[Misc] Modify CacheConfig import

[jeejeelee]

Purpose

When using the latest main branch of transformers, encounter the following import error

ERROR 08-23 02:59:11 [registry.py:430]   File "/root/Code/vllm_dev/vllm/vllm/attention/layers/encoder_only_attention.py", line 8, in <module>
ERROR 08-23 02:59:11 [registry.py:430]     from transformers import CacheConfig
ERROR 08-23 02:59:11 [registry.py:430] ImportError: cannot import name 'CacheConfig' from 'transformers' (/root/Code/vllm_dev/transformers/src/transformers/__init__.py)

In addition, based on the parameters of the Attention, CacheConfig should come from vllm.config.

Test Plan

Test Result

(Optional) Documentation Update

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[gemini-code-assist]

Code Review

This pull request correctly resolves an ImportError for CacheConfig in vllm/attention/layers/encoder_only_attention.py. The change updates the import path from transformers to vllm.config, which is the correct source for this class as used by the Attention superclass. The fix is accurate and necessary to resolve the described runtime error.

[heheda12345]

Sorry for that... But quite strange that type checker failed to catch this problem.