Convolutional neural networks on dataflow engines

None, None; None, None; None, None; None, None; None, None

Convolutional neural networks on dataflow engines

Conference Paper (2017)

Author(s)

Nils Voss (Imperial College London, Maxeler Technologies)

Marco Bacis (Politecnico di Milano, Maxeler Technologies)

Oskar Mencer (Maxeler Technologies)

Georgi Gaydadjiev (Imperial College London, Maxeler Technologies)

Wayne Luk (Imperial College London)

Affiliation

External organisation

Deep Learning CNN Inference FPGA DFE DSE

DOI related publication

https://doi.org/10.1109/ICCD.2017.77 Final published version

To reference this document use

https://resolver.tudelft.nl/uuid:10f15013-5c51-4720-9849-f0e578e6afb5

More Info

expand_more

Publication Year

2017

Language

English

Affiliation

External organisation

Article number

8119250

Pages (from-to)

435-438

ISBN (electronic)

9781538622544

Event

35th IEEE International Conference on Computer Design, ICCD 2017 (2017-11-05 - 2017-11-08), Boston, United States

Downloads counter

180

Abstract

In this paper we discuss a high performance implementation for Convolutional Neural Networks (CNNs) inference on the latest generation of Dataflow Engines (DFEs). We discuss the architectural choices made during the design phase taking into account the DFE chip properties. We then perform design space exploration, considering the memory bandwidth and resources utilisation constraints derived from the used DFE and the chosen architecture. Finally, we discuss the high performance implementation and compare the obtained performance against other implementations, showing that our proposed design reaches 2,450 GOPS when running VGG16 as a test case.