MambaQuant: Quantizing the Mamba Family with Variance Aligned Rotation Methods

• URL: http://arxiv.org/abs/2501.13484v2
• Date: Thu, 06 Feb 2025 08:05:38 GMT
• Title: MambaQuant: Quantizing the Mamba Family with Variance Aligned Rotation Methods
• Authors: Zukang Xu, Yuxuan Yue, Xing Hu, Zhihang Yuan, Zixu Jiang, Zhixuan Chen, Jiangyong Yu, Chen Xu, Sifan Zhou, Dawei Yang,
• Abstract summary: Mamba is an efficient sequence model that rivals Transformers. Existing quantization methods, which have been effective for CNN and Transformer models, appear inadequate for Mamba. We propose MambaQuant, a post-training quantization framework consisting of: 1) Karhunen-Loeve Transformation (KLT) enhanced rotation, rendering the rotation matrix adaptable to diverse channel distributions, and 2) Smooth-Fused rotation, which equalizes channel variances and can merge additional parameters into model weights.
• Score: 10.08926186041001
• License:
• Abstract: Mamba is an efficient sequence model that rivals Transformers and demonstrates significant potential as a foundational architecture for various tasks. Quantization is commonly used in neural networks to reduce model size and computational latency. However, applying quantization to Mamba remains underexplored, and existing quantization methods, which have been effective for CNN and Transformer models, appear inadequate for Mamba models (e.g., Quarot suffers a 21% accuracy drop on Vim-T$^\dagger$ even under W8A8). We have pioneered the exploration of this issue and identified several key challenges. First, significant outliers are present in gate projections, output projections, and matrix multiplications. Second, Mamba's unique parallel scan further amplifies these outliers, leading to uneven and heavy-tailed data distributions. Third, even with the application of the Hadamard transform, the variance across channels in weights and activations still remains inconsistent. To these ends, we propose MambaQuant, a post-training quantization (PTQ) framework consisting of: 1) Karhunen-Loeve Transformation (KLT) enhanced rotation, rendering the rotation matrix adaptable to diverse channel distributions. 2) Smooth-Fused rotation, which equalizes channel variances and can merge additional parameters into model weights. Experiments show that MambaQuant can quantize both weights and activations into 8-bit with less than 1% accuracy loss for Mamba-based vision and language tasks. To the best of our knowledge, MambaQuant is the first comprehensive PTQ design for the Mamba family, paving the way for further advancements in its application.

Related papers

• PTQ4VM: Post-Training Quantization for Visual Mamba [9.446971590056945]
  We propose Post-Training Quantization for Visual Mamba (PTQ4VM), which introduces two key strategies: Per-Token Static (PTS) quantization and Joint Learning of Smoothing Scale and Step Size (JLSS) PTQ4VM can be applied to various Visual Mamba backbones, converting the pretrained model to a quantized format in under 15 minutes without notable quality degradation.
  arXiv  Detail & Related papers  (2024-12-29T07:21:33Z)
• Mamba-SEUNet: Mamba UNet for Monaural Speech Enhancement [54.427965535613886]
  Mamba, as a novel state-space model (SSM), has gained widespread application in natural language processing and computer vision. In this work, we introduce Mamba-SEUNet, an innovative architecture that integrates Mamba with U-Net for SE tasks.
  arXiv  Detail & Related papers  (2024-12-21T13:43:51Z)
• Bi-Mamba: Towards Accurate 1-Bit State Space Models [28.478762133816726]
  Bi-Mamba is a scalable and powerful 1-bit Mamba architecture designed for more efficient large language models. Bi-Mamba achieves performance comparable to its full-precision counterparts (e.g., FP16 or BF16) and much better accuracy than post-training-binarization (PTB) Mamba baselines.
  arXiv  Detail & Related papers  (2024-11-18T18:59:15Z)
• MaskMamba: A Hybrid Mamba-Transformer Model for Masked Image Generation [63.73137438677585]
  MaskMamba is a novel hybrid model that combines Mamba and Transformer architectures. It achieves a remarkable $54.44%$ improvement in inference speed at a resolution of $2048times 2048$ over Transformer.
  arXiv  Detail & Related papers  (2024-09-30T04:28:55Z)
• The Mamba in the Llama: Distilling and Accelerating Hybrid Models [76.64055251296548]
  We show how to distill large Transformers into linear RNNs by reusing the linear projection weights from attention layers with academic GPU resources. The resulting hybrid model achieves performance comparable to the original Transformer in chat benchmarks. We also introduce a hardware-aware speculative decoding algorithm that accelerates the inference speed of Mamba and hybrid models.
  arXiv  Detail & Related papers  (2024-08-27T17:56:11Z)
• Snakes and Ladders: Two Steps Up for VideoMamba [10.954210339694841]
  In this paper, we theoretically analyze the differences between self-attention and Mamba. We propose VideoMambaPro models that surpass VideoMamba by 1.6-2.8% and 1.1-1.9% top-1. Our two solutions are to recent advances in Vision Mamba models, and are likely to provide further improvements in future models.
  arXiv  Detail & Related papers  (2024-06-27T08:45:31Z)
• Venturing into Uncharted Waters: The Navigation Compass from Transformer to Mamba [77.21394300708172]
  Transformer, a deep neural network architecture, has long dominated the field of natural language processing and beyond. The recent introduction of Mamba challenges its supremacy, sparks considerable interest among researchers, and gives rise to a series of Mamba-based models that have exhibited notable potential. This survey paper orchestrates a comprehensive discussion, diving into essential research dimensions, covering: (i) the functioning of the Mamba mechanism and its foundation on the principles of structured state space models; (ii) the proposed improvements and the integration of Mamba with various networks, exploring its potential as a substitute for Transformers; (iii) the combination of
  arXiv  Detail & Related papers  (2024-06-24T15:27:21Z)
• Q-Mamba: On First Exploration of Vision Mamba for Image Quality Assessment [15.320011514412437]
  We take the first exploration of the recently popular foundation model, i.e., State Space Model/Mamba, in image quality assessment. We propose Q-Mamba by revisiting and adapting the Mamba model for three crucial IQA tasks. Our proposed StylePrompt enables better perception transfer capability with less computational cost.
  arXiv  Detail & Related papers  (2024-06-13T19:21:01Z)
• An Empirical Study of Mamba-based Language Models [69.74383762508805]
  Selective state-space models (SSMs) like Mamba overcome some shortcomings of Transformers. We present a direct comparison between 8B-context Mamba, Mamba-2, and Transformer models trained on the same datasets. We find that the 8B Mamba-2-Hybrid exceeds the 8B Transformer on all 12 standard tasks.
  arXiv  Detail & Related papers  (2024-06-12T05:25:15Z)
• Demystify Mamba in Vision: A Linear Attention Perspective [72.93213667713493]
  Mamba is an effective state space model with linear computation complexity. We show that Mamba shares surprising similarities with linear attention Transformer. We propose a Mamba-Inspired Linear Attention (MILA) model by incorporating the merits of these two key designs into linear attention.
  arXiv  Detail & Related papers  (2024-05-26T15:31:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.