Pre-work¶
This section will give you InstrucLab background as well as guide you through the neccessary prerequisites and installations
Background¶
What is an LLM?¶
Large language models (LLMs) are a category of foundation models trained on immense amounts of data making them capable of understanding and generating natural language and other types of content to perform a wide range of tasks.
Granite¶
Granite is a family of IBM XXXXXXXXXXXXXXXXXXXXXXXXXX
What is InstructLab?¶
InstructLab is an open-source AI project that aims to make it easier for anyone to contribute to Large Language Models (LLMs). It's a community-driven initiative that allows users to add "skills" or "knowledge" to existing LLMs in a way that's both accessible and effective.
Why InstructLab¶
There are many projects rapidly embracing and extending permissively licensed AI models, but they are faced with three main challenges:
- Contribution to LLMs is not possible directly. They show up as forks, which forces consumers to choose a “best-fit” model that isn’t easily extensible. Also, the forks are expensive for model creators to maintain.
- The ability to contribute ideas is limited by a lack of AI/ML expertise. One has to learn how to fork, train, and refine models to see their idea move forward. This is a high barrier to entry.
- There is no direct community governance or best practice around review, curation, and distribution of forked models.
InstructLab is here to solve these problems.
The project enables community contributors to add additional "skills" or "knowledge" to a particular model.
InstructLab's model-agnostic technology gives model upstreams with sufficient infrastructure resources the ability to create regular builds of their open source licensed models not by rebuilding and retraining the entire model but by composing new skills into it.
Take a look at "lab-enhanced" models on the InstructLab Hugging Face page.
How does InstructLab work?¶
InstructLab leverages synthetic data generation to create datasets used to tune models. This allows users to tune models for specific tasks and teach models new knowledge without needing large amounts of user data. The user provided data is added to a taxonomy tree in the form of YAML files and markdown documents as needed for knowledge. The InstructLab training process then generates sythetic data from the files in the taxonomy true and samlples from the data using the Lab tuning process. More details and specifics in this process can be found here, in the original paper detailing this process created by IBM researchers.
Visual Studio Code¶
Mac installation steps¶
Open up a terminal, and install homebrew, if you haven't already.
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
After the installation is complete, install vscode via brew.
brew install --cask visual-studio-code
Windows installation steps¶
Note
If you are running this workshop on Windows, as of this writing it's an unsupported platform. You may run into some issues, and as they say "There be dragons."
Install Code via the website here.
Python¶
Python is a whole programming language. There are multpile ways to install it, and here is the offical website. Please take a moment and if you can't run the following command, reach out to a teaching assissant or instructor to help you get resolved.
python --version
Python 3.12.4
Please confirm that your python --version is at least 3.11+ for the best experience.
Ollama¶
Mac installation steps¶
Download via the Ollama website¶
Download Ollama via the website.
Unzip the folder, and move the Ollama app to your applications folder.
Terminal Installation¶
Open up a terminal, and install homebrew.
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
After the installation is complete, install ollama via brew.
brew install ollama
Windows installation steps¶
Install ollama via the website here.
With this you should have the knowledge and applications you need, so let's start the workshop!