Creating a good repeatable machine learning environment is often glossed over when working on a new python project. I would like to share how I manage and maintain my Python environment.
Mambaforge
Mambaforge is a minimal installer for conda that also includes mamba and is configured with conda-forge as the default and only channel. Anaconda or Conda is known for being slow when trying to resolve conflicts between dependencies and installing new packages. Mamba is a replacement for Conda. It’s written in C++ for faster execution. Even on a M1 mac it can be a great choice.
Anaconda has also recently changed its TOS and it’s not too clear how this affects us. Sticking to mamba would be a better choice according to me.
Installation
You can install the latest version of Mambaforge by running the commands below or using this script.
wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-Linux-x86_64.sh
bash Mambaforge-Linux-x86_64.sh -b
~/mambaforge/bin/mamba initExit your current shell session and start another one. Type the command conda --help to check that conda environment variables are correctly set.
usage: conda [-h] [-V] command ...
conda is a tool for managing and deploying applications, environments and packages.
Options:
…You will observe that conda config --show channels conda-forge is set as the default channel.
channels:
- conda-forgeCreating a New Environment
Simply use the commands below create a new environment called my_env with python 3.9 interpreter.
conda create -n my_env python=3.9
conda activate my_envUse pip install or conda install to install the package of your choice to your environment. ## Useful Commands Other than the commands mentioned above you may find these commands below useful.
# All conda env created
conda env list
# All packages in current conda env
conda list
# System information
conda info