How to obtain a float32 LAMMPS-runnable pretrained model

[alphalm4]

Hello developers,
I'm using the best two models (2L-OAM-L, 2L-OMAT-L) from grace foundational model list for benchmark uMLIPs,
and now especially doing something like speed benchmark.

We have already gathered initial results and shared them as a preprint,
but recently I realized that the grace models use float64 by default, whereas SevenNet runs in float32, which might mislead the results.

Is there any way to obtain float32-model for LAMMPS MD of GRACE-2L-OAM-L?
I have the cached model in $HOME/.cache/grace/GRACE-2L-OMAT-large-ft-AM and read about Change model's floating point precision but still not clear how to proceed it.

Thanks!

[AntBoc]

Hi,
Running LAMMPS in fp32 is currently not possible, only in python. But this is something we are working on, I'll come back here when we have it running.