Fetching the paper…

BESA: Pruning Large Language Models with Blockwise Parameter-Efficient Sparsity Allocation · Around