Hierarchical Sparse Plus Low Rank Compression of LLM

Pawan Kumar, Aditi Gupta

Modern large language models (LLMs) place extraordinary pressure on memory and compute budgets, making principled compression indispensable for both deployment and continued traini…