The practicals are a core, long-standing component of the Indaba. The 2024 practicals will enable participants to take a deep dive into core topics in artificial intelligence while implementing applied examples of the latest machine learning ideas. The aim is to equip the community with skills and knowledge so that foundational topics are accessible to all!

The practicals are developed by a team of advanced machine learning practitioners, this year led by Kira Dusterwald (UCL), James Allingham (Google DeepMind), and Raoul de Charette (Inria). Attendees will explore each topic using Jupyter notebooks and Python. To encourage their use as an ongoing resource, participants can also access the practicals—including more advanced and bonus material—after the Indaba.

Format and topics

This year’s practicals are hosted over three two-hour sessions: two cover foundational content, and one day is applications-based. The applications day is dedicated to applying fundamental algorithms to relevant, practical problems. The five application-based practicals (and the RL practical) were sought through a community call. Over 40 excellent applications were received during this process.

Below is the full list of practicals per session. Participants are encouraged to read up on each topic (more details to follow here and on GitHub) and choose which practical to attend in each session.

Day 1 (foundations 1):

  • Introduction to ML [using JAX]
  • Introduction to Probabilistic Thinking and Programming
  • Graph Neural Networks

Day 2 (foundations 2):

  • Responsible AI
  • LLM foundations
  • Diffusion Models: Building your own Stable Diffusion
  • From zero to ‘2048’: Building RL environment with Jax

Day 3 (applications):

  • AI for Biology
  • Fine-tuning and resource-efficient LLMs for NLP
  • From Centralised to Decentralised Training: An Intro to Federated Learning
  • Recommender Systems

Set-up for attendees

As in the last editions of the Indaba, practicals will be run on Python using Jupyter notebooks. We use JaX as the core machine learning module, as is now familiar to the Indaba community. To increase accessibility and ease of use, this year practicals will all be able to be downloaded and run offline on local machine CPUs, with components requiring GPUs highlighted for later exploration via Google Colab.

We are anticipating difficulty with Internet connectivity in Senegal, and hence for practicals to run smoothly it is important to pre-install the software required to run the Indaba notebooks locally. Attendees should please follow these instructions to download Jupyterlab, Python and the relevant modules/files for each practical in advance of arrival in Dakar! 

Attendees will be able to choose and download the notebooks for the practicals that they want to attend in the week prior to the Indaba.

Contact Information

Questions? Send us an email at kira@ / james@ / raoul@ deeplearningindaba.com.

Looking forward to coding with you in Senegal! 🙂