João Morado

Hi, Say you have the following tert-butyl molecule.  You want to perform a typical 2-stage RESP procedure where in the first step you allow all charges to vary freely...

Alternatively, it is also possible to set the symmetries (also manually) in the .ff files that ParaMol reads/writes. See this example: https://paramol.readthedocs.io/en/latest/Examples/Example_5/Example_5.html And specifications of the .ff file: https://paramol.readthedocs.io/en/latest/Files_specification.html

It should be possible to calculate the objective function in parallel using the OpenCL platform. Support for the CUDA platform is not available and has not been implemented. Also, always...

I agree this would make sense. Should we go ahead and create a PR for this?

@svarner9, could you try the current implementation available [here](https://github.com/sef43/openmm-ml/tree/nequip)? It uses the NNPOps neighbor list, so I anticipate it might be slightly faster for a system of the size you're...

Just for the records, I'm posting here a comparison between the energies I get when using this OpenMM-ML interface and the ASE-like `NequIPCalculator`. The script I'm using to calculate the...

That's true, thanks for pointing that out. Regarding the forces, this is what I get (values in kJ/mol/nm): Toluene ``` NequIPCalculator forces: [[ 318.61404 -1153.1539 783.135 ] [ 379.53235 455.95953...

This is done from my side. If someone could take a look and review the changes, that would be great. Performance benchmarks on test models can be found [here](https://github.com/mir-group/nequip/issues/288#issuecomment-2094822295). Many...

Many thanks for the thorough review, @peastman! Most of it should be now resolved. Thanks for testing, @svarner9. I think the slow performance you're seeing is not related to that...

@peastman @Linux-cpp-lisp, I've trained a model with these settings: ```yaml default_dtype: float64 model_dtype: float64 allow_tf32: true ``` and the energy and force differences between ASE and OpenMM are indeed _very_...