Conda Skeleton为不同的Python版本安装相同的包

Question

I want to install sklearn-pandas for data science on my conda environments so I switch to my Python2 Conda environment and use skeleton to build it in:

conda skeleton pypi sklearn-pandas
conda-build sklearn-pandas
conda install --use-local sklearn-pandas

The problem is, this builds in a bunch of python 3 dependencies, and at the end of it all when I run python --version I get

Python 3.6.6

This means that it had to upgrade the python version of the environment to accommodate the package it installed.

I think this library should be compatible with Python 2, it even says in its Git repo "build universal wheel compatible with python 2 and python 3".

What arguments do I need to pass to conda skeleton, build, and install so that it installs and builds it for Python 2 and doesn't do this unwanted upgrade?

Answer 1

Figured this out - you only need to add one flag to the conda-build step. It's:

conda-build --python 2.7 sklearn-pandas

Although in the conda skeleton stage of the installation it appears to bring in py36 packages, once you do this conda-build step with the flag it replaces them with py27.