pipeline-programming
Build, run, batch, debug, and smoke-check Towhee custom pipelines with pipe.input, RuntimePipeline, AutoConfig, and AutoPipes.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- pipeline-programming
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Build, run, batch, debug, and smoke-check Towhee custom pipelines with pipe.input, RuntimePipeline, AutoConfig, and AutoPipes.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/debugging-and-profiling.md4.8 KB
- 参考文档references/pipeline-api.md11.0 KB
- 参考文档references/troubleshooting.md6.1 KB
- 脚本scripts/pipeline_smoke.py4.1 KB
03 SKILL.md 正文
Towhee pipeline programming
Use this sub-skill when the user needs to define or diagnose a Towhee custom pipeline with pipe.input(...), chained node methods, output(...), RuntimePipeline.__call__, batch(...), debug(...), or flush(). It also covers safe, local-only checks for AutoConfig and network-aware guidance for AutoPipes.pipeline(...).
Fast routing
- For pipeline node syntax, callable contracts, schemas,
RuntimePipeline,batch,flush,AutoConfig, andAutoPipes, use Pipeline API. - For profiler/tracer usage, debug result inspection, include/exclude filtering, and batch debug, use Debugging and profiling.
- For failures involving schema names, missing output columns,
debug()flags, Hub/cache/network misses, or optional Hub operator downloads, use Troubleshooting. - For a no-network smoke check in the target environment, run pipeline_smoke.py with
--verbosewhen useful.
Scope boundaries
- Stay here for custom pipeline construction, local lambda/callable nodes, node semantics, runtime execution, batch/debug/flush,
AutoConfig.*, andAutoPipes.pipelineselection caveats. - Route operator template creation, Hub operator packaging,
opsregistration patterns, and CLI operator scaffolding tooperator-hub-and-cli. - Route HTTP/GRPC services, Docker, Triton server startup,
towhee server, and deployment packaging toserving-and-triton. - Route
DataCollectiondisplay, entity/type wrappers, serializers, and data loader details todata-utilities. - Route model training,
Trainer, andNNOperator.train/setup_trainerworkflows totraining-and-models.
Default safe workflow
python scripts/pipeline_smoke.py --verbose
If the smoke check fails at import time, fix the active Python environment before attempting Hub, model, Triton, or deployment workflows. If the smoke check passes, prefer lambda/callable pipelines for quick local diagnosis before introducing Hub operators that may download code, models, or optional dependencies.
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).