model-catalog
Select PINTO_model_zoo models by task family, model number/directory/name, format flags, and remarks without reopening source evidence.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- model-catalog
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Select PINTO_model_zoo models by task family, model number/directory/name, format flags, and remarks without reopening source evidence.
- disable-model-invocation
- true
02 资源文件 2
- 参考文档references/catalog-selection.md3.8 KB
- 参考文档references/troubleshooting.md2.5 KB
03 SKILL.md 正文
Model Catalog
Use this sub-skill when the task is to shortlist a model from the bundled catalog, interpret format flags, or map a loose request like ONNX hand pose or 132_YOLOX to a concrete folder.
Fast route
-
Check the license gate first: do not recommend a model for use, redistribution, or execution until the target folder's license terms are acceptable.
-
Query the bundled helper for a short list:
python ../../scripts/query_model_catalog.py --category "2D/3D Hand Detection" --format ONNX python ../../scripts/query_model_catalog.py --format OV --contains pose python ../../scripts/query_model_catalog.py --number 132 python ../../scripts/query_model_catalog.py --directory 132_YOLOX python ../../scripts/query_model_catalog.py --list-formats -
Read references/catalog-selection.md for category, flag, directory, name, and remark semantics.
-
Read references/troubleshooting.md when a filter returns no match, a folder number is ambiguous, a backend flag does not fit the target runtime, or the remarks imply a resolution/input constraint.
-
Hand off to
../model-acquisition/for downloads and artifact handling,../inference-demos/for running an existing model, or../conversion-and-deployment/for format conversion, quantization, or deployment planning.
Boundary with sibling sub-skills
- Need to find or rank candidates only -> stay here.
- Need download scripts, artifact acquisition, or cookie/network planning ->
../model-acquisition/. - Need to run an existing demo/test script or interpret runtime output ->
../inference-demos/. - Need to convert, quantize, export, or deploy a model artifact ->
../conversion-and-deployment/.
What stays here
- catalog lookup and shortlist generation
- format legend interpretation
- no-network model selection
- resolution/shape hint reading from remarks
- number/directory/name mapping
Bundled files
- references/model-catalog.json — the self-contained catalog data.
- references/catalog-selection.md — selection rules, filter examples, and flag meanings.
- references/troubleshooting.md — common failure modes and recovery steps.
- ../../scripts/query_model_catalog.py — offline query helper for the bundled catalog.
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).