WinoTrain: Winograd-Aware Training for Accurate Full 8-bit Convolution Acceleration

Pierpaolo Mori, Shambhavi Balamuthu Sampath, Lukas Frickenstein, Manoj Rohit Vemparala, Nael Fasfous, Alexander Frickenstein, Walter Stechele, Claudio Passerone

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Efficient inference is critical in realizing a low-power, real-time implementation of convolutional neural networks (CNNs) on compute and memory-constrained embedded platforms. Using quantization techniques and fast convolutional algorithms like Winograd, CNN inference can achieve benefits in latency and in energy consumption. Performing Winograd convolution involves (1) transforming the weights and activations to the Winograd domain, (2) performing element-wise multiplication on the transformed tensors, and (3) transforming the results back to the conventional spatial domain. Combining Winograd with quantization of all its steps results in severe accuracy degradation due to numerical instability. In this paper we propose a simple quantization-aware training technique, which quantizes all three steps of the Winograd convolution, while using a minimal number of scaling factors. Additionally, we propose an FPGA accelerator employing tiling and unrolling methods to highlight the performance benefits of using the full 8-bit quantized Winograd algorithm. We achieve 2× reduction in inference time compared to standard convolution on ResNet-18 for the ImageNet dataset, while improving the Top-1 accuracy by 55.7 p.p. compared to a standard post-training quantized Winograd variant of the network.

Original languageEnglish
Title of host publication2023 60th ACM/IEEE Design Automation Conference, DAC 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798350323481
DOIs
StatePublished - 2023
Event60th ACM/IEEE Design Automation Conference, DAC 2023 - San Francisco, United States
Duration: 9 Jul 202313 Jul 2023

Publication series

NameProceedings - Design Automation Conference
Volume2023-July
ISSN (Print)0738-100X

Conference

Conference60th ACM/IEEE Design Automation Conference, DAC 2023
Country/TerritoryUnited States
CitySan Francisco
Period9/07/2313/07/23

Fingerprint

Dive into the research topics of 'WinoTrain: Winograd-Aware Training for Accurate Full 8-bit Convolution Acceleration'. Together they form a unique fingerprint.

Cite this