@inproceedings{lemphers-etal-2026-chunder,
title = "{CHUNDER}: Enabling Million-Token {Q}wen3 Serving on {AMD} {ROC}m",
author = "Lemphers, David and
Twelftree, Maxwell and
He, An-Chi and
Yang, Yue",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.85/",
pages = "1429--1443",
ISBN = "979-8-89176-460-6",
abstract = "Million-token context serving is becoming a critical requirement for industry LLM systems for real deployments. However, practical 1M-context serving for open-source models has remained heavily CUDA-centric, largely because the required long-context runtime features are not fully available on AMD ROCm. With AMD GPUs now reported at around 19{\%} market share, this gap matters. We address it by enabling 1M-token Qwen3 serving on AMD GPUs with ROCm. We present , a runtime stack that enables million-token serving for the widely used open-source Qwen3 family on AMD ROCm. achieves this by intercepting vLLM{'}s attention routing at runtime and redirecting queries, keys, and values to JIT-compiled Triton Dual Chunk Attention (DCA) kernels, without forking the serving engine. We then harden the surrounding runtime to make full 1M-token execution practical: DCA-aware paged metadata, tiled single-request chunked prefill, MoE token chunking that reduces per-rank transient memory from approximately 91.6 GiB to 10.6 GiB, and FP8 GEMMs with hardened hipBLASLt kernel selection. We evaluate across multiple million-token context tasks and show consistent serving performance and exact retrieval correctness. To our knowledge, this is the first reported end-to-end Qwen3 million-token serving path on AMD ROCm integrated with vLLM."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lemphers-etal-2026-chunder">
<titleInfo>
<title>CHUNDER: Enabling Million-Token Qwen3 Serving on AMD ROCm</title>
</titleInfo>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Lemphers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maxwell</namePart>
<namePart type="family">Twelftree</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">An-Chi</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Million-token context serving is becoming a critical requirement for industry LLM systems for real deployments. However, practical 1M-context serving for open-source models has remained heavily CUDA-centric, largely because the required long-context runtime features are not fully available on AMD ROCm. With AMD GPUs now reported at around 19% market share, this gap matters. We address it by enabling 1M-token Qwen3 serving on AMD GPUs with ROCm. We present , a runtime stack that enables million-token serving for the widely used open-source Qwen3 family on AMD ROCm. achieves this by intercepting vLLM’s attention routing at runtime and redirecting queries, keys, and values to JIT-compiled Triton Dual Chunk Attention (DCA) kernels, without forking the serving engine. We then harden the surrounding runtime to make full 1M-token execution practical: DCA-aware paged metadata, tiled single-request chunked prefill, MoE token chunking that reduces per-rank transient memory from approximately 91.6 GiB to 10.6 GiB, and FP8 GEMMs with hardened hipBLASLt kernel selection. We evaluate across multiple million-token context tasks and show consistent serving performance and exact retrieval correctness. To our knowledge, this is the first reported end-to-end Qwen3 million-token serving path on AMD ROCm integrated with vLLM.</abstract>
<identifier type="citekey">lemphers-etal-2026-chunder</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.85/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1429</start>
<end>1443</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CHUNDER: Enabling Million-Token Qwen3 Serving on AMD ROCm
%A Lemphers, David
%A Twelftree, Maxwell
%A He, An-Chi
%A Yang, Yue
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F lemphers-etal-2026-chunder
%X Million-token context serving is becoming a critical requirement for industry LLM systems for real deployments. However, practical 1M-context serving for open-source models has remained heavily CUDA-centric, largely because the required long-context runtime features are not fully available on AMD ROCm. With AMD GPUs now reported at around 19% market share, this gap matters. We address it by enabling 1M-token Qwen3 serving on AMD GPUs with ROCm. We present , a runtime stack that enables million-token serving for the widely used open-source Qwen3 family on AMD ROCm. achieves this by intercepting vLLM’s attention routing at runtime and redirecting queries, keys, and values to JIT-compiled Triton Dual Chunk Attention (DCA) kernels, without forking the serving engine. We then harden the surrounding runtime to make full 1M-token execution practical: DCA-aware paged metadata, tiled single-request chunked prefill, MoE token chunking that reduces per-rank transient memory from approximately 91.6 GiB to 10.6 GiB, and FP8 GEMMs with hardened hipBLASLt kernel selection. We evaluate across multiple million-token context tasks and show consistent serving performance and exact retrieval correctness. To our knowledge, this is the first reported end-to-end Qwen3 million-token serving path on AMD ROCm integrated with vLLM.
%U https://aclanthology.org/2026.emnlp-industry.85/
%P 1429-1443
Markdown (Informal)
[CHUNDER: Enabling Million-Token Qwen3 Serving on AMD ROCm](https://aclanthology.org/2026.emnlp-industry.85/) (Lemphers et al., EMNLP 2026)
ACL
- David Lemphers, Maxwell Twelftree, An-Chi He, and Yue Yang. 2026. CHUNDER: Enabling Million-Token Qwen3 Serving on AMD ROCm. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 1429–1443, Budapest, Hungary. Association for Computational Linguistics.