I am currently trying to run my python script on my iOS application. My script is an ML application that consists of training through a given data set in real-time with dependencies on NumPy, cv2, torch, and torch-vision.
I could run the application on Android through Chaquopy successfully. But although several solutions are tried including Pythonista, Pyto, Beeware/briefcase, Kivy, and Pytorch-mobile, none of them worked yet for iOS.
Today, I am having difficulties understanding the reason for Pythonista's lack of support for various python packages that are already available for both desktop and android.
Is it something the source codes should be ported for iPhone CPU`s?
In the near term, should I wait for the required support to be given or continue to dig in?
Here are some references and my other questions as well:
-
https://forum.omz-software.com/topic/5328/can-pytorch-be-in-pythonista
-- It says that pytorch has distributed specific to each OS and architecture. -
https://pypi.org/project/torch/#description
-- PyPi says that the pytorch needs at least these dependencies:numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing_extensions future six requests dataclasses
-- It has been seen from here that http://omz-software.com/pythonista/docs/ios/index.html, only limited subset looks available for Pytonista. Does it mean that path is too long to expect quick resolution? -
https://stackoverflow.com/questions/50167270/how-to-compile-a-whole-python-library-including-dependencies-so-that-it-can-be
-- Although it is demoted, Could compiling the python into the static library with dependencies in a way that it can be basically called by a C program?
Thanks in advance.