportError: libmpi.so.40: cannot open shared object file: No such file or directory"
# in error.dat, here shows:" Traceback (most recent call last):

  File "/data/home/yinyuan/DMFTwDFT-old/bin/RUNDMFT.py", line 398, in <module>
    DMFT.EKIN0 = float(lines[-1].split()[4])
ValueError: could not convert string to float: Nd[-1]"

I guess something wrong with my environment but I  do not know how to fix it. On my login node, anaconda5.3.1 was installed with a python2.7.15. The default python on my servers is python2.7.5. Moreover, I want to show you my PBS script:
#!/bin/bash
#PBS -N dmft_run_test
#PBS -l nodes=cn34:ppn=8
#PBS -l walltime=999:00:00
export I_MPI_FABRICS=shm:tmi
export I_MPI_CC=icc
export I_MPI_CXX=icpc
export I_MPI_F77=ifort
export I_MPI_F90=ifort
module load intel/compiler_2016
module load intel/impi_5.1.3
cd $PBS_O_WORKDIR
/data/home/yinyuan/DMFTwDFT-old/bin/RUNDMFT.py >> out.dat 2>error.dat

Hope you can give some advice to fix this error. Best thanks!!!

Best wishes!
Yuan

Uthpala Herath <ukh...@mix.wvu.edu> 於 2022年7月7日週四 下午10:52寫道:

Uthpala Herath

unread,
Jul 8, 2022, 12:14:26 PM7/8/22
to Yuan Yin, DMFTwDFT
Can you try to install mpi4py with conda and see if it helps?

On Fri, Jul 8, 2022 at 11:25 AM Yuan Yin <13030...@gmail.com> wrote:
I find another issue when I log in to a node (like cn34) and do "locate libmpi.so.40". Here comes only path "/usr/mpi/gcc/openmpi-3.1.4-hfi/lib64".
When I do "locate libmpi.so.40" on the login node (login01), here comes:"/usr/local/lib/libmpi.so.40" and "/usr/mpi/gcc/openmpi-3.1.4-hfi/lib64/libmpi.so.40".
pic.jpg

Can you tell me how to deal with that? Thanks!!!

Best wishes
Yuan

Yuan Yin <13030...@gmail.com> 於 2022年7月8日週五 下午11:08寫道:
Yes, I can import mpi4py on any node on my servers. When I "locate libmpi.so.40", the path is "/usr/mpi/gcc/openmpi-3.1.4-hfi/lib64". My compile environment is 1)intel/compiler_2016, 2)intel/impi_5.1.3, 3)gcc4.8.5.
Is there any possible mistake when I compile the codes? The Makefile to compile the DMFTwDFT code is the default intel one. One more thing is that the mpi4py is installed via pip~ but not conda~, could that be the reason?


Best thanks to you.

Uthpala Herath <ukh...@mix.wvu.edu> 於 2022年7月8日週五 下午10:02寫道:
Reply all
Reply to author
Forward
0 new messages