paper-with-me

홈 › Papers

SparseSSM: Efficient Selective Structured State Space Models Can Be Pruned in One-Shot

2025-06-11 · Kaiwen Tuo, Huan Wang

State-space language models such as Mamba match Transformer quality while permitting linear complexity inference, yet still comprise billions of parameters that hinder deployment. Existing one-shot pruning methods are tailored to attention blocks and fail to account for the time-shared and discretized state-transition matrix at the heart of the selective state-space module (SSM). In this paper, we introduce SparseSSM, the first training-free pruning framework that extends the classic optimal brain surgeon (OBS) framework to state space architectures. Our layer-wise algorithm (i) derives an approximate second-order saliency score that aggregates Hessian-trace information across time steps, (ii) incorporates a component sensitivity analysis to guide feed-forward network (FFN) pruning, which also sheds light on where redundancy resides in mamba architecture, (iii) can be easily extended to semi-structured and structured sparsity. Empirically, we prune 50% of SSM weights without fine-tuning and observe no zero-shot accuracy loss, achieving the current state-of-the-art pruning algorithm for Mamba-based LLMs.

📄 PDF Abstract BibTeX arXiv:2506.09613

Code (0)

등록된 구현이 없습니다.

Tasks

MambaState Space Models

Methods 이 논문이 사용한 방법론

Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Adam 설명 없음
BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…
Attention 설명 없음
Mamba Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module.…
Multi-Head Attention 설명 없음
Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…
Label Smoothing Label Smoothing is a regularization technique that introduces noise for the labels. This accounts for the fact that datasets may have mistakes in them, so maximizing the…

Similar Papers 제목 키워드 기반

Restoring Pruned Large Language Models via Lost Component Compensation

2025-10-22 · Zijian Feng, Hanzhang Zhou, Zixiao Zhu, Tianjiao Li 외 arxiv

Pruning is a widely used technique to reduce the size and inference cost of large language models (LLMs), but it often causes performance degradation. To mitigate this, existing restoration methods typically employ param…

parameter-efficient fine-tuning

CBOL-Tuner: Classifier-pruned Bayesian optimization to explore temporally structured latent spaces for particle accelerator tuning

2024-12-02 · Mahindra Rautela, Alan Williams, Alexander Scheinker

Complex dynamical systems, such as particle accelerators, often require complicated and time-consuming tuning procedures for optimal performance. It may also be required that these procedures estimate the optimal system …

Bayesian OptimizationEfficient Explorationglobal-optimizationparameter estimation

Sample-aware Adaptive Structured Pruning for Large Language Models

2025-03-08 · Jun Kong, Xinge Ma, Jin Wang, Xuejie Zhang

Large language models (LLMs) have achieved outstanding performance in natural language processing, but enormous model sizes and high computational costs limit their practical deployment. Structured pruning can effectivel…

Bayesian Optimization

Distilled Gradual Pruning with Pruned Fine-tuning

2024-02-15 · IEEE Transactions on Artificial Intelligence 2024 2 · Federico Fontana, Romeo Lanzino, Marco Raoul Marini, Danilo Avola 외

Neural Networks (NNs) have been driving machine learning progress in recent years, but their larger models present challenges in resource-limited environments. Weight pruning reduces the computational demand, often with …

Image ClassificationKnowledge DistillationSelf-Knowledge Distillation

AP: Selective Activation for De-sparsifying Pruned Neural Networks

2022-12-09 · Shiyu Liu, Rohan Ghosh, Dylan Tan, Mehul Motani

The rectified linear unit (ReLU) is a highly successful activation function in neural networks as it allows networks to easily obtain sparse representations, which reduces overfitting in overparameterized networks. Howev…

Network Pruning