---
cited_by:
  - '[[thoughts/inference/kv-transfer#5-reading-order|reading-map §5]]'
date: '2026-11-11'
kind: pr
local: []
merged: '2026-07-28'
number: 49612
source:
  - https://github.com/vllm-project/vllm/blob/601fa9a74e8cd2ae0a4fa127d68b91bbf363a24e/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py
  - https://github.com/vllm-project/vllm/blob/601fa9a74e8cd2ae0a4fa127d68b91bbf363a24e/vllm/distributed/kv_transfer/kv_connector/v1/nixl/pull_worker.py
  - https://github.com/vllm-project/vllm/blob/601fa9a74e8cd2ae0a4fa127d68b91bbf363a24e/vllm/distributed/kv_transfer/kv_connector/v1/nixl/push_worker.py
state: merged
tags:
  - vllm
  - vllm-pr
  - hetero-tp-kv-layout
title: '#49612 [KV Connector] Support NIXL heterogeneous P/D block sizes for hybrid models'
topic: hetero-tp-kv-layout
url: https://github.com/vllm-project/vllm/pull/49612
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/49612
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/49612.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Mamba state blocks are indivisible, so their descriptors always use local page geometry and their desc ids are never ratio-expanded; attention descriptors remain expanded to remote-block granularity. · full dump: [[thoughts/inference/vllm-prs/49612|#49612]]

