vllm icon indicating copy to clipboard operation
vllm copied to clipboard

[Feature]: `kv_transfer_params` not returned for multiple subrequests

Open jianzs opened this issue 4 months ago • 0 comments

🚀 The feature, motivation and pitch

Currently, when handling HTTP requests with multiple subrequests, the response only includes kv_transfer_params for one subrequest, making it impossible to access KV transfer information for other subrequests.

Related PR: #17751 Reference code: https://github.com/vllm-project/vllm/blob/e384f2f10824df7789c6da35256cf957788c0208/vllm/entrypoints/openai/serving_completion.py#L514-L520

Alternatives

No response

Additional context

No response

Before submitting a new issue...

  • [x] 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.

jianzs avatar Jun 20 '25 09:06 jianzs