An FPGA-Based Reconfigurable Accelerator for Convolution-Transformer Hybrid EfficientViT
This work addresses the hardware efficiency problem for deploying efficient Vision Transformers on embedded devices, representing an incremental improvement over existing accelerators.
The paper tackles the challenge of deploying EfficientViT, a state-of-the-art Vision Transformer with a Convolution-Transformer hybrid architecture, on embedded devices by proposing an FPGA-based accelerator that achieves up to 780.2 GOPS in throughput and 105.1 GOPS/W in energy efficiency.
Vision Transformers (ViTs) have achieved significant success in computer vision. However, their intensive computations and massive memory footprint challenge ViTs' deployment on embedded devices, calling for efficient ViTs. Among them, EfficientViT, the state-of-the-art one, features a Convolution-Transformer hybrid architecture, enhancing both accuracy and hardware efficiency. Unfortunately, existing accelerators cannot fully exploit the hardware benefits of EfficientViT due to its unique architecture. In this paper, we propose an FPGA-based accelerator for EfficientViT to advance the hardware efficiency frontier of ViTs. Specifically, we design a reconfigurable architecture to efficiently support various operation types, including lightweight convolutions and attention, boosting hardware utilization. Additionally, we present a time-multiplexed and pipelined dataflow to facilitate both intra- and inter-layer fusions, reducing off-chip data access costs. Experimental results show that our accelerator achieves up to 780.2 GOPS in throughput and 105.1 GOPS/W in energy efficiency at 200MHz on the Xilinx ZCU102 FPGA, which significantly outperforms prior works.