[Bug]: Error when running multimodal large models with --enable-prefix-caching

[jiyanxin]

Your current environment

The output of `python collect_env.py`

When I was using vllm to launch the Qwen2-VL model service, I configured the parameter --enable-prefix-caching. An error occurred when the service requested the second image. It seems that during the current use, this parameter is not compatible with multimodal large models. Do we have plans to fix this bug for compatibility in the future?

🐛 Describe the bug

When I was using vllm to launch the Qwen2-VL model service, I configured the parameter --enable-prefix-caching. An error occurred when the service requested the second image. It seems that during the current use, this parameter is not compatible with multimodal large models. Do we have plans to fix this bug for compatibility in the future?

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[DarkLight1337]

Are you using the latest version of the PR branch for Qwen2-VL? This should be fixed by a recent PR #8028

Edit: Could you show how you're inputting the images?

[github-actions]

This issue has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this issue should remain open. Thank you!