CI Gemma 4 support

[qgallouedec]

This PR comes after #5452 and only adds tests for Gemma4 training

---

Note

Low Risk
Low risk: changes are limited to test coverage and the internal tiny-model generation script, with version-gated skips to avoid running on unsupported transformers releases.

Overview
Extends the tiny-model generator to include Gemma 4 VLM (google/gemma-4-E2B-it) and adds a Gemma4-specific config path that mutates text_config/vision_config on the loaded config to bypass rope validation.

Updates VLM training-related parametrized tests (DPO/GRPO/RLOO/SFT and utils) to include trl-internal-testing/tiny-Gemma4ForConditionalGeneration, guarded by skipif(transformers<5.5.0).

^{Reviewed by Cursor Bugbot for commit 9a41509. Bugbot is set up for automated code reviews on this repo. Configure here.}

[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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 55cefc4bc5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 6824819. Configure here.}

[albertvillanova]

Thanks.

I included "CI" in the PR title to make clear it only adds CI tests, but feel free to better rename it.