openff-toolkit
Use OpenFF Toolkit for molecule processing, SMIRNOFF force fields, topology/system preparation, and optional chemistry backend troubleshooting.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- openff-toolkit
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Use OpenFF Toolkit for molecule processing, SMIRNOFF force fields, topology/system preparation, and optional chemistry backend troubleshooting.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/package-overview.md2.7 KB
- 参考文档references/repo-provenance.md2.7 KB
- 参考文档references/repo-routing-metadata.json435 B
- 参考文档references/troubleshooting.md3.8 KB
- 脚本scripts/check_openff_toolkit.py6.4 KB
03 SKILL.md 正文
OpenFF Toolkit
Use this repo skill when a task involves the Open Force Field Toolkit (openff.toolkit), especially Molecule, Topology, SMIRNOFF .offxml force fields, toolkit wrappers, parameter assignment, molecule/topology conversion, conformers, charges, PDB handling, OpenMM/Interchange handoff, or force-field troubleshooting.
Read First
references/package-overview.mdsummarizes package purpose, install shape, important concepts, and cross-skill workflow order.references/troubleshooting.mdcovers cross-cutting install/import, optional dependency, backend, force-field discovery, and workflow routing failures.references/repo-provenance.mdrecords the source snapshot and evidence paths used to create this skill; read it before deciding whether the skill is stale.scripts/check_openff_toolkit.py --jsonchecks importability, version, available force fields, core classes, and backend wrapper status in the current Python environment.
Route by Task
| Task signal | Use this entry point | Why |
|---|---|---|
| Build a molecule from SMILES, mapped SMILES, SDF/MOL/SMI/PDB, RDKit/OpenEye objects, conformers, partial charges, stereochemistry, tautomers, SMARTS, visualization, or molecule graph checks | sub-skills/molecules-and-io/SKILL.md |
Focuses on single-molecule API usage and safe IO/validation recipes. |
| Assemble systems from molecules, load PDB files, handle proteins/HETATM records, manage hierarchy metadata, convert to/from OpenMM/MDTraj, write PDB, add constraints, or prepare topologies for parameterization | sub-skills/topology-and-systems/SKILL.md |
Owns multi-molecule Topology workflows and PDB/biopolymer assumptions. |
Load or edit .offxml, discover installed OpenFF force fields, inspect/modify SMIRNOFF handlers, label assigned parameters, serialize force fields, create Interchange/OpenMM systems, or troubleshoot SMIRNOFF errors |
sub-skills/smirnoff-force-fields/SKILL.md |
Owns ForceField and SMIRNOFF parameter workflows. |
Choose RDKit/OpenEye/AmberTools/NAGL/BuiltIn wrappers, inspect GLOBAL_TOOLKIT_REGISTRY, control registry precedence, diagnose missing optional dependencies, unsupported formats, or charge method failures |
sub-skills/toolkit-backends/SKILL.md |
Owns optional backend and registry behavior. |
Common Workflow Order
- Check the environment with
scripts/check_openff_toolkit.py --jsonif importability, force-field discovery, or optional backend availability is uncertain. - Create or validate
Moleculeobjects withsub-skills/molecules-and-io/SKILL.md. - Build the
Topologyor load a PDB/system withsub-skills/topology-and-systems/SKILL.md. - Load/apply/edit a SMIRNOFF force field with
sub-skills/smirnoff-force-fields/SKILL.md. - Return to
sub-skills/toolkit-backends/SKILL.mdwhen the failure is caused by an unavailable wrapper, unsupported file format, charge method, or registry ordering.
Installation and Import Baseline
Prefer package-manager installation for normal users because OpenFF Toolkit depends on compiled chemistry/simulation packages:
mamba create -n openff-toolkit -c conda-forge openff-toolkit
Minimal import check:
from openff.toolkit import ForceField, Molecule, Topology
from openff.toolkit.utils.toolkits import GLOBAL_TOOLKIT_REGISTRY
molecule = Molecule.from_smiles("CCO")
topology = Topology.from_molecules([molecule])
force_field = ForceField("openff-2.3.0.offxml")
print(molecule.n_atoms, topology.n_molecules, force_field.registered_parameter_handlers)
print(GLOBAL_TOOLKIT_REGISTRY.registered_toolkit_versions)
If ForceField("openff-2.3.0.offxml") fails, inspect available force fields and package installation before changing molecule/topology code.
Safe Defaults
- Use Python versions supported by the installed release; this repository snapshot declares Python
>=3.12. - Prefer conda-forge packages for RDKit, AmberTools, OpenMM, Interchange, OpenFF force fields, and OpenFF units.
- Keep toolkit selection explicit for reproducibility: pass
toolkit_registry=or a wrapper when behavior must match RDKit, OpenEye, AmberTools, or BuiltIn paths. - Treat PDB files as topology/system inputs, not reliable molecule graph sources unless hydrogens, elements, bonds, and unique molecules are known.
- Treat
.offxmlfiles as SMIRNOFF force-field sources; preserve originals when troubleshooting version, handler, or parameter errors.
What Not To Do
- Do not assume OpenEye, AmberTools, or NAGL are available just because
openff.toolkitimports. - Do not use a bare PDB file as the only source of small-molecule bond orders; prepare
unique_moleculeswhen loading PDB systems. - Do not mutate
GLOBAL_TOOLKIT_REGISTRYglobally unless the task explicitly needs process-wide behavior; prefer temporary context managers or explicit arguments. - Do not route force-field parameter problems to molecule IO only; use the SMIRNOFF sub-skill once molecules/topologies are valid.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).