Hi @kirsebom, are you still working on this merge? I remember there were some dependency issues with scipy, but I don't recall exactly what they were

Yes, this branch requires Scipy >=1.6, which requires Python >=3.7.

I am reluctant to enforce Python >=3.7. What do you think? Perhaps we could implement the solution that we discussed with Matt?

Also, related to this, can we change the specifications in setup.py so we allow for Python 3.8? I believe tensorflow has a version compatible with python 3.8 now, right?

Yes, python 3.8 is now supported. I can change for the next release.

What was the idea there with Matt again? Only install scipy 1.6 when the gammatone module is imported?

Something like that, yes. I don't remember Matt's precise suggestion. Let's ask him again today at the weekly meeting.

Ok, Let's ask him.

I don't think that installing a new package on the user's environment is a good idea because it might mess other things up if that environment is not being used exclusively for ketos. A less aggressive approach would be to just give an error message. Something like:

"You are trying to use a feature that requires scipy>1.6 and python>3.7. Please try again in an environment that meets these requirements."

We can even give more detailed instructions on how to prepare an environment with pyenv or anaconda.

I like your last suggestion!

OBS: Development of Gammatone-related features is now taking place on the nicole branch...