Improve discoverability of your project with Hugging Face Hub

[qubvel]

Hey!

I'm Pavel, from the open-source team at Hugging Face. Congrats on your work and acceptance to ECCV2024!

It would be great to make the checkpoints available on HF, along with dedicated model cards, which also enhance the visibility of your work. People can find them when searching https://huggingface.co/models?pipeline_tag=mask-generation&sort=trending.

This also comes with download stats, so that you see how many times people download your models.
e.g. clip-vit-large-patch14 has been downloaded 42M times last month: https://huggingface.co/openai/clip-vit-large-patch14.

1. Upload checkpoints in separate model repos.
   Ideally, each checkpoint is pushed to a separate model repository. Here's a guide for uploading models to the 🤗 hub: https://huggingface.co/docs/hub/models-uploading. For this use case, I assume that the PyTorchModelHubMixin class is best suited, it creates a config.json along with safetensors for each model.

2. Make the model Transformers compatible.
   In case you want your models to be usable through the Transformers library with trust_remote_code=True, I highly recommend following this guide: https://huggingface.co/docs/transformers/custom_models. It allows people to use your backbones using the AutoModel and AutoModelForMaskGeneration APIs. We recently did the same with the MambaVision author, as shown here: https://huggingface.co/collections/nvidia/mambavision-66943871a6b36c9e78b327d3.

Let me know if you need any help regarding this!

Thanks,

Pavel
ML Engineer @ HF 🤗

[bit-scientist]

I believe @byminji can help.