Onnx dynamic load

Introduction

This is a wrapper used to load onnxruntime dynamically to inference model.

Feature

• No c interface exposed.
• Library loader is thread-safe.
• OnnxApiWrapper is thread-safe and support multi-thread model infer.
• Dynamic library load.
• Library closed once unneeded, no more resources occupation.

Usage

1. Download: git clone --recursive https://github.com/atp798/onnx_dynamic_load.git

2. Build or copy library named 'onnxruntime_pybind11_state.so' to one of the path in environment variable 'LD_LIBRARY_PATH'.

3. Build: mkdir build && cd build && cmake ../ && make

4. Run sample: ./onnxruntime_dynamic_loader.out

License

MIT license (© 2020 ATP)