Release manifold-microscope artifacts on Hugging Face

[NielsRogge]

Hi @koulakis 🤗

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, add Github and project page URLs.

I saw your manifold-microscope repository and the note that it is still under construction for reproduction. It'd be great to make the checkpoints (like the $\beta$-VAE weights) and the extended datasets (the repurposed dSprites and COIL-20 variants) available on the 🤗 hub once they are ready, to improve their discoverability and visibility for the research 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.

For custom PyTorch models like your VAEs, you can leverage the PyTorchModelHubMixin class which adds from_pretrained and push_to_hub to any custom nn.Module. 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 benchmark 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/need any help regarding this!

Cheers,

Niels
ML Engineer @ HF 🤗

[koulakis]

Hi @NielsRogge 😃

Thanks for reaching out and for the helpful suggestions! Adding the paper to Hugging Face was already on my to-do list, so your message came at a good time!

I have now created a Hugging Face account and submitted/claimed the paper. Once the paper authorship is verified, I will link it and the code repository on Daily Papers.

Uploading datasets and models may take a bit longer, since I would first like to further clean up and update the codebase so that the released artifacts are easy to use and properly documented.

Regarding the datasets and model checkpoints, I would currently only plan to upload the artifacts related to dSprites, with proper attribution to the original dataset and a clear description of the modifications. Unfortunately, COIL-20 does not seem to have an explicit redistribution license, so I would avoid uploading COIL-20-derived datasets or checkpoints.

Thanks again for the invitation and for offering help. I will let you know once the paper page and repository links are in place.

Cheers,
Marios