跳到正文
研踪PaperTrail
← 研踪 PaperTrail
正在查看已观测的 v1。查看最新 v2 →

arXiv:2611.23878 · v1

HETEROVLM: Edge Multi-Request Vision-Language Model Serving on a Heterogeneous Versal SoC with AI Engines and Programmable Logic

首次提交 2026-11-18 13:30
版本更新 2026-11-18 13:30
雷达观测 2026-11-19 19:08
arXiv ↗PDF ↗

摘要

Edge devices in vehicles and robots serve concurrent requests with different deadlines to a vision-language model (VLM), each mixing a compute-bound encoder and prefill with a memory-bound decode. GPU serving targets datacenter throughput with per-iteration host scheduling: an urgent request waits for the running iteration, and small-batch decode costs much energy per token. Prior accelerators on Versal AI Engine (AIE) SoCs map one model onto a fixed AIE dataflow graph for one input stream, without decode or request scheduling. HETEROVLM keeps the AIE graph static and makes it schedulable. The AIE array runs the encoder, prefill and attention, and the programmable logic (PL) holds GEMV engines for small decode steps and a scheduler that orders short work units by class and deadline. Concurrent requests share one decode step that reads the weights once. An urgent request preempts at a work-unit boundary, and the paused request resumes from its results in DDR. On an AMD Versal VC1902, text and camera questions get their first tokens in 140 and 303 ms. HETEROVLM decodes a single request 1.3x and 4.0x faster than NVIDIA Tesla T4 and Jetson Orin Nano GPUs, with 61% and 73% of their energy per token, and uses 51% of the energy per token of an NVIDIA GeForce RTX 3090 Ti. The same bitstream serves multiple VLMs by loading their own instruction programs and weights. A demo is available at https://anonymous.4open.science/r/demo heterovlm-EFM2/.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

仅对比实际观测的元数据,不代表 PDF 正文或实验变化。版本可能不连续;带删除线为移除,下划线为新增。

摘要变化

EdgeVision-language devicesmodels (VLMs) play a pivotal role in enabling autonomous vehicles and robots serveto concurrentperceive requestsand withreason differentabout deadlinesthe tophysical aworld. vision-languageEdge modelcomputers (in autonomous systems typically serve concurrent VLM) inference requests with various deadlines, each mixing a compute-bound encoder and prefill stage with a memory-bound decode stage. GPU serving targetsoptimizes datacenterfor high throughput in datacenters, with per-iteration host scheduling:, which can cause an urgent request waitsto wait for thea runninglong iterationtime, and small-batch decode costs much energy per token. Prior accelerators on Versal AI Engine (AIE) SoCs map one model onto a fixed AIE dataflow graph for one input stream, without decode or request scheduling, which limits their capability in serving multiple requests. We present HETEROVLM, which is optimized for multi-request VLM serving on a resource-constrained heterogeneous platform. It keeps the AIE graph static and makes it schedulable. The AIE array runs the encoder, prefill and attention, and the programmable logic (PL) holdsimplements GEMV engines for small decode steps and a scheduler that orders short work units by class and deadline. Concurrent requests share one decode step that reads the weights once. An urgent request preempts at a work-unit boundary, and the paused request resumes from its results in DDR memory. On an AMD Versal VC1902, text and camera questions get their first tokens in 140 and 303 ms. HETEROVLM decodes a single request 1.3x and 4.0x faster than NVIDIA Tesla T4 and Jetson Orin Nano Super GPUs, withconsuming only 61% and 73% of their energy per token, and usesor 51% of the energy per token of an NVIDIA GeForce RTX 3090 Ti. TheA samesingle bitstream servesruns multiple VLMs by loading, theirswitching ownbetween instructionthem programsat andrun weightstime. A demo is available at https://anonymous.4open.science/r/demo heterovlm-EFM2/.

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v2 · 2026-11-19 11:48 · 最新观测版本

HETEROVLM: Edge Multi-Request Vision-Language Model Serving on a Heterogeneous Versal SoC with AI Engines and Programmable Logic

Vision-language models (VLMs) play a pivotal role in enabling autonomous vehicles and robots to perceive and reason about the physical world. Edge computers in autonomous systems typically serve concurrent VLM inference requests with various deadlines, each mixing a compute-bound encoder and prefill stage with a memory-bound decode stage. GPU serving optimizes for high throughput in datacenters, with per-iteration host scheduling, which can cause an urgent request to wait for a long time, and small-batch decode costs much energy per token. Prior accelerators on Versal AI Engine (AIE) SoCs map one model onto a fixed AIE dataflow graph for one input stream, without decode or request scheduling, which limits their capability in serving multiple requests. We present HETEROVLM, which is optimized for multi-request VLM serving on a resource-constrained heterogeneous platform. It keeps the AIE graph static and makes it schedulable. The AIE array runs the encoder, prefill and attention, and the programmable logic (PL) implements GEMV engines for small decode steps and a scheduler that orders short work units by class and deadline. Concurrent requests share one decode step that reads the weights once. An urgent request preempts at a work-unit boundary, and the paused request resumes from its results in DDR memory. On an AMD Versal VC1902, text and camera questions get their first tokens in 140 and 303 ms. HETEROVLM decodes a single request 1.3x and 4.0x faster than NVIDIA Tesla T4 and Jetson Orin Nano Super GPUs, consuming only 61% and 73% of their energy per token, or 51% of the energy per token of an NVIDIA GeForce RTX 3090 Ti. A single bitstream runs multiple VLMs, switching between them at run time. A demo is available at https://anonymous.4open.science/r/demo heterovlm-EFM2/.

查看此版本解读 →查看此版本原文 ↗
v1 · 2026-11-18 13:30

HETEROVLM: Edge Multi-Request Vision-Language Model Serving on a Heterogeneous Versal SoC with AI Engines and Programmable Logic

Edge devices in vehicles and robots serve concurrent requests with different deadlines to a vision-language model (VLM), each mixing a compute-bound encoder and prefill with a memory-bound decode. GPU serving targets datacenter throughput with per-iteration host scheduling: an urgent request waits for the running iteration, and small-batch decode costs much energy per token. Prior accelerators on Versal AI Engine (AIE) SoCs map one model onto a fixed AIE dataflow graph for one input stream, without decode or request scheduling. HETEROVLM keeps the AIE graph static and makes it schedulable. The AIE array runs the encoder, prefill and attention, and the programmable logic (PL) holds GEMV engines for small decode steps and a scheduler that orders short work units by class and deadline. Concurrent requests share one decode step that reads the weights once. An urgent request preempts at a work-unit boundary, and the paused request resumes from its results in DDR. On an AMD Versal VC1902, text and camera questions get their first tokens in 140 and 303 ms. HETEROVLM decodes a single request 1.3x and 4.0x faster than NVIDIA Tesla T4 and Jetson Orin Nano GPUs, with 61% and 73% of their energy per token, and uses 51% of the energy per token of an NVIDIA GeForce RTX 3090 Ti. The same bitstream serves multiple VLMs by loading their own instruction programs and weights. A demo is available at https://anonymous.4open.science/r/demo heterovlm-EFM2/.

查看此版本解读 →查看此版本原文 ↗