Installing Tensorflow and Keras in the base container


#1

Would it be possible to have Tensorflow and Keras installed in the base container?

I am currently working on a Flask server that is designed to serve a keras model. Part of this is going to be discussed in a conference talk I am giving next week.

Right now even just attempting to install 1 of these packages exceeds the storage quota.

Locally if I install Flask, gunicorn, tensorflow and keras it comes to 491M.

I have got the requirements compiled to a slug under Heroku, and just about got the requirements installed with pythonanywhere if I swap out tensorflow for theano. However one of my goals is to have an example that others can jump right into a working copy and adapt to their needs, which is why I love Glitch. Heroku and pythonanywhere don’t even come close to this flexibility.

I completely appreciate that Python is not officially supported and understand if this wouldn’t be possible

Thanks :blush: