Skip to content

v2.0.0

PathGennie v2.0.0
@namus namus tagged this 02 Aug 11:34 
First release with the shared adaptive-sampling core validated end-to-end on
real MD across all three backends.

Highlights
  - QM/MM reactive workflow: path generation, path-CV refinement, free energy,
    and an NEB reference, demonstrated on SN2, E2 and SN1 systems.
  - Engine-agnostic path refinement via EngineSampler, working on AMBER,
    GROMACS and OpenMM.
  - Weighted Ensemble with burn-in and a steady-state rate estimator.
  - Multi-GPU swarm execution verified on two physical A100s.

Breaking
  - OpenMM is now the [openmm] extra rather than a core dependency, so
    AMBER/GROMACS-only users no longer install it. Use
    pip install 'pathgennie[openmm]'.

Test suite 244 passing; CI covers core, ML (torch) and MD (OpenMM + MDAnalysis)
dependency lanes.
Assets 2
Loading