You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
During #151, caching of the openmm_system was proposed as an optimization that could be valuable when we compute many gradients/energies for the same molecule in the same set of jobs. We currently cache the generated off_forcefield, so the same mechanism can be utilized for openmm_system.
The key/hash used for the cache must be selected with care. It must be insensitive to rotations or translations of the molecule, but should be sensitive to charge states, connectivity, and forcefield parameters. From @peastman:
One option for this is to serialize the System to XML with XmlSerializer.serialize(system) then compute a hash from the string. This will detect any change to the System or the Forces it contains, but will be unaffected by changes to particle positions.
The text was updated successfully, but these errors were encountered:
During #151, caching of the
openmm_system
was proposed as an optimization that could be valuable when we compute many gradients/energies for the same molecule in the same set of jobs. We currently cache the generatedoff_forcefield
, so the same mechanism can be utilized foropenmm_system
.The key/hash used for the cache must be selected with care. It must be insensitive to rotations or translations of the molecule, but should be sensitive to charge states, connectivity, and forcefield parameters. From @peastman:
The text was updated successfully, but these errors were encountered: