Best practice for developing Julia packages with multi-language supports (Python)

Hi, I’m trying to make a Julia package that has multi-language supports, especially Python.


The following repo exports some predefined neural networks, compatible with Flux.jl.

I’d like to make the neural networks PyTorch-compatible.
So, I thought that two options may be available.

  1. Using some packages, export networks from Flux.jl to PyTorch.
  2. Write a new script for PyTorch-compatible neural networks.

I tried to do 1. but I failed it. So I’m trying to write a new script for PyTorch-version neural networks. That is, some scripts like


  1. Is it a good idea to add some Python scripts within a Julia package?
  2. Is there any alternative way to accomplish what I desire? Like, exporting Flux.jl-based neural networks to a PyTorch one.