Questions about DFT calculation of OMat24

[hexagonrose]

What would you like to report?

Hello! I'm interested in fine-tuning the pretrained model trained with OMat24.
I'm working on creating my own fine-tuning dataset and would like to know the precise DFT calculation settings used for OMat24 to ensure compatibility.

1. The version of VASP and pymatgen?
2. As fas as I know, it's better to use ISMEAR=0 to metallic systems. But, MPRelaxSet makes INCAR with ISMEAR=-5 for metallic systems. When performing the calculations, was there anything changed in the input file created by MPRelaxSet other than the ALGO parameter?

[lbluque]

Hi @hexagonrose,

1. We used VASP 6.3.2
2. Apart from the ALGO parameters we used VASP 54 pseudopotentials (this is the most important to ensure compatibility). We did run ISMEAR=-5 for metals, but you are correct ISMEAR=0 is more appropriate for metallic systems.

Note that you do not have to necessarily match the exact settings if you are going to fine-tune. For example finetuning with MPTrj gives very accurate models for the MPTrj settings.