Python Interface Question

[brian.w....@gmail.com]

Hi all,

I'm trying to get an MD simulation working with ANN potentials I've trained and I'm running into problems with the python interface. First my environment is:

ubuntu 16.04

python 3.5.5

numpy 1.14.0

ase 3.10.0

cython 0.28.4

gfortran 5.4.0

aenet2.0.3

I seem to have a problem importing the aenet.ase_calculator when I run the aenet-md.py script. The error I get is:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/orlando/anaconda3/lib/python3.5/site-packages/aenet-0.1.0a1-py3.5-linux-x86_64.egg/aenet/ase_calculator.py", line 24, in <module>
    from aenet.core import ANNPotentials
ImportError: /home/orlando/anaconda3/lib/python3.5/site-packages/aenet-0.1.0a1-py3.5-linux-x86_64.egg/aenet/core.cpython-35m-x86_64-linux-gnu.so: undefined symbol: mpi_allreduce_

When I compiled aenet2.0.3 I used the Makefile.gfortran_mpi file and I made a library using the make lib -f ./makefiles/Makefile.gfortran_mpi. Not too sure what the problem is so any suggestions would be greatly appreciated!

Best,

Brian Doherty

[Nongnuch Artrith]

Hi all,

Sorry, our interface implementation is only working in serial --- not with mpi.  That is probably the reason.  I have not tested the interface with anaconda Python either.

Best,

Nong

[Daniel Massote]

Dear all,

I am facing the same issue, though I am using serial version. I am using anaconda. Does anyone made it work that way?

Best,

Prof. Daniel Massote

UFJF - Brazil

[Daniel Massote]

Actually I was able to compile it again, without mpi error. I compiled everything with gfortran serial. When trying to run aenet-md.py, I got the error: "double free or corruption (out)".

I am not sure if I am missing anything, but the README files do not state the clearly the correct order of compilation. In python3 folder, we should run "python setup.py build_ext --inplace", but in the main README it only says that "python setup install --user". In the end, I run both commands in the order I mention them.

Best,

Daniel

[Nongnuch Artrith]

Dear Daniel,

Thanks for your interest in using aenet.

Please see a working example (all command lines) in the following Google Colab notebook [1]: https://github.com/atomisticnet/MLP-beginners-guide

To run the example, you first need to create your own copy of the notebook by choosing “File→Save Copy in Drive”.

Best,

Nong

[1] A.M. Miksch, T. Morawietz, J. Kästner, A. Urban, N. Artrith, 

“Strategies for the Construction of Machine-Learning Potentials for Accurate and Efficient Atomic-Scale Simulations”, submitted (2021). https://arxiv.org/abs/2101.10468

===============================

Nongnuch Artrith, PhD
Research Scientist (PI) 

Department of Chemical Engineering and 

Columbia Center for Computational Electrochemistry 

Columbia University 500 W. 120th St., 1006 Mudd
New York, NY 10027
Tel: 212-854-0105 or 857-523-8452
E-mail: nart...@atomistic.net and na2...@columbia.edu
Web: http://nartrith.atomistic.net and http://ann.atomistic.net

Also funded by:
Center for Functional Nanomaterials (CFN)
Brookhaven National Laboratory
===============================

[Daniel Massote]

Thanks Nong,

I missed this text. I followed everything. One thing that I noticed was the "pip install". I run exactly and still got the error. I tried to find exactly where it was crashing, and run these commands in ipython:

In [1]: import aenet

In [2]: from aenet.ase_calculator import ANNCalculator

In [3]: import ase

In [4]: import ase.spacegroup

In [5]: import ase.visualize

In [6]: atoms = ase.io.read("structure4536.xsf")

In [7]: ase.io.write('tio2.traj', atoms)

In [8]: ase.io.write('tio2.png', atoms)

In [9]: pot = {"Ti": "Ti.15t-15t.nn-04885", "O": "O.15t-15t.nn-04885"}

In [10]: calc = ANNCalculator(pot)

double free or corruption (out)

Looks like the calculator is not able to load the potentials. I tested with my own generated potential and also the one provided in the TiO2 example. Both crashed at the same point.

I think it may be some software version. All my system is updated. Is there any limit to python or ase?

Sorry about this.

Best,

Daniel

[Nongnuch Artrith]

Hi Daniel,

The commands from the Google Colab notebook are not meant to be run in your own IPython terminal, since they download and compile aenet and the Python interface. I would recommend first running the notebook from within Google Colab.

The example uses Python 3 and the most recent version of ASE.

Best,

Nong

--
You received this message because you are subscribed to the Google Groups "aenet" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aenet+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aenet/e02c0193-15b5-42ea-9bab-6b391e856b67n%40googlegroups.com.

[Daniel Massote]

Thanks for helping out.

Best,

Daniel