[Documentation] How to get correct outputTensor shape in C++?

Describe the documentation issue

I have inference on a model like this:

the input is fixed, while in model the output didn't have a ostensive output shape but in python I get this OK, since I can get the correct output.

But in C++, I need to get certain output of tensor (or at least get from model,) then some operation can be done.

I get the output shape from ORT C++ is 0:

:69] [1,0,4,]
D 9/28 11:4:27.954 ...core/backends/ort.cc infer:69] [1,0,]
D 9/28 11:4:27.954 ...core/backends/ort.cc infer:69] [1,0,]

how to do it?

std::vector<Ort::Value> ort_inputs;
  for (int i = 0; i < inputNames.size(); ++i) {
    ort_inputs.emplace_back(Ort::Value::CreateTensor<float>(
        memoryInfo, static_cast<float *>(inputs[i].data), inputs[i].get_size(),
        inputShapes[i].data(), inputShapes[i].size()));
  }

  std::vector<Ort::Value> outputTensors =
      session.Run(Ort::RunOptions{nullptr}, inputNames.data(),
                  ort_inputs.data(), 1, outputNames.data(), outputNames.size());

Page / URL

https://stackoverflow.com/questions/73875800/onnxruntime-c-how-to-get-outputtensor-dynamic-shape

Start with the C++ entry points shown in the report: Ort::Value::CreateTensor and session.Run, and compare the reported output shapes with the linked Stack Overflow question. Document how a C++ caller should obtain dynamic output tensor shapes and identify what “correct output” means for this case; the documentation should address the reported zero dimensions.