vllm-project/vllm

[Feature]: Support DCP with FP8 KV Cache

Open

#32,010 建立於 2026年1月9日

在 GitHub 查看
 (3 留言) (0 反應) (0 負責人)Python (80,034 star) (16,816 fork)batch import
feature requesthelp wanted

描述

🚀 The feature, motivation and pitch

DCP is great Fp8 kv cache is great We should support both at the same time:

(EngineCore_DP0 pid=2734260) ERROR 01-08 23:29:29 [core.py:900] RuntimeError: Worker failed with error 'DCP not support fp8 kvcache now.', please check the stack trace above for the root cause

Alternatives

No response

Additional context

No response

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.

貢獻者指南

技術棧
python
領域
backendmachine learning
議題類型
feature
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
4
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
over 1 week
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
fresh
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
mostly clear
前置要求
familiarity with vllm's KV cacheunderstanding of DCP
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
15
研究方向
The error occurs in core.py:900. Investigate the DCP implementation (likely in vllm/distributed) and the FP8 KV cache code path (likely in vllm/kv cache). The goal is to enable DCP to handle FP8 KV cache, possibly by adding a check or conversion. Review existing tests for DCP and FP8 KV cache. Consider looking at PRs or issues related to DCP or FP8 for context.