Abstract:Domain-specific hardware is becoming a promising topic in the backdrop of improvement slow down for general-purpose processors due to the foreseeable end of Moore's Law. Machine learning, especially deep neural networks (DNNs), has become the most dazzling domain witnessing successful applications in a wide spectrum of artificial intelligence (AI) tasks. The incomparable accuracy of DNNs is achieved by paying the cost of hungry memory consumption and high computational complexity, which greatly impedes their deployment in embedded systems. Therefore, the DNN compression concept was naturally proposed and widely used for memory saving and compute acceleration. In the past few years, a tremendous number of compression techniques have sprung up to pursue a satisfactory tradeoff between processing efficiency and application accuracy. Recently, this wave has spread to the design of neural network accelerators for gaining extremely high performance. However, the amount of related works is incredibly huge and the reported approaches are quite divergent. This research chaos motivates us to provide a comprehensive survey on the recent advances toward the goal of efficient compression and execution of DNNs without significantly compromising accuracy, involving both the high-level algorithms and their applications in hardware design. In this article, we review the mainstream compression approaches such as compact model, tensor decomposition, data quantization, and network sparsification. We explain their compression principles, evaluation metrics, sensitivity analysis, and joint-way use. Then, we answer the question of how to leverage these methods in the design of neural network accelerators and present the state-of-the-art hardware architectures. In the end, we discuss several existing issues such as fair comparison, testing workloads, automatic compression, influence on security, and framework/hardware-level support, and give promising topics in this field and the possible challenges as well. This article attempts to enable readers to quickly build up a big picture of neural network compression and acceleration, clearly evaluate various methods, and confidently get started in the right way.

Hardware-friendly compression and hardware acceleration for transformer: A survey

Accelerating Framework of Transformer by Hardware Design and Model Compression Co-Optimization

Accommodating Transformer Onto FPGA

Hardware Accelerator for Multi-Head Attention and Position-Wise Feed-Forward in the Transformer

A Length Adaptive Algorithm-Hardware Co-design of Transformer on FPGA Through Sparse Attention and Dynamic Pipelining

Transformer-based models and hardware acceleration analysis in autonomous driving: A survey

FTRANS: Energy-Efficient Acceleration of Transformers using FPGA

TransFRU: Efficient Deployment of Transformers on FPGA with Full Resource Utilization

A Survey on Efficient Training of Transformers

An Efficient Hardware Accelerator for Sparse Transformer Neural Networks

Model Compression and Hardware Acceleration for Neural Networks: A Comprehensive Survey

A Heterogeneous Chiplet Architecture for Accelerating End-to-End Transformer Models

Full Stack Optimization of Transformer Inference: a Survey

Recent Advances in Efficient Computation of Deep Convolutional Neural Networks

Ayaka: A Versatile Transformer Accelerator with Low-Rank Estimation and Heterogeneous Dataflow

ViA: A Novel Vision-Transformer Accelerator Based on FPGA

A Runtime-Adaptive Transformer Neural Network Accelerator on FPGAs

H3D-Transformer: A Heterogeneous 3D (H3D) Computing Platform for Transformer Model Acceleration on Edge Devices

TATAA: Programmable Mixed-Precision Transformer Acceleration with a Transformable Arithmetic Architecture

FPGA-Based Vit Inference Accelerator Optimization

A Survey on Transformer Compression