Sign In
insight - Adaptive Global Pruning
Gradient-Free Adaptive Global Pruning for Pre-trained Language Models: A Novel Framework for Efficient Compression
Efficiently compressing large language models through global pruning with low memory consumption.
Gradient-Free Adaptive Global Pruning for Pre-trained Language Models: A Novel Framework for Efficient Compression
AdaGP proposes a novel framework for global pruning of large language models, achieving significant performance improvements in high-sparsity regimes.