StruM: Structured Mixed Precision for Efficient Deep Learning Hardware Codesign

TL;DR

StruM introduces a structured mixed-precision inference method co-designed with hardware, reducing computational demands and power consumption in deep learning accelerators without retraining.

Contribution

The paper presents a novel mixed-precision quantization approach that leverages weight variance, avoiding retraining, and is co-designed with a hardware accelerator for efficiency.

Findings

Up to 50% reduction in weight precision with negligible accuracy loss.

31-34% reduction in processing element power consumption.

23-26% area reduction at the PE level.

Abstract

In this paper, we propose StruM, a novel structured mixed-precision-based deep learning inference method, co-designed with its associated hardware accelerator (DPU), to address the escalating computational and memory demands of deep learning workloads in data centers and edge applications. Diverging from traditional approaches, our method avoids time-consuming re-training/fine-tuning and specialized hardware access. By leveraging the variance in weight magnitudes within layers, we quantize values within blocks to two different levels, achieving up to a 50% reduction in precision for 8-bit integer weights to 4-bit values across various Convolutional Neural Networks (CNNs) with negligible loss in inference accuracy. To demonstrate efficiency gains by utilizing mixed precision, we implement StruM on top of our in-house FlexNN DNN accelerator [1] that supports low and mixed-precision execution. Experimental results depict that the proposed StruM-based hardware architecture achieves a 31-34% reduction in processing element (PE) power consumption and a 10% reduction in area at the accelerator level. In addition, the statically configured StruM results in 23-26% area reduction at the PE level and 2-3% area savings at the DPU level.