[BUG] dp test reports weighted avg of Virial errors even if the dataset does not contain virial info #4563
Description
Bug summary
I'm comparing the dp test result over ANI dataset. The dataset does not contain virial labels. However, it reports virial metrics on weighted average of errors, while test output for each system does not.
# ---------------output of dp test---------------
# testing system : /bohr/lambench-ood-3z0s/v6/LAMBench-TestData-v1/ANI/9
# number of test data : 376
Energy MAE : 2.464647e-01 eV
Energy RMSE : 3.475536e-01 eV
Energy MAE/Natoms : 2.738497e-02 eV
Energy RMSE/Natoms : 3.861707e-02 eV
Force MAE : 1.202944e-01 eV/A
Force RMSE : 2.087194e-01 eV/A
# -----------------------------------------------
# ----------weighted average of errors-----------
# number of systems : 50
Energy MAE : 1.867755e-01 eV
Energy RMSE : 2.697340e-01 eV
Energy MAE/Natoms : 1.355690e-02 eV
Energy RMSE/Natoms : 2.290897e-02 eV
Force MAE : 8.484314e-02 eV/A
Force RMSE : 1.472937e-01 eV/A
Virial MAE : 3.295064e+00 eV
Virial RMSE : 5.175478e+00 eV
Virial MAE/Natoms : 2.031481e-01 eV
Virial RMSE/Natoms : 3.304690e-01 eV
# -----------------------------------------------
DeePMD-kit Version
devel
Backend and its version
torch
How did you download the software?
pip
Input Files, Running Commands, Error Log, etc.
Steps to Reproduce
dp --pt test -m /tmp/lambench/change-bias-dp_dpa2_v2.4.0_1126_800w.pth -s /bohr/lambench-ood-3z0s/v6/LAMBench-TestData-v1/ANI -l dptest_output.txt --head Domains_Drug
Further Information, Files, and Links
No response