Setting Up an AI Python Environment on Debian with Conda: The Ultimate Guide

Why Use Conda for AI Development?

Conda is a powerful environment and package manager that simplifies managing Python environments, dependencies, and even packages that require compilation. For AI development, it is ideal because:

It supports GPU-based libraries like PyTorch and TensorFlow.
It avoids “dependency hell”.
It lets you isolate environments for different projects.

Prerequisites

A Debian 12 or newer installation (minimal install or full desktop).
Sudo privileges or root access.
Basic familiarity with the terminal.

Step 1: Update and Upgrade Your System

Open a terminal and update your package lists:

sudo apt update && sudo apt upgrade -y

Install essential build tools:

sudo apt install -y build-essential curl wget git

Step 2: Install Miniconda (Recommended)

Miniconda is a lighter version of Anaconda, which is perfect if you want to keep things lean and install only what you need.

Download the Miniconda Installer:

cd /tmp
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Verify the integrity (optional but recommended):

sha256sum Miniconda3-latest-Linux-x86_64.sh

Compare the hash with the official site:
Miniconda Hashes

Run the Installer:

bash Miniconda3-latest-Linux-x86_64.sh

Follow the prompts:

Accept the license.
Choose the install location (default is usually fine).
Allow it to initialize Conda in .bashrc.

Restart your shell:

source ~/.bashrc

Verify installation:

conda --version

Step 3: Create a New AI Environment

Create a Conda environment named ai with Python 3.10:

conda create -n ai python=3.10

Activate the environment:

conda activate ai

Step 4: Install Core AI & ML Packages

Install key packages used in AI research and development:

conda install -y numpy pandas matplotlib seaborn scikit-learn jupyterlab ipykernel

Install deep learning frameworks:

PyTorch (with CPU support):

conda install pytorch torchvision torchaudio cpuonly -c pytorch

TensorFlow (CPU version):

pip install tensorflow

Optional visualization and NLP tools:

conda install -y spacy nltk plotly

Install HuggingFace Transformers and Datasets:

pip install transformers datasets

Step 5: Set Up Jupyter Lab (IDE for Experiments)

Launch Jupyter Lab:

jupyter lab

You can also make the environment available to Jupyter globally:

python -m ipykernel install --user --name=ai --display-name "Python (AI)"

Step 6: Save Your Environment for Sharing/Reinstalling

To export:

conda env export > ai-environment.yml

To recreate later:

conda env create -f ai-environment.yml

Step 7: Cleanup & Tips

List environments:

conda env list

Remove an environment:

conda remove --name ai --all

Keep Conda updated:

conda update -n base -c defaults conda

Optional: VS Code Integration

Install VS Code and the Python extension:

sudo apt install code

Set the interpreter in VS Code to your ai environment via Ctrl+Shift+P → Python: Select Interpreter → Choose “Python (AI)”.

Summary

You now have a clean, powerful, and isolated Python environment on Debian, tailored for AI/ML development using Conda. Whether you’re training neural networks or experimenting with LLMs, you’re ready to go.