[Question] about how to get IR version (onnx surgeon)

[phamkhactu]

Thanks for excellent repo.

I've encounter with problem. I convert model successfully from onnx to onnx surgeon graph. However, when I infer it, I get error:

    sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model)
onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Load model from tts_surgeon.onnx failed:/onnxruntime_src/onnxruntime/core/graph/model.cc:146 onnxruntime::Model::Model(onnx::ModelProto&&, const PathString&, const IOnnxRuntimeOpSchemaRegistryList*, const onnxruntime::logging::Logger&, const onnxruntime::ModelOptions&) Unsupported model IR version: 9, max supported IR version: 8
``

My version:

• onnx: 1.14.1
• onnxruntime: 1.14.1
• onnx surgeon: 0.5.2

With my check version from [onnxruntime](https://onnxruntime.ai/docs/reference/compatibility.html), I know that IR 8 is suitable for onnxruntime, but i dont know the way to check.

Thanks.

[phamkhactu]

I change to IR 9 with onnxruntime=1.16.0 fixed this issue. But the time inference between pytorch --> onnx and onnx --> surgeon onnx the same. What do you think about it?

[zerollzeng]

Sorry I don't quite get the question, your issue seems caused by onnxruntime, am I correct?

[phamkhactu]

Hi @zerollzeng, Yes, it causes by version, and I fixed it by changing suitable IR version