Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment Paper • 2405.03594 • Published May 6 • 7
