Does installing packages with a Python version, then downgrading it cause compatibility problems (e.g. module not found)?

TLDR: Does using a python version other than that which was used to initially install the packages in a conda env lead to compatibility problems?

In after creating a conda environment with python=3.11, and installing several packages from a requirements.txt file, all libraries except one were successfully installed. After trying different ways to install the last library, I figured that it might be a python version issue and so I executed 'conda install python=3.10' and retried. The last package was successfully installed but, after trying to run the python program I get an error saying that another library cannot be found. Is that due to the fact that packages were installed using a later version than the ones used to run the program?

Tried to check if it is a python interpreter issue in VS Code. It was not.

The comments highlight important aspects, but don't explain the mechanics. So, let's walk through it.

Conda and Pip Installation Behavior

When you create a Conda environment with python=3.11 it will install Python 3.11 and Pip. Within the environment, it creates a lib/python3.11/site-packages directory, which is where subsequent Python modules (the importable code chucks that Python packages deliver) will be installed.

If the requirements.txt file was a Pip requirements file, then the installation with implicitly run python -m pip install -r requirements.txt and all packages will be installed to the lib/python3.11/site-packages.

Also, any Python packages you install with Conda will similarly get installed there, but Conda does the extra bookkeeping of tracking what you have defined as your environment (python=3.11, plus all Conda packages). It does not track the Pip-installed packages.

Conda-only just works

If you only had Conda-installed packages and then have Conda downgrade to python=3.10, it will also correctly uninstall all the lib/python3.11/site-packages and reinstall with matching lib/python3.10/site-packages versions.

Conda ignores Pip-installed packages

However, it's a different story for Pip-installed packages. Conda will not adjust these because they aren't included in Conda's bookkeeping. Hence, you may still have a bunch of packages residually left in the lib/python3.11/site-packages but they are no longer usable by the python=3.10 that you have installed.

So, you must reinstall any Pip-installed packages.

Better Strategy: YAML Environment Definition

A better strategy is to use a YAML file to manage everything. Something like:

my_py310.yaml

name: my_py310
channels:
  - conda-forge
dependencies:
  - python=3.10
  - pip

  ## Pip-installed packages
  - -r requirements.txt

then use

conda env create -n my_py310 -f my_py310.yaml

This way, if you had to downgrade, you can edit or create a new version of this file and then create a new environment from it with one command.

Any environment that mixes Conda and Pip packages is recommended to use this approach as per the documentation.