SA-ANT: Efficient Low-Bit Group-Wise Quantization for Large Language Models via Sign-Asymmetric Adaptive Numeric Type.
Xinkuang Geng, Siting Liu, Hui Wang, Jie Han, Honglan Jiang
Browse the full DATE paper archive.
Xinkuang Geng, Siting Liu, Hui Wang, Jie Han, Honglan Jiang
Browse the full DATE paper archive.