deployment-export
Export PaddleGAN checkpoints to static inference models and plan deployment targets.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- deployment-export
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Export PaddleGAN checkpoints to static inference models and plan deployment targets.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/deployment-targets.md5.1 KB
- 参考文档references/export-and-inference.md6.1 KB
- 参考文档references/troubleshooting.md4.8 KB
03 SKILL.md 正文
deployment-export
Use this sub-skill when a PaddleGAN checkpoint must become a static inference artifact, when an exported model needs inspection, or when deployment flags and backend prerequisites need to be planned without running heavy validation by default.
Covers
- Checkpoint to static export.
- Export artifact inspection.
- Paddle Inference Python planning, including TensorRT flags.
- Serving, Lite, and C++ deployment boundaries at reference level.
- TIPC harnesses as validation evidence, not default execution.
Does not cover by default
- Training from scratch or resume flows ->
training-configs. - App-style predictor demos ->
image-and-face-appsorvideo-and-audio-apps. - Dataset building ->
data-preparation. - Mobile builds, Serving startup, or C++ compilation unless explicitly authorized.
- Heavy native inference or benchmark runs unless the task explicitly asks for them.
Bundled helpers
scripts/export_model.py— bundled export wrapper.scripts/check_exported_model.py— exported model directory/prefix checker.references/export-and-inference.md— export grammar, file naming, and inference flags.references/deployment-targets.md— Serving, Lite, C++, TensorRT, and TIPC boundaries.references/troubleshooting.md— common failures and recovery steps.
Quick workflow
- Identify the model family and confirm whether the checkpoint is single-net, multi-net, or custom-export.
- Confirm the
inputs_sizestring matches the number and order of exported inputs. - Confirm checkpoint keys match the expected network names.
- Export with the bundled wrapper.
- Check the resulting prefix tree with the bundled checker.
- Plan the runtime:
- standard Python inference for quick checks,
- TensorRT only when the installed Paddle wheel/lib supports it,
- Serving/Lite/C++ only when their prerequisites are known.
- Treat TIPC as evidence and routing context, not as the default execution path.
Special cases
- CycleGAN and similar multi-net base exports should keep distinct prefixes such as
netG_AandnetG_B; a forced shared prefix would clobber one export. - Wav2Lip uses a single
netGcheckpoint load path. - FirstOrder/FOM exports a paired
fom_dy2st/tree withkp_detectorandgeneratorprefixes. - TensorRT planning is only meaningful when the installed Paddle build exposes TensorRT support.
Start here
- Export rules, input-size grammar, and inference flags: references/export-and-inference.md
- Deployment targets and reference-only boundaries: references/deployment-targets.md
- Common export and inference failures: references/troubleshooting.md
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.