This repository contains slides and notebooks for a workshop on Transformers and Large Language Models (LLMs), covering both user and developer perspectives.
Learning objectives:
- Understand the strengths and limitations of LLMs, including practical tips and an overview of current regulations
- Get to know Blablador, Helmholtz’s in-house alternative to other LLM providers
- Discover transformers — the architecture and mechanisms behind powerful LLMs
- Learn how transformers can be applied in your research
- Learn how to fine-tune an LLM for a specific task
The course will be fully online:
Zoom link
Meeting ID: 833 8752 6921
Passcode: 615095
We strongly recommend launching the provided notebook with Google Colab in order to execute the code in a self-contained and verified working environment. In this case, no setup is required.
In case you wish not to use Google Colab, you can install and run the notebooks locally by following the guide below.
git clone https://github.com/HelmholtzAI-Consultants-Munich/llm-transformers-course.git
Or enter the repository and press Code -> Download ZIP
Enter the directory you have downloaded:
cd llm-transformers-course
Create the virtual environment.
You can either use a virtual environment or a conda environment:
# virtual environment
python3 -m venv .venv
source .venv/bin/activate
# or conda
conda create -n llm_course python=3.13
conda activate llm_course
Then, follow jupyter notebook installation requirements here.
jupyter notebook Transformer_finetuning_tutorial.ipynb
A browser window will open, navigate to the .ipynb file. Alternatively, you can open it with VSCode.
Warning: Newer versions of Python sometimes don't work with jupyter notebooks. If you encounter 404 errors when running the notebook, switch to Python 3.11.
Notebook solutions are now available. Please see the solutions notebook for solutions to the exercises.
- Karol Szustakowski, Helmholtz Munich
- Donatella Cea, Helmholtz Munich
- Marcela Astrid, Helmholtz Munich
- Corrado Pancotti, Helmholtz Munich
- Leo Kaindl, Helmholtz Munich
It is possible to either create a local environment and install all the necessary packages (using the requirements.txt file), or to run the notebooks directly in the browser by clicking the “Open in Colab” button (recommended). This second option does not require any installation, but you will need access to a Google account.
Comments and input are very welcome! If you have a suggestion or you think something should be changed, please open an issue or submit a pull request.
This project is licensed under the MIT License.