Generating training data from trajectory

[jonathan-booth]

I have a few trajectories where, at each frame, I have a text file where each line is an atom and we have values for the atom type, the x,y,z positions, and the x,y,z components of the force vector. Each text file also has the total potential energy. I have trajectories for different sized systems. How do I convert these data into something that Nequip (Allegro really) can work with? I've read the docs but not found an answer sorry if I missed something obvious.

Thanks in advance

[Linux-cpp-lisp]

Hi @jonathan-booth ,

You need to convert it into extXYZ format; the easy way is to parse your format yourself in Python, make ase.Atoms objects (see their docs) and then write them out with ase.io.write in the extxyz format. At present only extxyz format supports variable system size inputs in our code.

[Linux-cpp-lisp]

Oh, closing as duplicate: mir-group/allegro#74