I possess an ONNX model stored at a specific path, which is integrated into an established software pipeline where modifications are restricted. My objective is to develop a deep learning model and evaluate its compatibility with the existing pipeline. Currently, I am trying to create a dummy ONNX model that mimics the input and output names, data types, and shapes of the current ONNX model that will produce a constant-valued output (for the sake of testing). Kindly inform me if there are additional aspects that I need to consider.
I am looking for a Python code, preferably based on PyTorch if it is feasible or necessary.