Release SoE artifacts (crypto models, encrypted datasets) on Hugging Face

[NielsRogge]

Hi @DataArcTech 🤗

Niels here from the open-source team at Hugging Face. I discovered your work on Arxiv and was wondering whether you would like to submit it to hf.co/papers to improve its discoverability. If you are one of the authors, you can submit it at https://huggingface.co/papers/submit.

The paper page lets people discuss about your paper and lets them find artifacts about it (your models, datasets or demo for instance), you can also claim the paper as yours which will show up on your public profile at HF, and add GitHub and project page URLs.

It'd be great to make the checkpoints (the "crypto" models based on Llama3 and Qwen2.5) and the encrypted synthetic datasets you introduced available on the 🤗 hub, to improve their discoverability and visibility for the privacy-preserving ML community. We can add tags so that people find them when filtering https://huggingface.co/models and https://huggingface.co/datasets.

Uploading models

See here for a guide: https://huggingface.co/docs/hub/models-uploading.

In this case, since you mention using LlamaFactory, the models should be easily compatible with the transformers library. Alternatively, one can leverage the hf_hub_download one-liner to download a checkpoint from the hub.

Uploading dataset

Would be awesome to make the synthetic datasets available on 🤗 , so that people can do:

from datasets import load_dataset

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

See here for a guide: https://huggingface.co/docs/datasets/loading.

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

Let me know if you're interested or need any help regarding this!

Cheers,

Niels
ML Engineer @ HF 🤗