You can use ONNXRuntime with a TensorRT backend, so one does not exclude the other.