### What this PR does / why we need it?
Delete accuracy testing of some models:
- Qwen2-VL-7B-Instruct
- Qwen2.5-VL-7B-Instruct
- gemma-2-9b-it
- DeepSeek-V2-Lite
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: MrZ20 <2609716663@qq.com>