model-export
Routes Torchreid checkpoint export to ONNX, OpenVINO, and TFLite-style artifacts with explicit model-name, input-shape, and optional dependency checks.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- model-export
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Routes Torchreid checkpoint export to ONNX, OpenVINO, and TFLite-style artifacts with explicit model-name, input-shape, and optional dependency checks.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/export-workflows.md4.3 KB
- 参考文档references/optional-dependencies.md2.2 KB
- 参考文档references/troubleshooting.md3.1 KB
- 脚本scripts/export_torchreid_model.py18.9 KB
03 SKILL.md 正文
model-export
Use this sub-skill when a Torchreid checkpoint must be packaged for deployment rather than trained or queried.
Use this route for
- Exporting a Torchreid checkpoint to ONNX.
- Chaining ONNX into OpenVINO and TFLite-style artifacts.
- Choosing input height/width, opset, dynamic axes, or FP16 conversion settings.
- Resolving model-name selection before export.
- Checking optional export dependencies and known limitations.
Do not use this route for
- Training a checkpoint first. Use
training-evaluation. - Feature extraction, embeddings, or model-key discovery only. Use
feature-extraction. - Project-specific model families that are not accepted by core
build_model; those project workflows are excluded long-tail gaps in this generated skill.
Read first
Skill-owned script
scripts/export_torchreid_model.py— argparse-driven helper with--helpand--dry-run, explicit--model-name,--weights,--imgsz,--include,--dynamic,--opset, safe optional-dependency checks, and non-destructive defaults unless--forceis supplied.
Typical workflow
- Confirm the checkpoint already exists and belongs to a core Torchreid model family.
- Supply
--model-namewhen the filename does not clearly identify the architecture. - Export ONNX first, then request OpenVINO and/or TFLite-style outputs only when the corresponding tools are installed.
- Verify the generated artifact path and any dependency or shape warnings.
Limits to keep explicit
--helpand--dry-runmust stay safe and side-effect free.- ONNX export can run on CPU when
onnxis installed. - OpenVINO and TFLite-style conversion remain optional and unverified unless their extra packages are present.
- Existing outputs should not be overwritten unless
--forceis supplied. - Do not claim a successful deployment export unless the artifact was actually written.
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).