---
cited_by:
  - '[[thoughts/inference/kv-layout#3-what-each-layout-costs-the-kernels|reading-map §3]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/worker/gpu/attn_utils.py
merged: '2026-07-11'
number: 47314
source:
  - https://github.com/vllm-project/vllm/blob/9c18e90f6c94b90ecdaa99b2230389ba40e0fc69/vllm/v1/worker/gpu/attn_utils.py
state: merged
tags:
  - vllm
  - vllm-pr
  - hetero-tp-kv-layout
title: '#47314 [BugFix] Fix packed HND KV cache reshape for FlashAttention'
topic: hetero-tp-kv-layout
url: https://github.com/vllm-project/vllm/pull/47314
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/47314
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/47314.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Fixes [#47054](https://github.com/aarnphm/aarnphm.github.io/issues/47054). · full dump: [[thoughts/inference/vllm-prs/47314|#47314]]

