export
Routes Chainer model export workflows for ONNX-Chainer and Caffe.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- export
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Routes Chainer model export workflows for ONNX-Chainer and Caffe.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/api-reference.md1.8 KB
- 参考文档references/troubleshooting.md1.7 KB
- 参考文档references/workflows.md1.9 KB
03 SKILL.md 正文
Export
Use this sub-skill when the user wants to export a Chainer model to ONNX or Caffe, generate export testcases, or debug conversion failures.
Typical requests
- "Export this model to ONNX."
- "How do I generate ONNX testcases?"
- "Why does
onnx_chainer.exportfail on my model?" - "How do I export a Chainer model to Caffe?"
- "Why is
chainer_model.prototxtorchainer_model.caffemodelmissing?"
Read these first
references/workflows.mdfor the export flow.references/api-reference.mdfor the exact export signatures and limits.references/troubleshooting.mdfor exporter-specific failure modes.
Use this script
../../scripts/export_smoke.pyfor a tiny ONNX and Caffe export check.
Include here
onnx_chainer.export(...)andonnx_chainer.export_testcase(...)- ONNX opset version limits and named inputs / outputs
- Caffe export via
chainer.exporters.caffe.export(...) - Export validation, file layout, and unsupported-layer debugging
Route elsewhere
- Ordinary model training or serialization ->
../training/ - MPI / distributed / multi-node export scenarios ->
../distributed/ - ChainerX-specific model behavior ->
../chainerx/
Quick mental model
A normal export workflow is:
- Build a tiny model or load a trained model.
- Feed a representative input.
- Export the model to ONNX or Caffe.
- Validate that the target files exist and the exported graph is structurally valid.
When you only need a fast check, use the bundled smoke script.
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).