Release XBridge mapping checkpoints on Hugging Face

[NielsRogge]

Hi @bingo123122121 🤗

Niels here from 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/2603.17512.
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.

It'd be great to make the XBridge checkpoints (the cross-model mapping layers) and the trilingual translation dataset used for alignment available on the 🤗 hub, to improve their discoverability and visibility for the community. We can add metadata 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, you could 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.

We encourage researchers to push each model checkpoint (for different LLM/NMT combinations) to a separate model repository, so that download stats and model cards work best. We can then link these checkpoints directly to the paper page.

Uploading dataset

It would also be awesome to make the trilingual alignment dataset available on 🤗 , so that people can easily replicate the training:

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 🤗

[bingo123122121]

Thanks, Niels! We are training a multilingual instruction-following version XBridge, and will open-source it on huggingface soon. Next week, probably!