Found 1 document(s)
Page 1 of 1
-
Non-uniform allocation can improve post-training pruning of large language models, yet quality can still degrade sharply at high sparsity. To control degradation as sparsity increases, we introduce LADDER (Layer-Wise All...