Checkpoints for QuEST (quantization aware training)

[NielsRogge]

Hi @BlackSamorez,

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.

It'd be great to see checkpoints for QuEST on the 🤗 hub, to improve their discoverability/visibility.
We can add tags so that people find them when filtering https://huggingface.co/models.

Are there any plans to release the pre-trained checkpoints? Let me know if you're interested/need any help regarding this!

Cheers,

Niels
ML Engineer @ HF 🤗

[BlackSamorez]

Hi!

I believe we submitted the paper to the hf.co/papers at the earliest opportunity already. Thanks!

We also provide a number of checkpoints for our method, mostly referenced in the verification notebooks in this repo (like this one). They are there mostly to demonstrates that there is no disconnect between the training process and the proposed inference kernels. However, I don't think they'll be of much practical use to the community otherwise (what we trained is mostly scaling laws verification), so we aren't looking into deeper HF integrations (for now).

[NielsRogge]

Thanks for providing context!

I noticed https://huggingface.co/daslab-testing/hadamard-testing-800m doesn't have a model card - would be great to add a minimal description and link it to https://huggingface.co/papers/2502.05003 (this can be achieved by simply adding the paper URL in the README)

[BlackSamorez]

I arranged proper READMEs and linked the paper. Thanks!

[NielsRogge]

Very nice! I opened a PR to add metadata: https://huggingface.co/ISTA-DASLab/QuEST-800M-INT4/discussions/1

Btw FYI you can also add the paper page to a collection