Release FlowBoost models and datasets on Hugging Face

[NielsRogge]

Hi @berczig 🤗

I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work through Hugging Face's daily papers as yours got featured: https://huggingface.co/papers/2601.18005.

The paper page lets people discuss your paper and find artifacts about it (like your flow-matching models or discovered geometric configurations). You can also claim the paper as yours, which will show up on your public profile at HF, and add your GitHub and project page URLs.

I saw on your GitHub repository that the code for FlowBoost is "coming soon". Would you be interested in hosting the pre-trained models and the discovered extremal datasets on the Hugging Face hub? Hosting on Hugging Face will give your work more visibility and enable better discoverability through metadata tags.

Uploading models

We have a guide here. If you are using PyTorch, you can even leverage the PyTorchModelHubMixin class which adds from_pretrained and push_to_hub methods to your model, making it very easy for others to use.

Uploading datasets

It would also be great to make the discovered configurations (like the circle packings or sphere packings) available on the 🤗 hub so that people can load them with:

from datasets import load_dataset
dataset = load_dataset("your-hf-org-or-username/your-dataset")

Besides that, there's the dataset viewer which allows people to explore the geometric data directly in the browser.

Let me know if you're interested or need any guidance!

Kind regards,

Niels
ML Engineer @ HF 🤗

[Baran-phys]

Hi @NielsRogge, this sounds great indeed. Thank you for contacting us. We have just updated our repo with v1 of the code. In the coming days, we will be on it to extend our results to be viewed in the Hugging Face hub as well.