---
cited_by:
  - '[[thoughts/inference/kv-transfer#5-reading-order|reading-map §5]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/worker/gpu/attn_utils.py
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/worker/gpu_model_runner.py
  - file:///Users/aarnphm/workspace/vllm/vllm/distributed/kv_transfer/kv_connector/utils.py
  - file:///Users/aarnphm/workspace/vllm/vllm/distributed/kv_transfer/kv_connector/v1/offloading/worker.py
  - file:///Users/aarnphm/workspace/vllm/vllm/model_executor/layers/mamba/abstract.py
  - file:///Users/aarnphm/workspace/vllm/vllm/model_executor/layers/attention_layer_base.py
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/worker/utils.py
merged: '2026-07-21'
number: 44456
source:
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/v1/worker/gpu/attn_utils.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/v1/worker/gpu_model_runner.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/distributed/kv_transfer/kv_connector/utils.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/distributed/kv_transfer/kv_connector/v1/offloading/worker.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/model_executor/layers/mamba/abstract.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/model_executor/layers/attention_layer_base.py
  - https://github.com/vllm-project/vllm/blob/6700813f86552a2e0787989dd9f468fddcb88bac/vllm/v1/worker/utils.py
state: merged
tags:
  - vllm
  - vllm-pr
  - hetero-tp-kv-layout
title: '#44456 [3/N][KV-Cache Layout Refactor] Standardize Mamba cache; drop `get_transfer_cache_regions`'
topic: hetero-tp-kv-layout
url: https://github.com/vllm-project/vllm/pull/44456
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/44456
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/44456.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Move Mamba (and conv/ssm) reshaping into a new `bind_kv_cache` hook; drop the now-unnecessary `TransferTopology.get_transfer_cache_regions` from the connector. This is an intermediate step toward [#42374](https://github.com/aarnphm/aarnphm.github.io/issues/42374) that does not yet introduce the… · full dump: [[thoughts/inference/vllm-prs/44456|#44456]]

