pipeline-components
Assemble and debug spaCy pipeline components, factories, registry wiring, and pipe analysis.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- pipeline-components
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Assemble and debug spaCy pipeline components, factories, registry wiring, and pipe analysis.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/built-in-components.md4.4 KB
- 参考文档references/component-factories-and-registry.md4.8 KB
- 参考文档references/pipeline-assembly-and-analysis.md6.5 KB
- 参考文档references/troubleshooting.md4.1 KB
- 脚本scripts/inspect_factories.py3.4 KB
03 SKILL.md 正文
pipeline-components
Validated against spaCy 3.8.15 on the installed CPU environment. Optional CUDA, Apple, transformer, and language-extra behavior was not verified here.
Use this sub-skill when the task is about Language.component, Language.factory, nlp.add_pipe, component order, registry lookup, source-vs-factory config wiring, or nlp.analyze_pipes.
Read first
- references/component-factories-and-registry.md for verified
Language.factory/Language.componentsignatures and component metadata. - references/pipeline-assembly-and-analysis.md for pipe ordering, enable/disable/select behavior, and
analyze_pipesusage. - references/built-in-components.md for the built-in component catalog and what each one roughly does.
- references/troubleshooting.md when
add_piperaises, config validation fails, or a source component gets mistaken for a factory. - scripts/inspect_factories.py for a safe registry and analysis smoke check.
What this sub-skill covers
- Registering stateless components with
Language.component. - Registering stateful factories with
Language.factoryand default configs. - Adding components by name, reordering them, and loading components from source pipelines.
- Reading and interpreting built-in factory metadata.
- Using
nlp.analyze_pipes()to understand assigns/requires/retokenizes relationships.
What to route elsewhere
- Tokenization,
Doc/Token/Span, matchers/rulers, serialization, and displaCy:documents-and-visualization. init config,debug config,train,evaluate,convert,package,validate:training-and-cli.- Install/import or backend concerns:
install-and-inspect. spacy projectorchestration:project-workflows.
Fast use pattern
- Check the exact component signature in the reference.
- Decide whether the user needs a stateless component or a stateful factory.
- Use the bundled registry inspection script to confirm the named component exists.
- Read
pipeline-assembly-and-analysis.mdwhen order, disable/enable, or source-vs-factory confusion appears.
Evidence basis
This sub-skill is grounded in the installed Language signatures, built-in factory source, pipeline tests, and public docs for component registration and pipeline processing.
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).