Automating Versatile Time-Series Analysis with Tiny Transformers on Embedded FPGAs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ling, Tianheng, Qian, Chao, Haßler, Lukas Johannes, Schiele, Gregor
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911162941046784
author Ling, Tianheng
Qian, Chao
Haßler, Lukas Johannes
Schiele, Gregor
author_facet Ling, Tianheng
Qian, Chao
Haßler, Lukas Johannes
Schiele, Gregor
contents Transformer-based models have shown strong performance across diverse time-series tasks, but their deployment on resource-constrained devices remains challenging due to high memory and computational demand. While prior work targeting Microcontroller Units (MCUs) has explored hardware-specific optimizations, such approaches are often task-specific and limited to 8-bit fixed-point precision. Field-Programmable Gate Arrays (FPGAs) offer greater flexibility, enabling fine-grained control over data precision and architecture. However, existing FPGA-based deployments of Transformers for time-series analysis typically focus on high-density platforms with manual configuration. This paper presents a unified and fully automated deployment framework for Tiny Transformers on embedded FPGAs. Our framework supports a compact encoder-only Transformer architecture across three representative time-series tasks (forecasting, classification, and anomaly detection). It combines quantization-aware training (down to 4 bits), hardware-aware hyperparameter search using Optuna, and automatic VHDL generation for seamless deployment. We evaluate our framework on six public datasets across two embedded FPGA platforms. Results show that our framework produces integer-only, task-specific Transformer accelerators achieving as low as 0.033 mJ per inference with millisecond latency on AMD Spartan-7, while also providing insights into deployment feasibility on Lattice iCE40. All source code will be released in the GitHub repository (https://github.com/Edwina1030/TinyTransformer4TS).
format Preprint
id arxiv_https___arxiv_org_abs_2505_17662
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Automating Versatile Time-Series Analysis with Tiny Transformers on Embedded FPGAs
Ling, Tianheng
Qian, Chao
Haßler, Lukas Johannes
Schiele, Gregor
Machine Learning
Transformer-based models have shown strong performance across diverse time-series tasks, but their deployment on resource-constrained devices remains challenging due to high memory and computational demand. While prior work targeting Microcontroller Units (MCUs) has explored hardware-specific optimizations, such approaches are often task-specific and limited to 8-bit fixed-point precision. Field-Programmable Gate Arrays (FPGAs) offer greater flexibility, enabling fine-grained control over data precision and architecture. However, existing FPGA-based deployments of Transformers for time-series analysis typically focus on high-density platforms with manual configuration. This paper presents a unified and fully automated deployment framework for Tiny Transformers on embedded FPGAs. Our framework supports a compact encoder-only Transformer architecture across three representative time-series tasks (forecasting, classification, and anomaly detection). It combines quantization-aware training (down to 4 bits), hardware-aware hyperparameter search using Optuna, and automatic VHDL generation for seamless deployment. We evaluate our framework on six public datasets across two embedded FPGA platforms. Results show that our framework produces integer-only, task-specific Transformer accelerators achieving as low as 0.033 mJ per inference with millisecond latency on AMD Spartan-7, while also providing insights into deployment feasibility on Lattice iCE40. All source code will be released in the GitHub repository (https://github.com/Edwina1030/TinyTransformer4TS).
title Automating Versatile Time-Series Analysis with Tiny Transformers on Embedded FPGAs
topic Machine Learning
url https://arxiv.org/abs/2505.17662