Stochastic test failures in RepartitionedHybridTopologyFactory_energies

I’m noticing some stochastic test failures in RepartitionedHybridTopologyFactory_energies. I’m opening this thread to start logging them for future assessment.

[zhang-ivy]

I tried turning on the sterics and removing the five attempts, but I’m now seeing this error for alanine dipeptide in vacuum:

E           AssertionError: The ratio in energy difference for the ZERO state is 0.455192615009072.
E            This is greater than the threshold of 0.0001.
E            real-zero: -22.36835783969803
E            alc-zero: -466.4774513254996
E           added-valence: -152.2754369353971

I think this is because the the validate_endstate_energies() function assumes that only valence energies are on when generating the geometry proposal and cannot handle testing the endstate energies when sterics are on.

[ijpulidos]

@zhang-ivy Should turn on the sterics and remove the five attempts.