problem-factories
Routes built-in optimizee problem selection, data-backed problem setup, custom loss factories, and util.get_config mappings.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- problem-factories
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Routes built-in optimizee problem selection, data-backed problem setup, custom loss factories, and util.get_config mappings.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/data-and-custom-problems.md3.6 KB
- 参考文档references/problem-catalog.md5.0 KB
- 参考文档references/troubleshooting.md2.6 KB
- 脚本scripts/inspect_problem_config.py13.7 KB
03 SKILL.md 正文
problem-factories
Use this sub-skill when you need to choose, inspect, or author an optimizee problem in learning-to-learn.
Route here for
- Built-in factories:
simple,simple_multi_optimizer,quadratic,ensemble,mnist, andcifar10. - README-facing problem names:
simple,simple-multi,quadratic,mnist,cifar, andcifar-multi. util.get_config(problem_name, path=None)mappings, checkpoint path rules, and optimizer-network assignments.- Custom problem factories that need dataset setup, queue creation, or stable variable naming.
Do not handle here
- Meta-loss construction, variable interception, or optimizer assignment mechanics. Use the
meta-optimizer-apisub-skill. - Training/evaluation CLI loops, epoch scheduling, or save/eval flow. Use the
training-evaluationsub-skill. - Network class internals or preprocessing modules. Use the
optimizer-networkssub-skill.
Read first
references/problem-catalog.mdreferences/data-and-custom-problems.mdreferences/troubleshooting.mdscripts/inspect_problem_config.py
Bundled script
scripts/inspect_problem_config.py— inspect a problem name, print a safe config summary, and avoid MNIST/CIFAR downloads by default.
Working rule
- Pick the public problem name or source factory you actually need.
- Check the catalog entry for its variables, side effects, and util mapping.
- If the problem is data-backed, confirm whether a path will trigger dataset loading or checkpoint lookup.
- Keep Python side effects outside the returned zero-arg loss builder.
- If the issue is really optimizer routing or training flow, hand off to the sibling sub-skill instead.
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).