OpenMM-ML uses a plugin architecture for defining potential functions. It supports a few potentials out of the box. Other packages can add support for additional potentials. To do that, the potential has to be registered by calling MLPotential.registerImplFactory(). The user doesn't need to do that directly. Instead, the package can do it automatically when you import it.
This provides a mostly seamless experience for users, but it does have one minor issue: your script has to import the outside library before you can create a MLPotential from it. If you leave out the import, the name of the potential function won't be recognized.
We can fix this by defining an entry point for potential functions. That would let MLPotential automatically discover all available potentials without any other packages needing to by imported first.
OpenMM-ML uses a plugin architecture for defining potential functions. It supports a few potentials out of the box. Other packages can add support for additional potentials. To do that, the potential has to be registered by calling
MLPotential.registerImplFactory(). The user doesn't need to do that directly. Instead, the package can do it automatically when you import it.This provides a mostly seamless experience for users, but it does have one minor issue: your script has to import the outside library before you can create a MLPotential from it. If you leave out the import, the name of the potential function won't be recognized.
We can fix this by defining an entry point for potential functions. That would let MLPotential automatically discover all available potentials without any other packages needing to by imported first.