Fine Tuning on DFT data with dispersion
#2
by rasoolahmad586 - opened
I want to fine tune MACE-MH-1 on my DFT data of surfaces. I have performed DFT calculations with D3 dispersion interaction. If I fine tune omat-pbe head, should I remove dispersion contribution from my DFT data before fine tuning the MACE potential? Or should I train with full DFT+D3 energy and forces? What strategy was used to train the foundation model regarding dispersion intercation.
rasoolahmad586 changed discussion title from Fine Tuning on DFT data to Fine Tuning on DFT data with dispersion