A 28nm 0.22uJ/Token Memory-Compute-Intensity-Aware CNN-Transformer Accelerator with Hybrid-Attention-Based Layer-Fusion and Cascaded Pruning for Semantic-Segmentation

P. Dong, Y. Tan, X. Liu, P. Luo, Y. Liu, L. Liang, Y. Zhou, D. Pang, M. Yung, D. Zhang, X. Huang, S-Y. Liu, Y. Wu, F. Tian, C-Y. Tsui, F. Tu, K-T. Cheng

January 2025

Abstract

Co-Corresponding Author, The First Hong Kong AI Chip at ISSCC

Type

Conference paper

Publication

International Solid-State Circuits Conference (ISSCC), San Francisco, USA