[seemslegit]

Why did you take up conda in the first place ? venv and pip are part of the python project and conda seems to be yet another package manager in the already crowded and shady space of npm, brew, ports apt yum etc. I've always been puzzled at how they managed to shoehorn themselves into the python ecosystem while offering no distinct value as far as I can see.

[downerending]

This is an HPC environment, so besides a nice, working version of TensorFlow, there are large numbers of pre-built scientific/bioinformatics packages. This is a huge win when you're running with a skeleton crew.

Also, note this is for system-wide, not per-user usage. I'm kind of vague on how virtualenv would handle this, plus it feels like it's changed a lot over time. (?)

[woodson]

They provide useful packages that aren't (easily) available via pip. For example, numpy with MKL or Tensorflow GPU builds that don't require/presuppose CPU support for new vector instructions. Moreover, their package installs don't break as randomly as some pip packages, presumably because someone bothered to do more testing of commonly used packages included in the default package repository.