SpenseGPT: One-shot Pruning for Sparse and Dense GEMM LLM Inference

SpenseGPT applies one-shot pruning to LLMs to enable both sparse and dense general matrix multiply operations during inference, targeting practical deployment speedups without retraining.