Jinjin Xu

LOST: Low-rank and Sparse Pre-training for Large Language Models

ELAS: Efficient Pre-Training of Low-Rank Large Language Models via 2:4 Activation Sparsity