model-ops-interpretability
Guides StellarGraph saved-model loading, custom Keras layers, calibration, ensembles, saliency and integrated gradients, randomness utilities, plotting, and optional Neo4j connector workflows.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- model-ops-interpretability
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides StellarGraph saved-model loading, custom Keras layers, calibration, ensembles, saliency and integrated gradients, randomness utilities, plotting, and optional Neo4j connector workflows.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/interpretability.md1.8 KB
- 参考文档references/neo4j.md1.9 KB
- 参考文档references/operations.md1.9 KB
- 参考文档references/troubleshooting.md1.6 KB
- 脚本scripts/calibration_smoke.py905 B
03 SKILL.md 正文
Model Operations and Interpretability
Use this sub-skill for operational tasks around StellarGraph models after or around training: saved-model loading, probability calibration, ensembles, interpretability, plotting/history, random seeds, and optional Neo4j connector workflows.
Read first
references/operations.mdforcustom_keras_layers, calibration, ensembles, history plotting, and seed utilities.references/interpretability.mdfor GCN/GAT saliency and integrated-gradient workflows.references/neo4j.mdfor optional Neo4j connector classes, service requirements, ID/features properties, and warnings.references/troubleshooting.mdfor saved model, calibration/ensemble, saliency, TensorFlow, and Neo4j failures.scripts/calibration_smoke.pyfor a safe tiny calibration smoke.
Route here when the user asks to
- load a Keras model containing StellarGraph custom layers;
- calibrate model outputs with
TemperatureCalibration,IsotonicCalibration, expected calibration error, or reliability diagrams; - build
EnsembleorBaggingEnsemblewrappers; - explain node/edge importance with GCN/GAT saliency maps or integrated gradients;
- check reproducibility utilities such as random seeds;
- use
Neo4jStellarGraph,Neo4jStellarDiGraph, or Neo4j GraphSAGE generators.
Route elsewhere
Core model training still belongs to the owning workflow route:
- node models:
../node-classification-gnns/SKILL.md; - link prediction/KG:
../link-prediction-kg/SKILL.md; - embeddings:
../embedding-workflows/SKILL.md; - graph/time-series:
../graph-time-series-workflows/SKILL.md.
Operating workflow
- Identify the operational surface: saved model, probability outputs, uncertainty/ensemble, saliency, or Neo4j-backed graph access.
- Verify package/TensorFlow import with the root diagnostic if the environment is suspect.
- For saved models, pass
stellargraph.custom_keras_layersas Kerascustom_objects. - For calibration, use logits or probabilities with the shapes required by the calibration class; validate with a tiny smoke before applying to model outputs.
- For saliency, keep generator sparse/dense settings consistent with the model and use node indices/classes that exist.
- For Neo4j, confirm
py2neo, service reachability, ID property, feature property, and uniqueness constraints before expecting model generators to work.
Safe check
python sub-skills/model-ops-interpretability/scripts/calibration_smoke.py --help
python sub-skills/model-ops-interpretability/scripts/calibration_smoke.py
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).