Oh man, this is my life right now, too. In my case, we're using tensorflow or tensorflow-gpu, depending on the host system and, unfortunately, only Conda offers tensorflow-gpu with built-in CUDA. Add to this that the tensorflow packages themselves are notoriously bad at specifying dependencies and that different versions of tensorflow(-gpu) are available on conda-forge, depending on your OS.

Tensorflow is the worst (along with ReAgent from FB).

I think it's because they have their own internal build systems, but they never play well with pip/conda et al.

One of my recent breakages was installing the recsim package, which pulled in tensorflow and broke my entire app. There's actually a recsim-no-tf package on PyPi, presumably because this happens to loads of people.

I see, I miss a lot of issues as don't use any GPU stuff, mainly flask + scipy and friends.... probably this is what saves me.

It's not even the GPU versions, even the CPU stuff causes issues.

The core problem is that pip will happily overwrite your existing dependencies when you attempt to install a new package.