vllm-project/vllm

[Feature]: Implement Concurrent Partial Prefills In V1 Engine

Open

#14,003 opened on Feb 28, 2025

View on GitHub
 (16 comments) (0 reactions) (0 assignees)Python (80,034 stars) (16,816 forks)batch import
feature requesthelp wantedunstale

Description

🚀 The feature, motivation and pitch

In V0, we support concurrent partial prefills to avoid TTFT latency with long requests. Implement it in V1

cc @WoosukKwon

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.

Contributor guide

Tech stack
python
Domain
backendperformancemachine learning
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
5
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Understanding of vLLM's V1 engine architectureKnowledge of concurrent request handlingFamiliarity with partial prefills concept
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
10
Research direction
Start by studying the V0 implementation of concurrent partial prefills in the vllm repository, focusing on the scheduler and model runner. Review issue comments for design constraints. Identify V1 engine differences and propose an integration that avoids TTFT latency overhead. Consider existing discussions or related PRs for guidance.