Improve discoverability on HF, make download stats work

[NielsRogge]

Hi @wanghao9610,

Congrats on this work! Niels here from the open-source team at Hugging Face. It's great to see you're releasing models on HF :) however some remarks:

Download stats

I also see that currently download stats aren't working for your models. This is due to the model repository containing various checkpoints, which do not contain a config.json file. See here for more info: https://huggingface.co/docs/hub/models-download-stats. Usually we recommend to push checkpoints each to a single repository.

Hence I was wondering whether you would be interested in leveraging either hf_hub_download or the PyTorchModelHubMixin to push model checkpoints to separate model repositories, each containing a config.json and safetensors file (which is considered safer than pickle). This would ensure downloads are working.

Discoverability

Next to that, we could also improve the discoverability of your models by adding a "zero-shot-object-detection" tag, along with a model card. See here for more info: https://huggingface.co/docs/huggingface_hub/en/guides/model-cards.

This ensures people can find your models at https://huggingface.co/models?pipeline_tag=zero-shot-object-detection&sort=trending.

Let me know if you need any help!

Cheers,

Niels
ML Engineer @ HF 🤗

[wanghao9610]

Thanks for your advice @NielsRogge , I will continue to add our code with hf_hub_download in the next few weeks, I may ask for your help. Thanks again in advanced.

[robinjoe93]

@wanghao9610 预计什么时候可以在hf上跑你们的模型？感觉你们的模型跟grounding dino1.5Pro效果差不多诶。

[wanghao9610]

@robinjoe93 hf transformers的适配有点繁琐，还有一些其他的工作需要完成。我会尽快的，预计8月中下旬可以完成