LAMMPS failed with model trained with float64

[saeedshahi]

Hi,

I trained an ML potential with float64 and it converged perfectly but when I want to use it with LAMMPS, it failed and I got this error "RuntimeError: expected scalar type Float but found Double".
the model trained with float32 works fine with LAMMPS.

Best,
Saeed

[Linux-cpp-lisp]

Hi @saeedshahi ,

Thank you for your interest in our code!

pair_nequip does not yet support float64 models (#21). The link contains details about how a model trained in float64 can be deployed for float32 inference.