Lammps is giving wrong energy values

[hdjani]

Hello, why the energies I got from Lammps using deepmd potential are completely differents from those in my training dataset?
here is attached the training results and below my lammps input:
#relaxation
units metal
boundary p p p
atom_style atomic
neighbor 1.0 bin
box tilt large
read_data structure-Pnma.lmp
run 10
change_box all triclinic
pair_style deepmd graph-compress.pb
pair_coeff * *
thermo_style custom step temp pe ke etotal press vol lx ly lz xy xz yz
thermo 10
dump 1 all custom 10 dump.relax id type x y z fx fy fz
fix 1 all box/relax tri 0.0
min_style cg
minimize 0 1.0e-5 10000 100000 #etol ftol maxiter maxeval

CsPbBr3-ML-Results.pdf

[njzjz]

fix 1 all box/relax tri 0.0

This command produces additional energy.
I have answered here: #4758 (comment)