Understanding the potential of fpga-based spatial acceleration for large language model inference

H Chen, J Zhang, Y Du, S ** vision transformer on edge
M Huang, J Luo, C Ding, Z Wei… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
Transformer-like network has shown remarkable high performance in both natural language
processing and computer vision. However, the huge computational demands in non-linear …

Msd: Mixing signed digit representations for hardware-efficient dnn acceleration on fpga with heterogeneous resources

J Wu, J Zhou, Y Gao, Y Ding, N Wong… - 2023 IEEE 31st …, 2023 - ieeexplore.ieee.org
By quantizing weights with different precision for different parts of a network, mixed-precision
quantization promises to reduce the hardware cost and improve the speed of deep neural …

Uint-packing: Multiply your dnn accelerator performance via unsigned integer dsp packing

J Zhang, M Zhang, X Cao, G Li - 2023 60th ACM/IEEE Design …, 2023 - ieeexplore.ieee.org
DSP blocks are undoubtedly efficient solutions for implementing multiply-accumulate (MAC)
operations on FPGA. Since DSP resources are scarce in FPGA, the advanced solution is to …

A hardware design framework for computer vision models based on reconfigurable devices

Z Fan, W Hu, F Liu, D Xu, H Guo, Y He… - ACM Transactions on …, 2024 - dl.acm.org
In computer vision, the joint development of the algorithm and computing dimensions cannot
be separated. Models and algorithms are constantly evolving, while hardware designs must …

DRViT: A Dynamic Redundancy-Aware Vision Transformer Accelerator via Algorithm and Architecture Co-design on FPGA

X Sun, Y Zhang, Q Wang, X Zou, Y Liu, Z Zeng… - Journal of Parallel and …, 2025 - Elsevier
The multi-modal artificial intelligence (MAI) has attracted significant interest due to its
capability to process and integrate data from multiple modalities, including images, text, and …

A Vision Transformer Inference Accelerator for KR260

Z Bao, H Li, W Zhang - Proceedings of the 2024 8th International …, 2024 - dl.acm.org
Vision Transformer (ViT) is pivotal in intelligent robotic vision tasks, its real-time
implementation on the edge relies on the design of ViT inference accelerators. KR260 was …