This repository is dedicated to providing examples of how to use the OAAX runtime and conversion toolchain. The examples are provided is self-contained folders, each containing a README file that explains how to build and run the example.
Please note that all examples can be used with any OAAX-compliant runtime. Hence, to run a model on different AI accelerator, you only need to replace the runtime library with the one that is compatible with the accelerator.
The repository is structured as follows:
- c-example: Contains an example of how to use the OAAX runtime in a C program.
- python-example: Contains an example of how to use the OAAX runtime in a Python program.
- parallel-models: Contains an example of how to use the OAAX runtime to run multiple models in parallel.
The examples provided assume that the optimized model file has been generated using a compatible conversion toolchain.
All conversion toolchains can be found in the contributions repository.
For example, to optimize an ONNX model using the reference conversion toolchain, you can run the following commands:
wget -c https://download.sclbl.net/OAAX/toolchains/conversion-toolchain-latest.tar
docker load -i conversion-toolchain-latest.tar
docker run -v $(pwd):/run -it conversion-toolchain:latest /run/model.onnx /run/The last command takes a model file available at $(pwd)/model.onnx and
generates an optimized model located in the /run/ folder.
The generated model is compatible for use by only the OAAX reference runtime.
You can find other contributed conversion toolchains in the contributions repository, in addition to documentation on how to use each one of them.