🧠 Debian AI Setup Summary

📌 Overview

This guide covers setting up a complete AI/ML development environment on Debian Trixie with NVIDIA GPU acceleration, using Miniforge for Python environment management.

🧹 System Cleanup

  • Removed Kali Linux repositories and packages to prevent dependency conflicts
  • Cleaned up APT package management with:bashsudo apt update && sudo apt upgrade sudo apt autoremove

🎮 NVIDIA GPU Setup

  • GPU: NVIDIA GeForce GT 1030
  • Driver Version: 550.163.01 (supports CUDA 12.4)
  • CUDA Toolkit: Installed via Debian repos

Commands:

bash

sudo apt install nvidia-driver nvidia-smi
nvidia-smi  # Verify GPU detection

🐍 Python Environment

  • Miniforge3 chosen over Anaconda for lighter, faster Conda experience
  • Alias conflict resolved: Removed alias python='/home/lvydvy/anaconda3/bin/python' from .bashrc
  • PATH prioritized:bashexport PATH=”/home/lvydvy/miniforge3/bin:$PATH”

🔥 PyTorch Installation (GPU)

  • Installed Version: 2.6.0+cu124
  • Method: pip (conda had dependency issues with Python 3.13)

bash

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

Verification:

python

import torch
print(torch.cuda.is_available())  # True
print(torch.cuda.get_device_name(0))  # NVIDIA GeForce GT 1030

🔢 TensorFlow Installation

bash

pip install tensorflow

Verification:

python

import tensorflow as tf
print(tf.config.list_physical_devices('GPU'))
# [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

📦 Other AI Packages

PackageVersionPurpose
scikit-learn1.6.1Classical ML
OpenCV4.13.0Computer Vision
ONNXOptionalModel interoperability

bash

conda install scikit-learn opencv onnx -c conda-forge -y

🧪 Final Test Script

python

import torch
import tensorflow as tf
import sklearn
import cv2

print(f"PyTorch: {torch.__version__}, CUDA: {torch.cuda.is_available()}")
print(f"TensorFlow: {tf.__version__}, GPUs: {tf.config.list_physical_devices('GPU')}")
print(f"scikit-learn: {sklearn.__version__}")
print(f"OpenCV: {cv2.__version__}")

✅ System Status

ComponentStatus
OSDebian Trixie
NVIDIA Driver✅ 550.163.01
CUDA Support✅ 12.4
PyTorch (GPU)✅ 2.6.0+cu124
TensorFlow✅ 2.19.1
scikit-learn✅ 1.6.1
OpenCV✅ 4.13.0

🎯 Key Lessons Learned

  1. Don’t mix Kali repos with Debian — they break system dependencies
  2. Match PyTorch CUDA version to your driver — PyTorch 2.6.0+cu124 works with driver 550
  3. Check Python aliases — Anaconda/Miniconda can override system Python
  4. Use pip for PyTorch when conda has Python 3.13 compatibility issues
  5. Always verify GPU detection with torch.cuda.is_available()

🚀 Next Steps

  • Explore model training with PyTorch/TensorFlow
  • Install Hugging Face Transformers:bashpip install transformers datasets
  • Set up Jupyter Lab:bashpip install jupyterlab

📚 Resources