Hello World (Using InterpreterSession directly)

Here we use torch::deploy to print Hello World to the console without using torch.package. Instead we simply acquire an individual InterpreterSession, and use it to print Hello World directly.

Printing Hello World with torch::deploy

#include <multipy/runtime/deploy.h>
#include <multipy/runtime/path_environment.h>
#include <torch/script.h>
#include <torch/torch.h>

#include <iostream>
#include <memory>

int main(int argc, const char* argv[]) {
  // create two interpreters
  multipy::runtime::InterpreterManager manager(2);

  // Acquire a session on one of the interpreters
  auto I = manager.acquireOne();

  // from builtins import print
  // print("Hello world!")
  I.global("builtins", "print")({"Hello world!"});
}

Here we introduce the pythonic nature of torch::deploy’s InterpreterSession``s by using them in a similar fasion to python objects. This allows us to add further flexibility to the code exported by ``torch.package by interacting with it in C++.

manager.acquireOne allows us to create an individual subinterpreter we can interact with.

InterpreterSession::global(const char* module, const char* name) allows us to access python modules such as builtins and their attributes such as print. This function outputs an Obj which is a wrapper around print. From here we call print by using {"Hello world!"} as its argument(s).

Build and execute

Assuming the above C++ program was stored in a file called hello_world_example.cpp, a minimal CMakeLists.txt file would look like:

cmake_minimum_required(VERSION 3.12 FATAL_ERROR)
project(multipy_tutorial)

set(MULTIPY_PATH ".." CACHE PATH "The repo where multipy is built or the PYTHONPATH")

# include the multipy utils to help link against
include(${MULTIPY_PATH}/multipy/runtime/utils.cmake)

# add headers from multipy
include_directories(${MULTIPY_PATH})

# link the multipy prebuilt binary
add_library(multipy_internal STATIC IMPORTED)
set_target_properties(multipy_internal
    PROPERTIES
    IMPORTED_LOCATION
    ${MULTIPY_PATH}/multipy/runtime/build/libtorch_deploy.a)
caffe2_interface_library(multipy_internal multipy)

# build our examples
add_executable(hello_world_example hello_world/hello_world_example.cpp)
target_link_libraries(hello_world_example PUBLIC "-Wl,--no-as-needed -rdynamic" dl pthread util multipy c10 torch_cpu)

From here we execute the hello world program

mkdir build
cd build
cmake -S . -B build/ -DMULTIPY_PATH="<Path to Multipy Library>" -DPython3_EXECUTABLE="$(which python3)" && \
cmake --build build/ --config Release -j
./hello_world