All Tools
P
Dev ToolsFreeOpen Source
PADDLE2ONNX
Export PaddlePaddle models to ONNX for multi-runtime deployment
Apache-2.0
ABOUT
PaddlePaddle checkpoints do not run natively on TensorRT, OpenVINO, or mobile ONNX engines. Paddle2ONNX exports the graph and weights to ONNX so CV and NLP teams deploy the same model across those runtimes without a custom converter per backend.
INSTALL
pip install paddle2onnxINTEGRATION GUIDE
1. Convert a PaddleDetection or PaddleOCR model to ONNX for TensorRT serving
2. Export PaddlePaddle graphs to OpenVINO for CPU inference
3. Move a Paddle model onto MNN, TNN, or NCNN for mobile deployment
4. Check and optimize exported ONNX graphs before production serving
TAGS
pythononnxpaddlepaddlemodel-exporttensorrtopenvinoinference