---

name: export-deployment description: "Export YOLOv3 weights and choose deployment formats across TorchScript, ONNX, OpenVINO, TensorRT, CoreML, and Paddle." disable-model-invocation: true metadata: disco-role: operating license: AGPL 3.0

Export and Deployment Sub-skill

Read this for export.py, deployment formats, DetectMultiBackend suffixes, backend-specific dependencies, ONNX/OpenVINO/TensorRT/CoreML/Paddle choices, and export troubleshooting.

Use

• Read references/export-formats.md for supported format matrix and backend constraints.
• Use scripts/yolov3_export_format_matrix.py to print or validate format names without exporting weights.
• Read references/troubleshooting.md for optional dependency, CPU/GPU, and unsupported TensorFlow export issues.

Important facts

• Public export command pattern: python export.py --weights yolov3-tiny.pt --img 64 --include torchscript.
• CPU-friendly export formats include TorchScript, ONNX, OpenVINO, CoreML, and Paddle when dependencies exist.
• TensorRT .engine export is CUDA/GPU-only.
• TensorFlow rows are retained in export_formats() for suffix detection, but TensorFlow export is not implemented by this repo.