EKI-App: Energy-efficient artificial intelligence in the data center by approximating deep neural networks for field-programmable gate arrays
Overview
The goal of the project is to increase the energy efficiency of AI systems for DNN inference by approximation methods and mapping on high-performance FPGAs. By adapting, further developing and providing a software tool chain based on the open source tool FINN for the automated, optimized and hardware-adapted implementation of DNNs on FPGAs and evaluating the resulting energy savings through precise measurements in real server systems, the project closes the existing gap for the practical use of FPGAs with their energy and/or performance benefits for AI users.
Key Facts
- Project type:
- Forschung
- Project duration:
- 01/2023 - 12/2025
- Contribution to sustainability:
- Responsible consumption and production, Industry, innovation, and infrastructure
More Information
Publications
FINN-HPC: Closing the Gap for Energy-Efficient Neural Network Inference on FPGAs in HPC
L. Jungemann, B. Wintermann, H. Riebler, C. Plessl, in: Proceedings of the 15th International Symposium on Highly Efficient Accelerators and Reconfigurable Technologies, ACM, New York City, n.d.
AuroraFlow, an Easy-to-Use, Low-Latency FPGA Communication Solution Demonstrated on Multi-FPGA Neural Network Inference
Show all publications
G. Pape, B. Wintermann, L. Jungemann, M. Lass, M. Meyer, H. Riebler, C. Plessl, in: Proceedings of the 15th International Symposium on Highly Efficient Accelerators and Reconfigurable Technologies, n.d.