I am running an ONNX model through TensorRT. I can verify that inference is running on the GPU through the results and nvsys profile logs. However, I would like to see the corresponding PTX binary that TensorRT generates for my input model.

Is there a specific flag or argument to do so?