ULSeq-TA: Ultra-Long Sequence Attention Fusion Transformer Accelerator Supporting Grouped Sparse Softmax and Dual-Path Sparse LayerNorm.
Jingyu WangLu ZhangXueqing LiHuazhong YangYongpan LiuPublished in: IEEE Trans. Comput. Aided Des. Integr. Circuits Syst. (2024)