Xinming Yan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xinming Yan has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Bo Liu
Southeast University, Nanjing, Jiangsu, China
6 shared publications - Hao Cai
Southeast University, Nanjing, Jiangsu, China
6 shared publications - Zihan Zou
School of Integrated Circuits, Southeast University, Nanjing, China
6 shared publications - Xilong Kang
School of Integrated Circuits, Southeast University, Nanjing, China
3 shared publications - Bo Hu
School of Integrated Circuits, Southeast University, Nanjing, China
2 shared publications - Chen Zhang
School of Integrated Circuits, Southeast University, Nanjing, China
2 shared publications
Latest Publications
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (7) , 2604-2617
Low Bit-Width LLM Acceleration via Symmetric Lookup Format and Compute-in-Decoding Paradigm
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (7) , 2604-2617
Low Bit-Width LLM Acceleration via Symmetric Lookup Format and Compute-in-Decoding Paradigm
- Open Access
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 432-444
S-DMA: Sparse Diffusion Models Acceleration via Spatiality-Aware Prediction and Dimension-Adaptive Dataflow
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 432-444
Open AccessS-DMA: Sparse Diffusion Models Acceleration via Spatiality-Aware Prediction and Dimension-Adaptive Dataflow
- 29 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 44 (11) , 4127-4140
PF²A-ViT: Parameter-Free and Feature-Aware Dynamic Token Pruning Accelerator With Complementary Quantization–Encoding for Vision Transformer
Vision transformers (ViTs) have achieved outstanding performance in visual applications. However, ViT’s increasing parameters and computation overhead limit its deployment on hardware. Previous ViT accelerators focus on optimizing the core attention mechanism of ViTs due to the high overheads of language transformer-based neural networks. Nevertheless, linear layers are the actual bottleneck of ViT inference, accounting for larger than 90% FLOPs on numerous ViTs because of the short and fixed token length of ViTs. To this end, we propose P$\text {F}^{2}$ A-ViT, an algorithm and accelerator co-design framework, comprehensively accelerating both core attention and linear layers. At the algorithm level, a parameter-free and feature-aware dynamic token pruning (PF2ATP) strategy is proposed to reduce the dimension of feature maps and dynamically adjust the pruning ratio according to the complexity of the feature without complex execution of subnetworks. Meanwhile, a mixed-precision quantization (MPQ) strategy combines Hessian trace and parameter-aware signal-to-quantization-noise ratio to boost the deployment efficiency of PF2ATP. In addition, a cluster-regroup-based weight encoding strategy is proposed to compensate for the bit-wise redundant information of the quantization strategy. At the hardware level, a token pruning module based on bitonic sorters is designed to fully leverage PF2ATP. Simultaneously, a 3-D PE array with reconfigurable 4/8-bit processing elements (PEs) is designed to implement the MPQ strategy and equipped with greedy bit-wise compensation decoders to exploit the encoding strategy. Extensive experiments on multiple ViTs demonstrate the achievements of P$\text {F}^{2}$ A-ViT: 1) maximally realize$3.89\times $ speedup,$5.54\times $ energy efficiency compared to state-of-the-art ViT accelerators and 2) occupying a 2.25-mm2 area and consuming 76-mW power in 28-nm technology.- 29 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 44 (11) , 4127-4140
PF²A-ViT: Parameter-Free and Feature-Aware Dynamic Token Pruning Accelerator With Complementary Quantization–Encoding for Vision Transformer
Vision transformers (ViTs) have achieved outstanding performance in visual applications. However, ViT’s increasing parameters and computation overhead limit its deployment on hardware. Previous ViT accelerators focus on optimizing the core attention mechanism of ViTs due to the high overheads of language transformer-based neural networks. Nevertheless, linear layers are the actual bottleneck of ViT inference, accounting for larger than 90% FLOPs on numerous ViTs because of the short and fixed token length of ViTs. To this end, we propose P$\text {F}^{2}$ A-ViT, an algorithm and accelerator co-design framework, comprehensively accelerating both core attention and linear layers. At the algorithm level, a parameter-free and feature-aware dynamic token pruning (PF2ATP) strategy is proposed to reduce the dimension of feature maps and dynamically adjust the pruning ratio according to the complexity of the feature without complex execution of subnetworks. Meanwhile, a mixed-precision quantization (MPQ) strategy combines Hessian trace and parameter-aware signal-to-quantization-noise ratio to boost the deployment efficiency of PF2ATP. In addition, a cluster-regroup-based weight encoding strategy is proposed to compensate for the bit-wise redundant information of the quantization strategy. At the hardware level, a token pruning module based on bitonic sorters is designed to fully leverage PF2ATP. Simultaneously, a 3-D PE array with reconfigurable 4/8-bit processing elements (PEs) is designed to implement the MPQ strategy and equipped with greedy bit-wise compensation decoders to exploit the encoding strategy. Extensive experiments on multiple ViTs demonstrate the achievements of P$\text {F}^{2}$ A-ViT: 1) maximally realize$3.89\times $ speedup,$5.54\times $ energy efficiency compared to state-of-the-art ViT accelerators and 2) occupying a 2.25-mm2 area and consuming 76-mW power in 28-nm technology.- Open Access
- 20 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 177-182
TWDP: A Vision Transformer Accelerator with Token-Weight Dual-Pruning Strategy for Edge Device Deployment
Vision Transformers (ViTs) have attracted significant attention due to their superior accuracy compared to convolutional neural networks (CNNs) in various computer vision tasks. However, their substantial computational load and significant memory footprint lead to excessive delay and considerable data storage overhead, posing challenges for resource-limited edge device deployment. To address these issues, we present TWDP, a vision transformer accelerator employing a Token-Weight Dual-Pruning strategy to enhance the efficiency of the inference process. Firstly, we propose a parameter-free self-adaptive token pruning method to skip redundant computations in an image-dependent manner. Secondly, we apply a Hessian-aware layer-wise N:M weight pruning approach to minimize storage overhead, memory access, and computational power consumption. Additionally, to manage the complex computing patterns in ViTs, an overlapping dataflow is utilized to further reduce temporal storage and inference latency. Implemented and evaluated under an industrial 28nm technology, the proposed TWDP framework reduces 66.1% weight storage requirements and achieves an energy efficiency of 2070.9 FPS/W. Compared to state-of-the-art architectures, TWDP obtains a 1.6× energy efficiency improvement with negligible accuracy loss, demonstrating the superiority of TWDP in edge device deployment scenarios.
- 20 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 177-182
Open AccessTWDP: A Vision Transformer Accelerator with Token-Weight Dual-Pruning Strategy for Edge Device Deployment
Vision Transformers (ViTs) have attracted significant attention due to their superior accuracy compared to convolutional neural networks (CNNs) in various computer vision tasks. However, their substantial computational load and significant memory footprint lead to excessive delay and considerable data storage overhead, posing challenges for resource-limited edge device deployment. To address these issues, we present TWDP, a vision transformer accelerator employing a Token-Weight Dual-Pruning strategy to enhance the efficiency of the inference process. Firstly, we propose a parameter-free self-adaptive token pruning method to skip redundant computations in an image-dependent manner. Secondly, we apply a Hessian-aware layer-wise N:M weight pruning approach to minimize storage overhead, memory access, and computational power consumption. Additionally, to manage the complex computing patterns in ViTs, an overlapping dataflow is utilized to further reduce temporal storage and inference latency. Implemented and evaluated under an industrial 28nm technology, the proposed TWDP framework reduces 66.1% weight storage requirements and achieves an energy efficiency of 2070.9 FPS/W. Compared to state-of-the-art architectures, TWDP obtains a 1.6× energy efficiency improvement with negligible accuracy loss, demonstrating the superiority of TWDP in edge device deployment scenarios.