Dataset refinement

[ald3hyde]

Dear all,

I have generated a large DFT dataset using VASP calculations, and I would like to reduce the number of configurations.
So, I mean, I'd like to exclude the unphysical or unstable configurations with a large error. Can I archive this problem with DP-GEN?

[njzjz]

You can use dpgen simplify and give a proper model_devi_f_trust_hi or model_devi_e_trust_hi.

However, I am wondering why not exclude high-energy points directly considering you have already had DFT energies.