protocols
Choose and configure OpenFE protocol classes, default settings, ProtocolDAG creation, API execution boundaries, OpenMM backend settings, and simulation failure modes.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- protocols
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Choose and configure OpenFE protocol classes, default settings, ProtocolDAG creation, API execution boundaries, OpenMM backend settings, and simulation failure modes.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/protocol-reference.md7.5 KB
- 参考文档references/settings-and-execution.md9.3 KB
- 参考文档references/troubleshooting.md9.2 KB
- 脚本scripts/inspect_protocol_defaults.py7.5 KB
03 SKILL.md 正文
OpenFE Protocols
Use this sub-skill when the task is about selecting an OpenFE protocol class, customizing protocol settings before construction, creating Transformation or ProtocolDAG objects, setting OpenMM backend options, or diagnosing protocol-level simulation failures.
Route Tasks
- Choose among
RelativeHybridTopologyProtocol,AbsoluteBindingProtocol,AbsoluteSolvationProtocol,SepTopProtocol, andPlainMDProtocol; see protocol reference. - Inspect and edit
default_settings()safely before protocol construction, includingprotocol_repeats, lambda/window settings, sampling lengths, output/checkpoint settings, OpenMM platform settings, and optional_adaptive_settings()caveats; see settings and execution. - Create API-side work with
Transformation.create()orprotocol.create(...), understand what aProtocolDAGcontains, and know whenexecute_DAGactually runs expensive simulations; see settings and execution. - Print protocol default settings without creating systems or running simulations using the defaults inspector.
- Troubleshoot wrong protocol topology, immutable protocol/settings surprises, invalid OpenMM settings, CUDA/PTX/platform issues, optional parameterization packages, analysis warnings, checkpoint/resume boundaries, and wall-time interruptions; see troubleshooting.
Safe Operating Rules
- Treat
execute_DAG,openfe quickrun, and any protocolcreate(...)followed by execution as expensive and potentially file-writing; inspect settings and construct objects only when the user wants API planning. - Customize the
Settingsobject before constructing the protocol. If a protocol already exists, rebuild a new protocol from a copied or fresh settings object instead of mutating the existing protocol. - Prefer
default_settings()for a known baseline. Use_adaptive_settings()only when the protocol documents it, the requiredChemicalSystemand mapping inputs are available, and the user accepts that the method is experimental. - Use the bundled defaults inspector for safe defaults discovery; it imports protocol classes and serializes settings, but it never builds chemical systems, submits jobs, runs OpenMM, or downloads data.
- Keep protocol guidance separate from execution orchestration: CLI command syntax and scheduler command generation belong in
../cli-workflows/SKILL.md, and post-run estimates or gathered tables belong in../results-analysis/SKILL.md.
Boundary Handoffs
- Ligand/protein/solvent loading,
ChemicalSystemconstruction, atom mapping, and ligand-network planning belong in../network-planning/SKILL.md. - CLI flags,
openfe quickrun, resume command syntax, repeat-safe shell/Slurm command generation, and gather command selection belong in../cli-workflows/SKILL.md. - Result JSON decoding,
ProtocolResult.get_estimate(), uncertainty interpretation, plots, and network result aggregation belong in../results-analysis/SKILL.md.
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).