The Elixir community is making significant progress in machine learning, and Hugging Face is a key enabler. A recent demonstration shows how to build a Whisper-based chat app using Livebook and deploy it to Hugging Face Spaces in under 15 minutes.
In this chat app, users communicate solely through audio messages, which are automatically transcribed to text using the Whisper model. This showcases Livebook's integration with Hugging Face models, support for multiplayer machine learning apps, and concurrent model serving—with the added ability to distribute model serving across a cluster.
Livebook is an open-source tool for interactive Elixir notebooks, part of the Elixir NX ecosystem for numerical computing, data science, and machine learning.
Hugging Face and Elixir
The Elixir community leverages Hugging Face across its machine learning stack. The Bumblebee library brings pre-trained neural network models from Hugging Face to Elixir, inspired by Hugging Face Transformers. The tokenizers library is an Elixir binding for Hugging Face Tokenizers. Additionally, Livebook can run inside Hugging Face Spaces as a Docker template, allowing users to run notebooks for free or deploy apps.
Your Turn
This integration empowers more people to use machine learning and share their work. To get started, install Livebook on Hugging Face Spaces and follow the video tutorial to build and deploy your first Livebook ML app.