Open chiang-yuan opened 3 months ago
So currently, this doesn't actually require CHGNet to be installed I don't think. I.e., you can import this and overwrite those makers to get a MACE or M3GNet elastic constant workflow. However, it could make sense to create several versions of the workflow, one for each universal force field?
One thing I realised looking at this is that we can remove the prev_calc_dir_argname
function from the BaseElasticMaker
and subclasses since this has now been standardised across all codes.
So currently, this doesn't actually require CHGNet to be installed I don't think. I.e., you can import this and overwrite those makers to get a MACE or M3GNet elastic constant workflow. However, it could make sense to create several versions of the workflow, one for each universal force field?
One thing I realised looking at this is that we can remove the
prev_calc_dir_argname
function from theBaseElasticMaker
and subclasses since this has now been standardised across all codes.
@utf also in the Abinit workflows?
Yes @utf you are definitely right - one could always replace default factory with the relaxmaker they like to use. However, I think this design is a bit counterintuitive and one might just want to set the forcefield calculator name and the code switch the input generator accordingly like what we did for MDMaker
Not a serious issue so maybe I shouldn't call it bug at the first place (I think it is the template..)
ElasticMaker
shouldn't use submaker as default factory otherwisechgnet
will be required to be installed as dependencyhttps://github.com/materialsproject/atomate2/blob/313f8098d58e3afcebd89a840aa489050afe2477/src/atomate2/forcefields/flows/elastic.py#L62-L70