Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

SpenseGPT: Practical One-shot Pruning Enabling Sparse and Dense GEMMs for LLM Inference

About

Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup. However, its strict 50% sparsity constraint often causes non-negligible accuracy degradation under post-training pruning. Meanwhile, existing relaxed sparsity formats either require specialized compiler support or introduce runtime overheads that limit end-to-end speedup. We propose Spense, a practical hybrid sparse-dense format that splits each weight matrix into a 2:4 sparse region and a dense region. This design relaxes the effective sparsity constraint while remaining compatible with existing high-performance sparse and dense GEMM libraries, avoiding both custom compiler support and input activation expansion. Building on this format, we introduce SpenseGPT, a one-shot post-training pruning method that produces sparse and dense regions. Notably, we show that selecting the right dense regions is important, and we devise two different strategies to choose them. Experiments on Qwen3-32B and Seed-OSS-36B demonstrate that our method achieves up to 1.2x end-to-end decoding speedup on B200 GPUs with FP8 precision, while preserving accuracy. To the best of our knowledge, this is the first one-shot pruning demonstration of real-world end-to-end LLM decoding speedup from semi-structured sparse tensor cores on recent GPUs such as B200s, while maintaining model quality.

Jaeseong Lee, Seung-won Hwang, Samyam Rajbhandari• 2026

Related benchmarks

TaskDatasetResultRank
Instruction FollowingIFEval
IFEval Accuracy83.73
854
Mathematical ReasoningAIME 2024
AIME 2024 Accuracy80
60
Code GenerationLiveCodeBench v6
LCB Score0.6057
16
Knowledge-intensive Question AnsweringGPQA Diamond
GPQA Score68.18
16
multi-task average performanceAIME 2024, GPQA Diamond, IFEval, LiveCodeBench v6
Average Score73.12
16
PruningQwen3-32B
Pruning Time (hours)1.33
4
PruningSeed-OSS-36B
Pruning Time (hours)1.52
4
Showing 7 of 7 rows

Other info

Follow for update