fate
Operate FATE 2.2.0 federated-learning deployments, FateFlow Pipeline workflows, local launchers, and component CLI/runtime inspection.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- fate
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Operate FATE 2.2.0 federated-learning deployments, FateFlow Pipeline workflows, local launchers, and component CLI/runtime inspection.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/package-overview.md4.7 KB
- 参考文档references/repo-provenance.md2.6 KB
- 参考文档references/repo-routing-metadata.json417 B
- 参考文档references/troubleshooting.md4.5 KB
- 脚本scripts/check_fate_install.py8.2 KB
03 SKILL.md 正文
FATE
Use this repository skill when a task involves FATE / FederatedAI/FATE 2.2.x: installing or checking the package, starting FateFlow-backed services, writing service-backed FateFlowPipeline jobs, running service-free local launchers, or inspecting the local component runtime exposed by python -m fate.components.
Start here
- Read
references/package-overview.mdfor package names, installed surfaces, backend boundaries, and the sub-skill map. - Run
scripts/check_fate_install.pyto summarize the active Python environment before giving commands that depend on installed FATE packages.- Add
--include-servicewhen the user needsfate_flow,pipeline, or FateFlow-backed Pipeline workflows. - Add
--strictonly when the task should fail on missing optional/service surfaces.
- Add
- Route to the smallest matching sub-skill below; keep root guidance for package-wide setup, routing, and cross-cutting troubleshooting.
- If a user reports behavior from a newer or older checkout, compare against
references/repo-provenance.mdbefore assuming this skill is current.
Sub-skill routing
| User task | Read this first |
|---|---|
| Install FATE, choose PyPI vs Docker/Compose/host deployment, initialize or check FateFlow services, diagnose ports/Docker/SSH/service startup | sub-skills/deployment/SKILL.md |
Upload data, build FateFlowPipeline DAGs, use Reader/PSI/preprocessing/train/evaluate/deploy/predict, validate upload YAML |
sub-skills/pipeline-workflows/SKILL.md |
Run without FateFlow using fate.arch.launchers, multiprocessing parties, dataframe readers, or direct fate.ml trainers |
sub-skills/local-launchers/SKILL.md |
Inspect component descriptors, component list, hyphenated task-schema, artifact types, task config schema, or custom component discovery |
sub-skills/component-runtime/SKILL.md |
Root operating boundaries
- The verified baseline is a CPU Python inspection/runtime surface. Do not claim GPU, DeepSpeed, Spark, Eggroll cluster, RabbitMQ, Pulsar, or Docker-cluster execution is verified unless the user provides and you verify that backend.
- Service-backed workflows require a running FateFlow service and initialized
pipelineclient. Local launchers and component CLI inspection do not require FateFlow. - Treat deployment shell scripts and cluster rollout scripts as reference workflows by default. Do not run destructive service, Docker, SSH, or root OS mutation commands unless the user explicitly requests that action and the target is clear.
- The bundled references and scripts are self-contained for operating guidance. Do not rely on the original repository checkout for runtime instructions.
- When in doubt between Pipeline and local launcher APIs, ask whether the user wants a FateFlow-backed job or a service-free local simulation.
Repository-wide references
references/package-overview.md— package/import/CLI surfaces, install footprint, backend limits, and capability ownership.references/troubleshooting.md— cross-cutting install/import, service, CLI, data/config, and backend diagnostics.references/repo-provenance.md— source commit, package versions, dirty-state note, and relative evidence paths for staleness checks.references/repo-routing-metadata.json— structured router placement metadata for this repo skill.
Safe root helper
python scripts/check_fate_install.py
python scripts/check_fate_install.py --include-service --strict
python scripts/check_fate_install.py --json
The helper only imports packages, checks versions, runs help commands, and reports backend/service surfaces. It does not start services, contact FateFlow, upload data, launch training, or mutate the host.
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).