Release SARMAE model checkpoints on Hugging Face

[NielsRogge]

Hi @wenquandan777 🤗

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/2512.16635.

It's fantastic to see that you've already shared the SAR-1M dataset on the Hub! It's a significant contribution to the remote sensing community. The paper page on Hugging Face allows people to discuss your work and discover related artifacts; you can also claim the paper as yours which will show up on your public profile, and add your GitHub and project page URLs.

I noticed in your GitHub repository that the SARMAE model checkpoints (ViT-B and ViT-L) and pre-training code are listed as "Coming Soon." Would you be interested in hosting these models on https://huggingface.co/models once they are ready for release?

Hosting on Hugging Face will give your work more visibility and enable better discoverability. We can add metadata tags to the model cards so that people find the models easier, and they can be directly linked to your paper page and the SAR-1M dataset.

If you're interested, I'm leaving a guide here. For custom PyTorch models, you can leverage the PyTorchModelHubMixin class which adds from_pretrained and push_to_hub methods, making it very easy for others to use your models.

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

Kind regards,

Niels
ML Engineer @ HF 🤗

[wenquandan777]

Hi @NielsRogge 🤗

Thank you very much for your kind message and for highlighting our work on Hugging Face — we truly appreciate the support from the open-source team.

We’re glad to hear that the SAR-1M dataset has been helpful to the community. Regarding the SARMAE models, the code is currently still being organized, but the pre-trained checkpoints will be released soon and we expect them to be available within about a week.

We would definitely be interested in hosting the models on Hugging Face once everything is ready. The improved visibility and integration you mentioned sound very valuable, and we appreciate you sharing the resources and guidance.

Thanks again for reaching out, and we’ll keep you updated once the models are released. Please let us know if there are any recommended best practices we should follow when preparing the release.

Kind regards,
Danxu Liu