Segmentation fault (core dumped): Error when running lammps about pair_nequip

[DIooooooooooooo]

@Linux-cpp-lisp I have already turned to Allegro after I realized it. Thanks for your remind anyway~

[anjohan]

Hi,

Sorry for the late reply - I have been at a conference.

First of all, I would suggest starting out with a simplified version of the script (read the data, define the potential, run NVE).

Secondly, with 2.5k atoms, there is a possibility that you are simply running out of memory, depending on the memory of your GPU and the configuration of the model. This can easily be tested by simply deleting a lot of atoms before the first run command.

[DIooooooooooooo]

Hi @anjohan
It seems that I just simply running out of memory as you said. With a system of 216 atoms I almost run out of memory.
Thanks for your reply!

[Linux-cpp-lisp]

@DIooooooooooooo note that in general our solution to not being able to fit the number of atoms you want to run on a single GPU is our Allegro model and corresponding LAMMPS plugin pair_allegro, which are designed to enable parallelized MD over many GPUs for calculations with many atoms.