Goal: Understand how Conda’s pip works with a User’s Python Pip Installs.
Quick Note: Default Python Version with a Conda Environment
Miniconda does ship with a default version of Python.
[salexan5@mblog1 conda]$ cd /project/arcc/salexan5/conda [salexan5@mblog1 conda]$ module load miniconda3/24.3.0 [salexan5@mblog1 conda]$ conda create -p py_env ## Package Plan ## environment location: /cluster/medbow/project/arcc/salexan5/conda/py_env # To activate this environment, use # $ conda activate /cluster/medbow/project/arcc/salexan5/conda/py_env [salexan5@mblog1 conda]$ conda activate /cluster/medbow/project/arcc/salexan5/conda/py_env (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ python --version Python 3.12.2 (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ which python /apps/u/opt/linux/miniconda3/24.3.0/bin/python
Conda Environments and pip Installs
Installing non-conda packages:
If a package is not available from conda or Anaconda.org, you may be able to find and install the package via conda-forge or with another package manager like
pip.Pip packages do not have all the features of conda packages and we recommend first trying to install any package with conda. If the package is unavailable through conda, try finding and installing it with conda-forge.
If you still cannot install the package, you can try installing it with pip. The differences between pip and conda packages cause certain unavoidable limits in compatibility but conda works hard to be as compatible with pip as possible.
Note:
Both pip and conda are included in Anaconda and Miniconda, so you do not need to install them separately.
It is possible to have pip installed outside a conda environment or inside a conda environment.
Try Installing numpy into our Conda environment
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pip install numpy Defaulting to user installation because normal site-packages is not writeable Requirement already satisfied: numpy in /home/salexan5/.local/lib/python3.12/site-packages (1.26.4) (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ python ~/py_test.py Python: 3.12.2 | packaged by Anaconda, Inc. | (main, Feb 27 2024, 17:35:02) [GCC 11.2.0] Numpy: 1.26.4
What’s happened?
Conda’s pip will look under the userbase for existing packages.
In this case ~/.local/lib/python3.12/site-packages/ where it will find the previously installed version of numpy/1.16.4
Confirm what is Currently Installed/Available
Check for conda installed packages:
# This list should be empty since we have essentially a blank conda environment # since we have not conda installed anything. (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ conda list # packages in environment at /cluster/medbow/project/arcc/salexan5/conda/py_env: # # Name Version Build Channel
Check for pip installed packages:
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pip list -v Package Version Location Installer ----------------------- ----------- ---------------------------------------------------------------- --------- anaconda-anon-usage 0.4.4 /apps/u/opt/linux/miniconda3/24.3.0/lib/python3.12/site-packages conda archspec 0.2.3 /apps/u/opt/linux/miniconda3/24.3.0/lib/python3.12/site-packages conda ... numpy 1.26.4 /home/salexan5/.local/lib/python3.12/site-packages pip ... pip 23.3.1 /apps/u/opt/linux/miniconda3/24.3.0/lib/python3.12/site-packages ...
Conda’s pip will check for package installs under the ~/.local/lib/PythonX.Y folder.
If the python package already exists it will be used.
Force a numpy update
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pip install numpy==2.0.0
Defaulting to user installation because normal site-packages is not writeable
...
Installing collected packages: numpy
Attempting uninstall: numpy
Found existing installation: numpy 1.26.4
Uninstalling numpy-1.26.4:
Successfully uninstalled numpy-1.26.4
Successfully installed numpy-2.0.0
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pip list -v
Package Version Location Installer
----------------------- ----------- ---------------------------------------------------------------- ---------
...
numpy 2.0.0 /home/salexan5/.local/lib/python3.12/site-packages pip
...
numpy has been updated, but still lives under: ~/.local/lib/python3.12/
Create a Self-Contained Conda Environment
We can force our conda environment to only use what is installed within it, by setting the PYTHONUSERBASE environment variable:
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ echo $CONDA_PREFIX /cluster/medbow/project/arcc/salexan5/conda/py_env (/cluster/medbow/project/arcc/salexan5/conda/py_env_3.12.4) [salexan5@mblog2 conda]$ export PYTHONUSERBASE=$CONDA_PREFIX (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pip list -v Package Version Location Installer ----------------------- -------- ------------------------------------------------------------------------------- --------- ... numpy 2.0.0 /cluster/medbow/project/arcc/salexan5/conda/py_env/lib/python3.12/site-packages pip ...
Conda’s pip now only looks under, and installs under, it’s own site-packages folder:
(/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ pwd /project/arcc/salexan5/conda (/cluster/medbow/project/arcc/salexan5/conda/py_env) [salexan5@mblog1 conda]$ ls py_env/lib/python3.12/site-packages/ numpy numpy-2.0.0.dist-info numpy.libs