14 October 2024
Convergence Center @ Purdue University
US/Eastern timezone

Attention at Silicon Speed: Towards Efficient Transformers on FPGAs

14 Oct 2024, 13:55
10m
Innovation Room (Convergence Center @ Purdue University)

Innovation Room

Convergence Center @ Purdue University

101 Foundry Dr, West Lafayette, IN 47906

Speaker

Rian Flynn (Purdue University (US))

Description

The High-Luminosity Large Hadron Collider (HL-LHC), anticipated to begin operations in 2029, will generate data at an astounding rate on the order of 100 terabits per second. To efficiently process and filter these data, the Compact Muon Solenoid (CMS) experiment
relies on the extremely low-latency Level-1 trigger, which uses Field-Programmable Gate Arrays (FPGAs). My project focuses on further optimizing this process by adapting efficient transformer models for implementation on FPGAs using the hls4ml package. This
talk will highlight my contributions, ongoing work to optimize FPGA implementations of transformer models, and my experiences as an A3D3 Postbaccalaureate Fellow.

Author

Rian Flynn (Purdue University (US))

Presentation materials