Add Hiera model

[p1atdev]

Model description

Hiera is a hierarchical vision transformer that is fast, powerful, and, above all, simple. It outperforms the state-of-the-art across a wide array of image and video tasks while being much faster.

Open source status

• [X] The model implementation is available
• [X] The model weights are available

Provide useful links for the implementation

GitHub Repo: https://github.com/facebookresearch/hiera/ (but licensed under CC BY-NC 4.0) arXiv: https://arxiv.org/abs/2306.00989

[Namangarg110]

Can I work on this ?

[amyeroberts]

@Namangarg110 Certainly! Feel free to open a PR when you're ready and ping us for review 🤗. To avoid issues from becoming too stale, we will prioritise the first open PR when reviewing over the first comment on issues.

[Namangarg110]

Thanks @amyeroberts. This is my first open-source issue. Would it be possible for you to please share any helpful resources or similar PR to understand the code structure?

[amyeroberts]

Sure!

• Docs page: https://huggingface.co/docs/transformers/en/add_new_model
• Example model PR: https://github.com/huggingface/transformers/pull/26668

Adding models is quite a big project. If you want to tackle something smaller for your first issue to get used to the workflow of contributing to transformers, resolving a good first issue is a great place to start.

[Namangarg110]

Thank you for the resources, @amyeroberts.

I've begun the task and have completed 50% of the work.

I recognize that contributing a new model can be exceptionally challenging, but I am eager to give it a try. :)