Is TVM only for Python? How can I use TVM to inference an ONNX model in C++?

Questions
1

Is there any documentation available on how to perform inference with an ONNX model in C++?

2

Yes, TVM does have interfaces for different languages. Most of the components of TVM are implemented in C++. I think a better solution is

  1. Import, convert, and compile the ONNX model in advance with Python. Now you will get a runtime executable.
  2. Then you can deploy the executable on the target device with TVM runtime (without Python).

You can refer to https://github.com/apache/tvm/tree/main/apps/bundle_deploy

1 Like
3

right, python is common used for compilation and deployment happens at the language of your choice