optimization-workflows
Route users who need flatten/unflatten helpers, structured optimizers, fixed-point differentiation helpers, and SciPy minimize integration using value_and_grad.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- optimization-workflows
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Route users who need flatten/unflatten helpers, structured optimizers, fixed-point differentiation helpers, and SciPy minimize integration using value_and_grad.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档references/api-reference.md3.4 KB
- 参考文档references/troubleshooting.md3.0 KB
- 参考文档references/workflows.md3.0 KB
- 脚本scripts/fixed_point_smoke.py936 B
- 脚本scripts/rosenbrock_minimize.py679 B
- 脚本scripts/structured_optimizer_smoke.py3.4 KB
03 SKILL.md 正文
Optimization Workflows
Use this sub-skill for optimization tasks built from nested parameter trees, fixed-point recurrences, or SciPy minimization loops.
Route here when the user needs
flattenorflatten_functo turn nested dict/list/tuple parameter trees into flat arrays and back.- Structured optimizers such as
sgd,rmsprop, oradamon nested containers. fixed_pointfor differentiable iterative updates.scipy.optimize.minimizecombined withvalue_and_grad.
Route elsewhere when the question is really about
value_and_gradsemantics, scalar-output constraints, or Jacobian/Hessian selection: differentiation-core.- General
autograd.scipywrapper behavior or SciPy wrapper coverage: numpy-scipy-primitives. - Custom primitive or staged VJP/JVP mechanics: extend-primitives.
Bundled runtime helpers
scripts/structured_optimizer_smoke.pyscripts/rosenbrock_minimize.pyscripts/fixed_point_smoke.py
These scripts use tiny synthetic fixtures only; they do not download data and there is no CLI entry point to discover.
What this sub-skill does not own
- Ordinary differentiation operators and higher-order autodiff semantics.
- Custom primitive registration or gradient-rule authoring.
- General NumPy/SciPy wrapper troubleshooting outside optimization workflows.
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).