Update modular_modernbert.py to support input_embeddings

[KoichiYasuoka]

What does this PR do?

Fixes #35344

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[tomaarsen]

I believe the norm is that only the base model, i.e. ModernBertModel here, has get_input_embeddings and set_input_embeddings. See for example modeling_bert.py and modeling_roberta.py.

Beyond that, the ModernBertForSequenceClassification class doesn't have an attribute called embeddings.

However, the Python 3.9 patch is very good, and I'll be making a separate PR for it.

[tomaarsen]

I opened #35458 with @KoichiYasuoka as co-author to implement the Python 3.9 support from this PR.
Unless we'd like to break the norm and add get_input_embeddings to the other model classes, we can probably leave this PR be, and only merge #35458.

• Tom Aarsen

[KoichiYasuoka]

Thank you @tomaarsen for separating issues. Well... I added get_input_embeddings in #29878 to handle input_embeds for "single character which was tokenized into multiple tokens". So I really need input_embeds for ModernBERT but this time I'm planing to emulate with set_input_embeddings to [UNUSED] token.

[tomaarsen]

Oh I see, I didn't realise there were non-base model classes that use get_input_embeddings. That changes things!

[KoichiYasuoka]

Thank you @tomaarsen and here I link to Ek and Bernardy (2020).