retrieval-qa
Use DeepPavlov retrieval, ranking, FAQ, SQuAD, ODQA, and KBQA workflows with correct config, data, index, and input/output choices.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- retrieval-qa
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use DeepPavlov retrieval, ranking, FAQ, SQuAD, ODQA, and KBQA workflows with correct config, data, index, and input/output choices.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/data-and-indexing.md9.1 KB
- 参考文档references/model-catalog.md7.0 KB
- 参考文档references/troubleshooting.md6.8 KB
- 脚本scripts/tiny_retrieval_config.py6.0 KB
03 SKILL.md 正文
retrieval-qa
Use this sub-skill when a DeepPavlov task involves questions, contexts, documents, document IDs, ranking scores, FAQ categories, answer spans, ODQA pipelines, or Wikidata-style KBQA queries.
Route quickly
- Generic DeepPavlov config syntax, nested
config_path,overwrite, registry, train/evaluate/predict modes, and component wiring: ../pipelines/SKILL.md. - Generic text classification, NER, entity extraction, spelling, syntax, morphology, embedders, or relation extraction outside retrieval/QA: ../text-models/SKILL.md.
- REST or socket deployment after a retrieval/QA config runs locally: ../serving/SKILL.md.
- Shared install/import/cache/backend problems: ../../references/troubleshooting.md.
Start here
- Choose the workflow family in references/model-catalog.md:
doc_retrieval,ranking,squad,odqa,kbqa, orfaq. - Prepare the required local data/index layout with references/data-and-indexing.md.
- For a safe local document-retrieval template, generate a two-document config with scripts/tiny_retrieval_config.py.
- Diagnose retrieval/QA-specific failures with references/troubleshooting.md before escalating to root troubleshooting.
Decision hints
- Use
squadwhen every question has an explicit context batch. - Use
odqawhen the user supplies only questions and expects retrieval plus answer extraction over a document collection. - Use
doc_retrievalwhen the deliverable is document IDs/titles/scores rather than answer spans. - Use
rankingfor response ranking or KBQA relation/path ranking; route ordinary classifiers totext-models. - Use
kbqafor Wikidata-style entity/relation/query generation outputs. - Use
faqfor fastText/logistic-regression FAQ intent/category lookup or simple CSV FAQ corpora.
Do not reopen the original repository checkout for runtime guidance; use the bundled references and scripts in this sub-skill tree.
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).