MMML Documentation¶
MMML combines molecular mechanics workflows with machine-learned force fields built on JAX.
Use this site for installation, the CLI reference, MD workflow guides, and development notes.
What is here¶
- Getting started — install with
uv, serve docs locally - PyCHARMM + MM/ML capabilities checklist — status checklist, examples, diagrams
- CLI overview —
mmml commands,mmml examples, tab completion, per-command--help - md-system YAML configs — single runs, campaigns, condensed-phase builders
- Packmol placement — default composition builder
- Protein force fields — CHARMM36 + jax-md peptide examples
- CHARMM CGenFF JAX clone — bonded + switched nonbonded JAX vs PyCHARMM
- MLpot guides — COM handoff switches, medium PBC, spatial MPI, long-range electrostatics
- Calculator profiling — JAX compile vs run, jax-pme primitives, cProfile / TensorBoard
- PyCHARMM on clusters — MPI launcher, threading, FFTW
- PyCHARMM C API: PBC box & pressure — KEY_LIBRARY get/set for cell side and CPT pressure tensor
- Package architecture — module layout and import graph
- Scientific code policy — reproducibility, provenance, failure integrity, and review checklist
- 1D dimer scan design — canonical modular API and self-describing result bundle
- Nudged elastic band (NEB) — PhysNet MEP / barrier sampling (
mmml neb) - Batched umbrella sampling — pure-ML distance umbrella NVT + MBAR
- Dimer tool inventory — canonical, campaign, validation, exploratory, and deprecated paths
- MPNN tool inventory — PhysNet / Spooky / DCM / EF ownership and shared kernels
- Calculator capability matrix — calculators, hybrid assembly, charge modes, MM modes, LR solvers, and drivers