Incorrect Lennard-Jones functional form

[jchodera]

There’s a bug in the Lennard-Jones functional form here

    return epsilon * (
        coefficients[0] * sigma_over_x ** order[0]
        - coefficients[1] * sigma_over_x ** order[1]

The prefactor should be 4*epsilon, not just epsilon.
This wouldn’t impact error statistics for fitting energies, but if you’re looking at MM parameter RMSE for LJ, it would matter.
Likely does not impact any experiments you’re running now.

[maxentile]

Well-spotted -- thanks for catching this!

Will fix this early next week, but I agree with assessment that this bug doesn't impact immediate results involving valence fitting.

This also suggests an issue with coverage of OpenMM consistency tests.

[yuanqing-wang]

fixed in #58

[yuanqing-wang]

This also suggests an issue with coverage of OpenMM consistency tests.

we have yet to figure out how to turn on/off the exceptions in OpenMM energies. so we're not doing nonbonded tests

[jchodera]

we have yet to figure out how to turn on/off the exceptions in OpenMM energies. so we're not doing nonbonded tests

I'm not quite sure I understand, but let me know what you are looking to do here and I can help.