Daniel van Strien on the Hugging Face hub and synthetic creation of a DPO dataset for Haiku

We are really excited about the next community meetup with Daniel van Strien from Hugging Face. For the people who don't know him, Daniel van Strien, is Hugging Face's machine learning librarian and he has also been one of the leaders behind the push for our Hugging Face prompt-collective :)
​
We discussed:
The work Hugging Face is doing to make the Hub better for data.
​- His beautiful haiku-dpo project which aims to make LLMs better at writing Haiku. The long-term goal is to see if DPO can be used to help make LLMs match an individual's aesthetic preferences. The project was created using distilabel and Argilla.
​- Why data is better together and we will share some thoughts and preliminary results from our prompt collective efforts.

Signup for coming meetups here: https://lu.ma/d720wy9f