AI Setup On Linux

Follow our guide to install AI models on Linux efficiently

This comprehensive tutorial will walk you through everything needed to install and run the Stable Diffusion image-generation model on Ubuntu/Debian. You’ll prepare your system, set up CUDA, create isolated Python environments, install all dependencies, download model weights, launch the web UI, and troubleshoot common issues—all with annotated screenshots and properly formatted code blocks.

1. System Prerequisites & GPU Setup

• Ubuntu 20.04+ or Debian 11+
• NVIDIA GPU with CUDA capability (compute > 6.0)
• 8+ GB RAM (16 GB recommended)
• Internet access to download packages and model weights

2. Install NVIDIA Driver, CUDA & cuDNN

Add NVIDIA repository and install driver and CUDA toolkit:

                    
                        sudo apt update
                        sudo apt install -y gnupg curl
                        curl -s -L https://nvidia.github.io/nvidia-apt-key.gpg | sudo apt-key add -
                        distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
                        curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list \
                          | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
                        sudo apt update
                        sudo apt install -y nvidia-driver-515 cuda-toolkit-11-7 libcudnn8 libcudnn8-dev
                    
                

After installation, reboot and verify:

                    
                        nvidia-smi
                    
                

3. Update & Install Base Packages

                    
                        sudo apt update && sudo apt upgrade -y
                        sudo apt install -y git wget unzip python3.10 python3.10-venv python3.10-dev build-essential cmake
                    
                

4. Create Project Directory & Virtual Env

                    
                        mkdir -p ~/stable-diffusion && cd ~/stable-diffusion
                        python3.10 -m venv .venv
                        source .venv/bin/activate
                        pip install --upgrade pip setuptools wheel
                    
                

5. Clone Web UI & Install Python Dependencies

                    
                        git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git .
                        pip install -r requirements.txt
                    
                

This installs PyTorch (with CUDA support), diffusers, transformers, and other libraries.

6. Download & Place Model Weights

                    
                        mkdir -p models/Stable-diffusion
                        cd models/Stable-diffusion
                        wget -O sd-v1-4.ckpt https://huggingface.co/CompVis/stable-diffusion-v-1-4-original/resolve/main/sd-v1-4.ckpt
                        cd ~/stable-diffusion
                      
                

7. Launch Web UI with GPU

                    
                        source .venv/bin/activate
                        python webui.py --precision full --no-half --enable-insecure-extension-access
                    
                

Access the UI at http://localhost:7860. The extra flag allows loading third-party extensions.

8. Optional: Install Extensions & Models

                    
                        # In Web UI > Extensions tab, install via Git URL
                        # Or manually:
                        cd extensions
                        git clone https://github.com/someuser/sd-webui-extension.git
                    
                

To add new models or extensions, simply clone into the extensions folder.

9. Troubleshooting

• If PyTorch doesn’t detect GPU: ensure CUDA paths in ~/.bashrc:

                              
                                  export PATH=/usr/local/cuda-11.7/bin:$PATH
                                  export LD_LIBRARY_PATH=/usr/local/cuda-11.7/lib64:$LD_LIBRARY_PATH
                              
                          

• If webui.py fails on import: run pip install --upgrade -r requirements.txt

10. Performance Tuning

To speed up output and reduce memory consumption:

                    
                        python webui.py --precision full --no-half --opt-sdp-attention
                        export TORCHTRT_HIGHER_PRECISION_DISABLED=1
                    
                

11. Automatic Updates

Set up a cron to automatically update once a week:

                    
                        crontab -e
                        0 3 * * 0 cd ~/stable-diffusion && git pull && source .venv/bin/activate && pip install -r requirements.txt
                    
                

12. Security & Permissions

To restrict access to the UI, you can run via SSH tunnel or with basic authentication:

                    
                        python webui.py --listen localhost
                        {
                          "basic_auth_username": "user",
                          "basic_auth_password": "pass"
                        }
                    
                

13. Cleanup & Disk Management

Remove old checkpoints and compress logs:

                    
                        find ~/stable-diffusion/models -type f -mtime +30 -delete
                        gzip ~/stable-diffusion/logs/*.log
                    
                

14. Verify Python & PyTorch Installation

                    
                        python --version
                        python -c "import torch; print(torch.cuda.is_available())"
                    
                

15. Run a GPU Stress Test

                    
                        python -m torch.cuda.benchmarks
                    
                

16. Optimize Disk I/O

                    
                        sudo apt install -y nvme-cli
                        sudo nvme format /dev/nvme0n1 --filesystem host
                        echo 'vm.swappiness=10' | sudo tee -a /etc/sysctl.conf
                    
                

17. Automated Startup Script

                    
                        [Unit]
                        Description=Stable Diffusion WebUI
                        After=network.target
                        [Service]
                        User=$USER
                        WorkingDirectory=/home/$USER/stable-diffusion
                        ExecStart=/home/$USER/stable-diffusion/.venv/bin/python webui.py --precision full --no-half
                        Restart=on-failure
                        [Install]
                        WantedBy=multi-user.target
                    
                

Conclusion

Now you have a complete, debugged and automated AI-setup on Linux: from basic dependencies to optimization, security and automatic startup.