Bash Scripts for System Setup

This directory contains scripts for automating system configuration, particularly useful for WSL/Ubuntu environments.

Recommended Execution Order

1. no_sudo.sh - Configure passwordless sudo (optional, reduces password prompts)
2. setup_dotfiles.sh - Link configuration files (dotfiles) from a central source.
3. install_essentials.sh - Base system packages and tools
4. cuda_install.sh - CUDA and ML environment setup

Note: Running no_sudo.sh first minimizes password prompts during installation. If security is a concern, skip this step and be prepared to enter your password multiple times.

Windows Setup for CUDA in WSL

Before running cuda_install.sh, ensure your Windows host meets these requirements:

1. System Requirements:

   • Windows 10/11 (Build 20150 or later)
   • WSL 2 enabled (run wsl --set-default-version 2 in PowerShell)
   • NVIDIA GPU with driver version 470.14 or newer

2. Windows-side Installation:

   • Install NVIDIA CUDA Toolkit for Windows
   • Reboot after installation
   • Verify with PowerShell:

     nvidia-smi

3. WSL Configuration:

   • Ensure WSL 2 is default:

     wsl --set-default-version 2

   • Install Ubuntu from Microsoft Store

4. Verification:

   • In PowerShell:

     nvidia-smi

   • In WSL:

     lspci | grep -i nvidia

5. Troubleshooting:

   • If nvidia-smi fails in WSL:
     1. Update Windows NVIDIA drivers
     2. Reinstall CUDA Toolkit
     3. Restart WSL (wsl --shutdown)
   • Check WSL version:

     wsl -l -v

Note: The Windows CUDA toolkit version should match or be newer than the version installed in WSL.

Script Documentation

1. no_sudo.sh

Purpose: Configures passwordless sudo for the current user.

Features:

• Creates backup of sudoers file
• Adds NOPASSWD rule for current user
• Sets proper file permissions

Usage:

./no_sudo.sh

WARNING:

• This reduces system security
• Only recommended for development environments
• Revert with: sudo rm /etc/sudoers.d/99_nopasswd_*

2. setup_dotfiles.sh

Purpose: Manages and synchronizes user configuration files (dotfiles) from a central location, ensuring a consistent environment across multiple shells or WSL instances.

Features:

• Supports two sources: a local directory on the Windows host or a remote Git repository.
• Automatically creates symbolic links for your configuration files (e.g., .bashrc, .gitconfig).
• Safely backs up any existing local dotfiles before creating new links.

Usage:

# Option 1: Sync from a folder on your Windows machine
./setup_dotfiles.sh --source windows --path "C:\Users\YourUser\dotfiles"

# Option 2: Sync from a Git repository
./setup_dotfiles.sh --source git --repo "https://github.com/your-user/dotfiles.git"

Options:

• --source [windows|git]: Required. Specify the source type.
• --path <path>: The absolute path to your dotfiles folder on Windows (used with --source windows).
• --repo <url>: The URL of your dotfiles Git repository (used with --source git).

3. install_essentials.sh

Purpose: Installs fundamental development tools and configures a basic environment.

Features:

• Installs essential packages (build-essential, git, vim, etc.).
• Configures bash aliases and terminal colors idempotently (won't add duplicates).
• Installs Node.js 18, Python tools, and Docker Community Edition.
• Provides guidance for VS Code integration within WSL.

Usage:

./install_essentials.sh

Notes:

• Requires sudo privileges
• Some changes require restarting the terminal or starting a new shell session.

4. cuda_install.sh

Purpose: Installs the CUDA toolkit and ML frameworks with GPU support in a self-contained Python environment.

Features:

• Auto-detects the recommended CUDA version based on popular ML frameworks.
• Performs full security verification of the CUDA installer (GPG+SHA256).
• Creates a Python virtual environment to avoid system-wide conflicts.
• Installs PyTorch, TensorFlow, and JAX with corresponding CUDA support.
• Includes robust error handling and automatic cleanup of temporary files.

Usage:

# Basic usage with defaults
./cuda_install.sh

# Specify a Python version and custom environment name
./cuda_install.sh --python-version 3.10 --env-name my-ml-env

Options:

• --python-version: Specify Python version (e.g., 3, 3.9, 3.10).
• --env-name: Set a custom name for the virtual environment.
• --env-path: Define a custom path for the virtual environment.

Requirements:

• NVIDIA drivers must be installed on the Windows host.
• An existing CUDA installation will be detected and the script will exit to prevent overwriting.

Security Considerations

1. no_sudo.sh significantly reduces security - use with caution.
2. setup_dotfiles.sh can execute remote code if using an untrusted Git repository.
3. cuda_install.sh and install_essentials.sh download and install packages from official and third-party repositories.
4. All scripts require internet access.

Maintenance

To update all scripts from your Git repository:

git pull origin main

To verify script integrity:

sha256sum *.sh