Code

FPGA Transformer Inference Survey - Source Code

Arjan Blankestijn

Uraz Odyurt, Amirreza Yousefzadeh [supervisors]

Abstract

This code artefact supports a systematic literature review of FPGA acceleration for Transformer inference. The Python pipeline combines paper metadata from IEEE Xplore and Scopus CSV exports with additional arXiv entries. It removes duplicate DOIs, applies publication-year and abstract filters, excludes existing reviews and training-focused studies, and ranks the remaining papers using keywords. It exports the filtered results as CSV and generates a publication-year histogram. Additional analysis scripts produce throughput and energy-efficiency plots.

The code accompanies the publication "Recent Developments in Transformer Inference Deployment on FPGA Platforms: A Survey".

Refer to the provided README file for further details.

Cite as »

Metadata

Type:

Code

Year:

2026

Repository:

Zenodo

DOI:

10.5281/zenodo.20185394

Links

Related artefacts

Utilised by:

Transformer Inference on FPGAs: Survey

Licence

This code artefact is licenced under CC BY 4.0.