# On-Device Qwen2.5: Efficient LLM Inference with Model Compression and Hardware Acceleration

arXiv: https://arxiv.org/abs/2504.17376 · HTML: https://mlsanity.com/paper/2504.17376/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Serving systems and request scheduling, one of 12 directions in Efficient LLM inference and serving.
- Builds on 13 of its 19 references that are in the map.
- First posted on arXiv 24 Apr 2025.

## Placed in

- Efficient LLM inference and serving / [Serving systems and request scheduling](https://mlsanity.com/inference/serving-systems-and-request-scheduling/)

## Builds on (most cited first, 13 of 13)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [PaLM 2 Technical Report](https://mlsanity.com/paper/2305.10403/) (2023)
- [Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learning](https://mlsanity.com/paper/1702.03118/) (2017)
- [TinyLlama: An Open-Source Small Language Model](https://mlsanity.com/paper/2401.02385/) (2024)
- [A Study of Generative Large Language Model for Medical Research and Healthcare](https://mlsanity.com/paper/2305.13523/) (2023)
- [Large Language Models for Robotics: A Survey](https://mlsanity.com/paper/2311.07226/) (2023)
- [ChatMusician: Understanding and Generating Music Intrinsically with LLM](https://mlsanity.com/paper/2402.16153/) (2024)
- [Understanding the Potential of FPGA-Based Spatial Acceleration for Large Language Model Inference](https://mlsanity.com/paper/2312.15159/) (2023)
- [LlamaF: An Efficient Llama2 Architecture Accelerator on Embedded FPGAs](https://mlsanity.com/paper/2409.11424/) (2024)

## Built on by (most cited first, 6 of 6)

- [Standardization of Psychiatric Diagnoses -- Role of Fine-tuned LLM Consortium and OpenAI-gpt-oss Reasoning LLM Enabled Decision Support System](https://mlsanity.com/paper/2510.25588/) (2025)
- [Standardization of Neuromuscular Reflex Analysis -- Role of Fine-Tuned Vision-Language Model Consortium and OpenAI gpt-oss Reasoning LLM Enabled Decision Support System](https://mlsanity.com/paper/2508.12473/) (2025)
- [Military AI Cyber Agents (MAICAs) Constitute a Global Threat to Critical Infrastructure](https://mlsanity.com/paper/2506.12094/) (2025)
- [Efficient Online RFT with Plug-and-Play LLM Judges: Unlocking State-of-the-Art Performance](https://mlsanity.com/paper/2506.05748/) (2025)
- [Train the Trainers -- An Agentic AI Framework for Peer-Based Mental Health Support in Battlefield Environments](https://mlsanity.com/paper/2605.16269/) (2026)
- [Recent Developments in Transformer Inference Deployment on FPGA Platforms: A Survey](https://mlsanity.com/paper/2609.01212/) (2026)

## Abstract (arXiv)

Transformer-based Large Language Models (LLMs) have significantly advanced AI capabilities but pose considerable challenges for deployment on edge devices due to high computational demands, memory bandwidth constraints, and energy consumption. This paper addresses these challenges by presenting an efficient framework for deploying the Qwen2.5-0.5B model on the Xilinx Kria KV260 edge platform, a heterogeneous system integrating an ARM Cortex-A53 CPU with reconfigurable FPGA logic. Leveraging Activation-aware Weight Quantization (AWQ) with FPGA-accelerated execution pipelines, the proposed approach enhances both model compression rate and system throughput. Additionally, we propose a hybrid execution strategy that intelligently offloads compute-intensive operations to the FPGA while utilizing the CPU for lighter tasks, effectively balancing the computational workload and maximizing overall performance. Our framework achieves a model compression rate of 55.08% compared to the original model and produces output at a rate of 5.1 tokens per second, outperforming the baseline performance of 2.8 tokens per second.