paper-with-me

홈 › Papers

LLMForge: Multi-Backend Hardware-Aware Neural Architecture Search with Infinite-Head Attention for Edge Language Models

2026-05-17 · Xinting Jiang, Junyi Luo, Ruichen Qi, Kauna Lei, Ben Laurie, Gregory Kielian, Mehdi Saligane arxiv

Sub-billion-parameter Transformer language models are increasingly deployed on edge devices, where the privacy, latency, and operating-cost advantages of on-device inference are constrained by tight memory-bandwidth, energy, and thermal budgets that make architectural choice and accelerator-specific cost central to efficient inference. We present LLMForge, a hardware-aware neural architecture search (NAS) framework whose three composable contributions together make edge-LM architecture search hardware-conditioned, since different substrates impose different hardware cost bottlenecks. Infinite-Head Attention (IHA) decouples the number of query heads, KV groups, and per-head query/key and value dimensions, expanding the feasible per-layer attention configuration space by approximately 400x over grouped-query attention within our search-space ranges. Forge-Former, an encoder-based surrogate for ranking architectural candidates, outperforms MLP and random-forest baselines. Forge-DSE, an NSGA-II-based design-space-exploration engine, pairs Forge-Former with a multi-backend hardware cost model spanning GPUs, systolic accelerators, and ring-dataflow edge accelerators. Across four different hardware substrates, the searches converge to visibly different architectures whose shapes track each substrate's cost bottleneck. On the multi-chip ring substrate, our co-search returns three 300M-scale deployment-aware variants on the Pareto front. Each is re-trained on FineWeb-Edu-10BT under matched recipe against SmolLM2-360M and Qwen-0.5B architecture baselines. The accurate variant has the lowest validation loss 2.798 and competitive benchmark performance with fewer parameters, the energy-optimized variant lowers energy per token by 40%, and the latency-optimized variant lowers TTFT and TPOT by 43%.

📄 PDF Abstract BibTeX arXiv:2605.17653

Code (0)

등록된 구현이 없습니다.

Tasks

Neural Architecture Search

Similar Papers 제목 키워드 기반

NAS-Driven Hardware Accelerator Exploration for Edge AI and Quantization Effects on the Pareto Space

2026-08-13 · Eleftherios Mylonas, Angelos Kouprizas, Michael Birbas, Alexios Birbas arxiv

Edge AI deployment demands neural architectures that are simultaneously accurate, computationally efficient, and hardware-deployable - a challenge addressed by hardware-aware Neural Architecture Search (NAS). While recen…

Neural Architecture Search

LLM-Guided Neural Architecture Search for Robust Co-Design of Physical Neural Networks

2026-06-09 · Tyler King, Timothee Leleu arxiv

Deploying neural networks on unconventional hardware demands architectures that co-optimize task accuracy and platform-specific constraints such as energy cost, physical non-idealities, and numerical precision. Existing …

Neural Architecture Search

TinyCNN: A Tiny Modular CNN Accelerator for Embedded FPGA

2019-11-15 · Ali Jahanshahi

In recent years, Convolutional Neural Network (CNN) based methods have achieved great success in a large number of applications and have been among the most powerful and widely used techniques in computer vision. However…

General Classificationimage-classificationImage Classification

Automated Backend-Aware Post-Training Quantization

2021-03-27 · Ziheng Jiang, Animesh Jain, Andrew Liu, Josh Fromm 외

Quantization is a key technique to reduce the resource requirement and improve the performance of neural network deployment. However, different hardware backends such as x86 CPU, NVIDIA GPU, ARM CPU, and accelerators may…

CPUDiversityGPUQuantization

Dooly: Configuration-Agnostic, Redundancy-Aware Profiling for LLM Inference Simulation

2026-05-08 · Joon Ha Kim, Geon-Woo Kim, Anoop Rachakonda, Daehyeok Kim arxiv

Selecting the optimal LLM inference configuration requires evaluation across hardware, serving engines, attention backends, and model architectures, since no single choice performs best across all workloads. Profile-base…