What I mean is, sometimes people make mistakes in the conversion from eV to A/ps. If there is a unit conversion error, it is more likely to reside in your script than anywhere else. I suggest you first verify by hand that the velocities output by LAMMPS add up to the kinetic energy of the system reported by LAMMPS. If not, then you are not correctly converting from A/ps to eV.
I notice that the mass ratio of H to Zr is about two orders of magnitude.