Passing result of torch.jit.script to C++ function

C++
1

I have the following Python code:

import torch.nn as nn
import torch as pt
import myCppModule

class Model(nn.Module):
    def __init__(self):
        super().__init__()

model = pt.jit.script(Model())
myCppModule.myCppFoo(model)

I want to implement some function in C++, myCppFoo, that takes “model” as argument. And then expose it to python with something like pybind11
I tried to implement this with:

#include<torch/torch.h>
#include<pybind11/pybind11.h>
#include<iostream>
void myCppFoo(const torch::jit::Module& module){
  std::cerr<<module.get_properties().front().name<<std::endl;
}
PYBIND11_MODULE(myCppModule, m) {
  m.def("myCppFoo", &myCppFoo);
}

However, when compiling the python library and running the python example, I get an error:

Traceback (most recent call last):
  File "/home/user/test/ex.py", line 10, in <module>
    myCppModule.myCppFoo(model)
TypeError: myCppFoo(): incompatible function arguments. The following argument types are supported:
    1. (arg0: torch::jit::Module) -> None

Invoked with: RecursiveScriptModule(original_name=Model)

Clearly the types are not compatible. How can I modify the C++ function to accept the type provided by torch.jit.script?
The docs suggest to serialize the module to a file in Python and deserialize in C++, but I would like to avoid writing an intermediate file.