Fix: Gemma3TextConfig rope scaling assignments

[RyanMullins]

What does this PR do?

Related to #41922, this PR corrects the assignment of the rope_scaling dictionary present on some Gemma 3 pre-trained models on HF Hub when normalizing to the new rope_parameters value.

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.

@zucchini-nlp PTAL since you have been handling the RoPE changes.

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: gemma3

[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.

[Rocketknight1]

I'm guessing the context for this is that rope_scaling clobbering rope_parameters (the old behaviour) gave the right config for older models but breaks on upcoming models that might have specific configs for e.g. sliding_attention?

[RyanMullins]

I'm guessing the context for this is that rope_scaling clobbering rope_parameters (the old behaviour) gave the right config for older models but breaks on upcoming models that might have specific configs for e.g. sliding_attention?

Yes, we've observed that rope_scaling was being applied to both full_attention and sliding_attention for the Gemma 3 checkpoints on HF Hub and in our attempts to re-convert them from the original Orbax checkpoint. The correct behavior is for rope_scaling to only be applied to full_attention if it exists in the config, and the sliding_attention config should always be default RoPE @ 10k.

[Rocketknight1]

Got it, thanks for the fix!