Currently, our workflows are tailored to VASP:
The RSS workflow, for example, directly uses VASP input parameters:
|
custom_incar: dict | None |
While this is fine for the first versions of Autoplex, we would need to aim for a generalization in the next steps. Similar to atomate2, I would expect that we can operate with DFTMakers instead. To ensure compatibility, we might need to carefully consider how such an addition could look.
The phonon workflow can already operate with Makers and a generalization would mostly mean making sure different DFT output files can be converted into files ready for machine learning.
Currently, our workflows are tailored to VASP:
The RSS workflow, for example, directly uses VASP input parameters:
autoplex/src/autoplex/auto/rss/flows.py
Line 125 in 1de3c34
While this is fine for the first versions of Autoplex, we would need to aim for a generalization in the next steps. Similar to atomate2, I would expect that we can operate with DFTMakers instead. To ensure compatibility, we might need to carefully consider how such an addition could look.
The phonon workflow can already operate with Makers and a generalization would mostly mean making sure different DFT output files can be converted into files ready for machine learning.