AI Development Environment Setup

This repository documents a reproducible Ubuntu-based AI development environment with NVIDIA GPU acceleration.

What this repo proves

• NVIDIA driver + CUDA runtime are correctly installed
• PyTorch (CUDA 12.8, Blackwell / sm_120) works on RTX 5090
• GPU workloads run both natively and inside Docker containers
• Environment can be reproduced via conda or Docker

Structure

ai-dev-setup/
├── environment.yml # Conda environment definition
├── docker/
│ └── Dockerfile # GPU-enabled Docker image
└── validation/
└── gpu_test.py # CUDA validation script

Reproduce conda environment

conda env create -f environment.yml
conda activate ai
python validation/gpu_test.py

Test GPU inside Docker

docker build -t ai-dev-gpu-test -f docker/Dockerfile .
docker run --rm --gpus all ai-dev-gpu-test

Notes

• Host system: Ubuntu 24.04
• GPU: NVIDIA GeForce RTX 5090
• CUDA runtime: 12.8