coordinates format for training

[benshpiro]

Hi, as an exercise I am creating frames for my data set from a LAMMPS simulation.
My system has periodic boundary conditions and so the lammps coordinates include the x,y,z components but also the box image components ix,iy,iz.
Is there a way to communicate the box image data to deepmd during the training, or do I need to translate all the coordinates to all be in only 1 box prior to the training?