When conda was updated to version 22.9, the environment where PyMC was installed (among others) was broken. I remember searching the issue and finding it was due to the new conda version, so downgrading it to v4.14.0 fixed the problem. I am wondering if the issue still exists and, in case, if someone knows if the conda team is aware of it and plans to address it.
(Full disclosure here: I wasn’t able to find any reference to the problem I faced a few weeks back when I ventured in a new search online and certainly don’t want to update conda to 22.9 to try to reproduce the issue as it wasn’t painless to downgrade and rebuild my environments…)
1 Like
Welcome!
Not sure if it helps, but many have moved to mamba for various reason (performance benefits being a big one). But I am not sure about the 22.9 version issue you referred to.
Thank you for your welcome and suggestion!
TBH, I would like to avoid having to deal with mamba, which I believe still depends on conda, to keep the system as tidy and neat as possible. Of course, if that’s the only way to go going forward and PyMC will require mamba, I’ll migrate too, but for now I’d like to stick to conda.
Back to my original question, if there is anyone who’s using conda and that experienced the same issue I faced when upgraded to version 22.9, do you know if a solution exists and/or if the issue will be addressed in the next conda release?
Just as a point of clarification, mamba can be installed into a conda environment if you already have conda installed, but mambaforge is (I believe) a drop-in replacement for miniconda. I don’t install conda anymore and am all the happier for it. But conda is definitely the “official” recommendation for installation of pymc, so whatever is going on should definitely be remedied.
1 Like