Minimal Python Setup for LLM Development on MacBook

Ashwin Kashyap

A minimal Python setup for LLM development on a MacBook is crucial for a clean and efficient workflow. Here’s a concise guide to current best practices.

To Conda or Not to Conda?

For a minimal setup focused on LLM transformer development, Conda is likely overkill. While Conda is a powerful environment and package manager, it can be heavy and introduce unnecessary complexity. A more lightweight and modern approach is to use a combination of pyenv to manage Python versions and venv to manage project-specific dependencies. This combination provides the flexibility of Conda without the added bulk.

pyenv lets you easily switch between multiple Python versions on a single machine, which is invaluable when different projects have different version requirements.

venv is a built-in Python module that creates isolated environments, ensuring that the packages for one project don’t interfere with others.

The Minimalist’s Setup Guide 🚀

Here’s a streamlined process for setting up your environment:

  1. Install Homebrew: If you don’t already have it, install the quintessential macOS package manager.
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
  1. Install pyenv: Use Homebrew to install pyenv.
brew install pyenv
  1. Configure pyenv: Add the following lines to your shell’s configuration file (e.g., .zshrc for Zsh or .bash_profile for Bash) to ensure pyenv is loaded automatically.
echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.zshrc
echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.zshrc
echo 'eval "$(pyenv init --path)"' >> ~/.zshrc

Remember to restart your shell or run source ~/.zshrc for the changes to take effect.

  1. Install a Python Version: With pyenv installed, you can now install your desired Python version. For LLM development, a recent version like 3.11 or later is recommended.
pyenv install 3.11.4
pyenv global 3.11.4

Managing Your Virtual Environments

Once pyenv has set your global Python version, you can create a virtual environment for your project using venv.

  1. Navigate to your project directory:
cd your-llm-project
  1. Create the virtual environment:
python -m venv venv

This creates a venv directory within your project folder.

  1. Activate the environment:
source venv/bin/activate

Your shell prompt will now indicate that you are in the virtual environment.

  1. Install your dependencies: You can now install your required packages, such as PyTorch, Transformers, and Datasets.
pip install torch transformers datasets
  1. Create a requirements.txt file: To ensure reproducibility, save your project’s dependencies to a requirements.txt file.
pip freeze > requirements.txt

Should You Check venv into Git?

No, you should not check your venv directory into your Git repository. This is a critical best practice for several reasons:

  • Bloat: The venv folder can be quite large, containing numerous packages and their dependencies, which will unnecessarily inflate your repository size.
  • Platform Dependency: Virtual environments are often platform-specific. A venv created on a MacBook may not work correctly on a Linux or Windows machine.
  • Redundancy: The requirements.txt file is the canonical source for your project’s dependencies. Anyone can recreate the virtual environment using this file with a simple pip install -r requirements.txt.

In a larger project that includes Go and Hugo, it’s even more important to keep your repository clean and focused on the source code. Your .gitignore file should always include an entry for the virtual environment directory:

# Python
venv/

By following these guidelines, you can maintain a lean, efficient, and reproducible Python environment for your LLM development endeavors, even within a multi-language project structure.