rules-and-trie
Guides HanLP deterministic rules, sentence and string utilities, Trie/TrieDict matching, and model-backed custom dictionary overlays.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- rules-and-trie
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides HanLP deterministic rules, sentence and string utilities, Trie/TrieDict matching, and model-backed custom dictionary overlays.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/custom-dictionaries.md857 B
- 参考文档references/trie-reference.md695 B
- 参考文档references/troubleshooting.md607 B
- 脚本scripts/rules_smoke.py884 B
- 脚本scripts/trie_smoke.py964 B
03 SKILL.md 正文
Rules And Trie
Use this sub-skill when a task mentions deterministic sentence splitting, hanlp.utils.rules, possible_tokenization, Trie, TrieDict, TupleTrieDict, custom dictionaries, dict_force, dict_combine, dict_tags, gazetteers, or longest-prefix matching.
Read First
- Read
references/trie-reference.mdfor verified trie and dictionary APIs. - Read
references/custom-dictionaries.mdfor model-backed tokenizer/POS/NER dictionary overlays. - Read
references/troubleshooting.mdfor matching surprises, tuple keys, whitespace, and model-download boundaries. - Run
scripts/rules_smoke.pyfor no-download rules/string utility checks. - Run
scripts/trie_smoke.pyfor no-download trie and dictionary checks.
Minimal Trie Use
from hanlp_trie import Trie, TrieDict
trie = Trie({'商品': 'goods', '和服': 'kimono'})
print(trie.parse('商品和服'))
print(trie.parse_longest('商品和服'))
custom = TrieDict({'重要': 'important'})
print(custom.tokenize('第一个词语很重要'))
Trie and rule utilities can be verified without model downloads. Model-backed dictionary overlays such as tok.dict_force, tok.dict_combine, pos.dict_tags, and ner.dict_whitelist require a loaded pretrained component before they can change model output.
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).