The header `trt_plugin_python.h` does not seem to be included with TensorRT or with this repo

[matthewfl]

The header trt_plugin_python.h isn't included with TensorRT distribution, and does not seem to be included with this repo. Where is this header supposed to come from?

https://github.com/NVIDIA/TensorRT-Incubator/blob/main/mlir-tensorrt/tensorrt/lib/Utils/NvInferPluginUtils.cpp#L35-L37

[matthewfl]

Looks like the header was renamed to NvInferPythonPlugin.h in the released version

[christopherbate]

Hi Matthew, it looks like an update to the CMake build system for handling that file didn't get sync'd out. I will make the fix.

The way it should be working is: we look for that file under the various names it has taken in the past several TensorRT releases. We then copy it to the trt_plugin_python.h in the build directory so that it can be referenced under a common name.

[matthewfl]

Thanks for the update. Suggestions, might it be better to use #if MLIR_TRT_COMPILE_TIME_TENSORRT_VERSION_GTE(10, 9, 0) against the TRT version inside of that file instead of using cmake to copy/link the file?