[Bugfix] gemma[2,3] interleaved attention when sliding window is disabled

[heheda12345]

I get the following errors when running VLLM_ATTENTION_BACKEND=FLASHINFER vllm serve google/gemma-2-2b-it

ERROR 04-25 02:35:23 [core.py:395]   File “***/vllm/model_executor/models/gemma2.py", line 149, in __init__
ERROR 04-25 02:35:23 [core.py:395]     config.interleaved_sliding_window is not None)
ERROR 04-25 02:35:23 [core.py:395]     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 04-25 02:35:23 [core.py:395]   File “***/lib/python3.11/site-packages/transformers/configuration_utils.py", line 210, in __getattribute__
ERROR 04-25 02:35:23 [core.py:395]     return super().__getattribute__(key)
ERROR 04-25 02:35:23 [core.py:395]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 04-25 02:35:23 [core.py:395] AttributeError: 'Gemma2Config' object has no attribute 'interleaved_sliding_window'

This PR fix it

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[DarkLight1337]

Reference Slack discussion: https://vllm-dev.slack.com/archives/C07QCGVDNUF/p1744754229620969

[kylehh]

Confirmed this fix works for both v0 and v1 engine for following test
VLLM_ATTENTION_BACKEND=FLASHINFER VLLM_ALLOW_LONG_MAX_MODEL_LEN=1 python3 examples/offline_inference/vision_language.py --model gemma3

[mgoin]

Thank you!