Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Conda & Mamba

Conda / Mamba (C++ reimplementation of Conda) is a package manager and environment management system. Can be replaced by Pixi.

Installation

Miniforge

https://conda-forge.org/download/

bash Miniforge3-Linux-x86_64.sh

Prevent conda from activating the base env by default

conda config --set auto_activate_base false

Stackoverflow Thread

Create default environments

Jupyter Lab environment

conda create -n jupyter nb_conda_kernels jupyterlab jupyterlab-git ipywidgets

Default data science environment

conda create -n default \
pandas polars geopandas statsmodels \
scikit-learn lightgbm xgboost catboost \
matplotlib seaborn plotly streamlit \
pydantic openpyxl fastexcel ipykernel

Want to change kernel in jupyter?

Base environment needs the nb_conda_kernels package. Every environment needs the ipykernel package.

conda install nb_conda_kernels

Updating Conda

Update base environment

conda update conda

Update all packages of activated environment

conda update --all

Environments

Create new environment and switch to it

conda create -n name_of_new_env
conda activate name_of_new_env

Create environment with specific python version:

conda create -n name_of_new_env python=3.12

Remove environment

conda deactivate # only relevant if environment to-be-deleted is activated

conda remove -n name_of_env --all
# alternative: conda env remove -n name_of_env

List all environments

conda env list

Adding channels to environment

Activate the environment, then:

conda config --env --add channels conda-forge

Sharing an environment

This can be useful if you want to document the package versions for a project or if you are collaborating with other people so they can work with the same environment.

conda env export > environment.yml

To exclude packages that were installed into the environment as a dependency add the flag --from-history. This can help for cross-platform compatibility, because conda will figure out platform-specific dependencies. However, this will not document the version of a package if the version was not explicitly stated when installing.

conda env export --from-history > environment.yml

To recreate the environment using the yml file:

conda env create -f environment.yml

Archived

Deep learning environment

Prereq:

sudo apt-get install libhdf5-serial-dev # for saving keras models efficiently

sudo apt install graphviz
conda create -n deeplearn
source activate deeplearn

conda install tensorflow-gpu

conda install matplotlib
conda install opencv # needed for some examples
conda install pydot # graphviz
conda install pillow # python imaging library

conda install cython # Why?

conda install ipykernel # for kernel selection in jupyter

# conda install matplotlib PyYAML opencv-python pydot pillow cython

Check tensorflow gpu

import tensorflow as tf
tf.test.is_gpu_available(
    cuda_only=False,
    min_cuda_compute_capability=None
)

Resources

Services
Cloud Services
Tools
Git