Use explicit tiny-Qwen2ForCausalLM-2.5 model_id param in CI tests

[albertvillanova]

Use explicit tiny-Qwen2ForCausalLM-2.5 model_id param in CI tests.

This change improves test flexibility and maintainability by allowing tests to easily run with different model IDs. All hardcoded references to the model "trl-internal-testing/tiny-Qwen2ForCausalLM-2.5" have been replaced with the parametrized variable, and the corresponding test function signatures were updated.

No changes were made to the core test logic or assertions; the focus is on improving test parametrization and maintainability.

Follow-up to:

• Use explicit tiny-Qwen2_5_VL model_id parameter in CI tests #4325

[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.

[qgallouedec]

In my opinion, the numerous repetitions of model_id="Qwen/tiny-Qwen2ForCausalLM-2.5" are not a problem. On the contrary, they allow the tests to be self-contained and easy to copy and paste. However, I find that using a fixture reduces consistency somewhat, because, for example, some tests in TestSFTTrainer will use the fixture model_id, but others (e.g., tests specific to GPT-OSS and VLMs) will keep an explicit model_id. Let me know what you think