pair_style with develop version of nequip/allegro

[matr1x-1]

Hello everyone,

I am having some problems with the pair_style. The allegro/nequip version are from the develop branch. When running lammps, the following error is raised: „Exception: expected scalar type Double but found Float“ (similar to this: #20).

Is there a different version of the pair_style that I have to use with the develop version of nequip and allegro?

Thank you!

[matr1x-1]

Using pair_style allegro6464 as suggested here works for me.

Edit: allegro6464 works with the pair_allegro from the 'main' branch and allegro from the 'develop' branch.

[matr1x-1]

One follow-up question on this: As written in the comment above, allegro6464 solves the problem when using an allegro model. However, when I try to run a nequip model (also from the develop branch) with pair_allegro (nequip branch), I get the same error again. As there is no option here to use e.g. nequip6464, what should I change in terms of the types that I use when training the model?

[Linux-cpp-lisp]

Sorry for the delay on your second question @matr1x-1 --- after the recent major update these options are no longer relevant, and NequIP GNN models trained with nequip will be in precision that work without any further configuration with pair_style nequip.