Kay Hayen

I will have yours as an excellent cheat sheet though! It's just that I will review it on the way for necessity. For example, the scipy one, I was wondering...

btw, what a wonderful code ``` try: import cython # type: ignore except ImportError: compiled: bool = False else: # pragma: no cover try: compiled = cython.compiled except AttributeError: compiled...

That is then used to potentially create a string, my goodness. And for that, it loads Cython in full glory into memory.

It's also an experiment, I might just fail to do it enough, lets see if it ever finishes for me, lol

So, it seem tensorflow can integrate with kubernetes, and we don't have to include it though by default, so hopefully that gets rid of a bunch of thing, I imagine...

As an aside, the yaml looks good to me. For me it seem PIL is being picked up, so I have some doubt about those top level dependencies, the other...

ok, got that down to 11k modules, but I noticed, I was using `''` for the no follow mode, kind of thinking that would be `'ignore'`, but it outputs this...

Something in the yaml is very wrongly formatted, in my visual code it became all red, missing dashes, wrong indentation, etc.

So, the crash is happening during import of `numpy` of all things, and only in a clean environment, not sure what that is about, it could well be a corruption...

Reduced to this: ``` import imgaug.augmenters.segmentation print("OK") ``` Since numpy is suspected to work, it cannot be that modules code itself: ``` from __future__ import print_function, division, absolute_import from abc...