I know that using a shared install that doesn’t change can be done.
I can’t seem to create a shared install that will allow someone to create an env local to them and install packages not in the “base” install.
The error currently blocking me is
NotWritableError: The current user does not have write permissions to a required path.
The suggested fix of course is to chown to my test user that I’m using.
I’m still a newbie to Anaconda, so this may not be the right answer but I don’t see any other suggestions so…
Try looking into having the user set pkgs_dirs in thier .condarc. Ex:
There’s documentation on how to create a shared cache (Configuring a shared package cache — Anaconda documentation), but I found this to be the wrong way to go. In our case we just ended up with a shared mess of files with all different protections that no one person (aside root) could then control. YMMV.
Hey, that pushed into the right direction of things.
After trying some stuff I determined that if I create an env then if I install a package it will install/write/whatever in my local env and not the “base” env that no one has write access to.
Now I’m gonna re-read the documentation to see what I missed when I read through it before.
How useful that will be for our folks I’ll find out. Mostly to save on disk space (even 1GB of space extra used by 100+ users adds up, nevermind 5+GB for the “base” env I’m creating for them).
We point most folks to conda for their pytorch/tensorflow needs since it’s such a pain to install within a python install.