add self.head_dim for VisionAttention in Qwen2-VL (#33211)
* add self.head_dim for VisionAttention in Qwen2-VL
* add self.head_dim for VisionAttention in Qwen2-VL
* fix ci
* black the test_modeling_qwen2_vl.py
* use ruff to format test_modeling_qwen2_vl.py
* [run-slow] qwen2_vl
* use tying for python3.8
* fix the import format
* use ruff to fix the ci error I001
* [run-slow] qwen2_vl
* remove unused import
* commit for rebase
* use ruff fix ci
* [run-slow] qwen2_vl
---------
Co-authored-by: root <liji>
