A Stochastic Rounding-Enabled Low-Precision Floating-Point MAC for DNN Training

TL;DR

This paper introduces a low-precision floating-point MAC unit with stochastic rounding and reduced accumulator precision, achieving hardware efficiency while maintaining DNN training accuracy across various vision tasks.

Contribution

It proposes a novel FP8-based MAC with FP12 accumulations and stochastic rounding, reducing hardware complexity and power consumption without sacrificing model accuracy.

Findings

Significant reduction in MAC area and power consumption.

Maintains near baseline accuracy across multiple vision tasks.

Optimized stochastic rounding implementation improves hardware efficiency.

Abstract

Training Deep Neural Networks (DNNs) can be computationally demanding, particularly when dealing with large models. Recent work has aimed to mitigate this computational challenge by introducing 8-bit floating-point (FP8) formats for multiplication. However, accumulations are still done in either half (16-bit) or single (32-bit) precision arithmetic. In this paper, we investigate lowering accumulator word length while maintaining the same model accuracy. We present a multiply-accumulate (MAC) unit with FP8 multiplier inputs and FP12 accumulations, which leverages an optimized stochastic rounding (SR) implementation to mitigate swamping errors that commonly arise during low precision accumulations. We investigate the hardware implications and accuracy impact associated with varying the number of random bits used for rounding operations. We additionally attempt to reduce MAC area and power by proposing a new scheme to support SR in floating-point MAC and by removing support for subnormal values. Our optimized eager SR unit significantly reduces delay and area when compared to a classic lazy SR design. Moreover, when compared to MACs utilizing single-or half-precision adders, our design showcases notable savings in all metrics. Furthermore, our approach consistently maintains near baseline accuracy across a diverse range of computer vision tasks, making it a promising alternative for low-precision DNN training.