MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

Large Language Models (LLMs) are distinguished by their massive parameter counts, which typically result in significant redundancy.