↓Skip to main content

Learning AI Poorly: What is Hugging Face?

·2 mins

(originally posted on LinkedIn)

If you’re kind of new and clicking around in the AI world and reading things you might see “Hugging Face” all over the place. And, if you are like me, you might be thinking, “What the hell is Hugging Face?” Heck, I’m pretty sure I linked to a Hugging Face space in that article last week about MusicGen. Maybe we should talk a little about what it is.

Hugging Face* started out in 2016 as a chatbot targeted toward teenagers. When that didn’t pan out, they pivoted to be a platform for machine learning. In May 2022 they raised a Series C round that valued them at $2 Billion. I’d say that panned out.**

Today, Hugging face is an AI community built upon an open source hosting platform for Natural Language Processing (NLP), Computer Vision, and really any other Machine Learning domain. It allows users to build, train, and deploy models on the platform. Hugging Face also reproduces and distributes pre-trained models as open source software so that anyone can use them. They aim to be the “GitHub of machine learning” and their mission “is to democratize good machine learning and maximize its positive impact across industries and society. Not only do we strive to advance Transformer models, but we also work hard on simplifying their adoption.”

They host open source, community created datasets, models, and spaces in what they call a “hub.” “Spaces” makes it easy for people to create and deploy ML powered demos and provides “16GB RAM, 2 CPU cores and 50GB of (not persistent) disk space” free of charge to run that demo. “Models” are pre-trained ML models you can use in your own solutions. “Datasets” are curated and available for you to use to train your own models. Of course there is plenty of documentation available to help you figure it all out.

So, yeah, not a super exciting article this week but if you didn’t know what Hugging Face was before hopefully now you sort of do.

  • Don’t be tricked, the whole point of this article is to see how many times I can write “Hugging Face” (only 10 times… crazy)

** Their history is a little more interesting than that. Their chatbot could interact with users and to do that they used NLP. They decided to open source their NLP models which the AI community loved. The founders realized this and started open sourcing large models like BERT and GPT. That kind of snowballed into what Hugging Face is today.

↑