BLAST: Block-Level Adaptive Structured Matrices for Efficient Deep Neural Network Inference

• URL: http://arxiv.org/abs/2410.21262v2
• Date: Wed, 30 Oct 2024 00:38:11 GMT
• Title: BLAST: Block-Level Adaptive Structured Matrices for Efficient Deep Neural Network Inference
• Authors: Changwoo Lee, Soo Min Kwon, Qing Qu, Hun-Seok Kim,
• Abstract summary: We introduce the Block-Level Adaptive STructured (BLAST) matrix to learn and leverage efficient structures prevalent in the weight matrices of linear layers within deep learning models. We demonstrate the efficiency of using the matrix for compressing both language and vision tasks.
• Score: 15.519068157865023
• License:
• Abstract: Large-scale foundation models have demonstrated exceptional performance in language and vision tasks. However, the numerous dense matrix-vector operations involved in these large networks pose significant computational challenges during inference. To address these challenges, we introduce the Block-Level Adaptive STructured (BLAST) matrix, designed to learn and leverage efficient structures prevalent in the weight matrices of linear layers within deep learning models. Compared to existing structured matrices, the BLAST matrix offers substantial flexibility, as it can represent various types of structures that are either learned from data or computed from pre-existing weight matrices. We demonstrate the efficiency of using the BLAST matrix for compressing both language and vision tasks, showing that (i) for medium-sized models such as ViT and GPT-2, training with BLAST weights boosts performance while reducing complexity by 70% and 40%, respectively; and (ii) for large foundation models such as Llama-7B and DiT-XL, the BLAST matrix achieves a 2x compression while exhibiting the lowest performance degradation among all tested structured matrices. Our code is available at https://github.com/changwoolee/BLAST.

Related papers

• Efficient Adaptation of Pre-trained Vision Transformer via Householder Transformation [53.88562288388169]
  A common strategy for. Efficient Fine-Tuning (PEFT) of pre-trained Vision Transformers (ViTs) involves adapting the model to downstream tasks. We propose a novel PEFT approach inspired by Singular Value Decomposition (SVD) for representing the adaptation matrix. SVD decomposes a matrix into the product of a left unitary matrix, a diagonal matrix of scaling values, and a right unitary matrix.
  arXiv  Detail & Related papers  (2024-10-30T12:08:30Z)
• Searching for Efficient Linear Layers over a Continuous Space of Structured Matrices [88.33936714942996]
  We present a unifying framework that enables searching among all linear operators expressible via an Einstein summation. We show that differences in the compute-optimal scaling laws are mostly governed by a small number of variables. We find that Mixture-of-Experts (MoE) learns an MoE in every single linear layer of the model, including the projection in the attention blocks.
  arXiv  Detail & Related papers  (2024-10-03T00:44:50Z)
• Two Sparse Matrices are Better than One: Sparsifying Neural Networks with Double Sparse Factorization [0.0]
  We present Double Sparse Factorization (DSF), where we factorize each weight matrix into two sparse matrices. Our method achieves state-of-the-art results, enabling unprecedented sparsification of neural networks.
  arXiv  Detail & Related papers  (2024-09-27T15:48:39Z)
• Search for Efficient Large Language Models [52.98684997131108]
  Large Language Models (LLMs) have long held sway in the realms of artificial intelligence research. Weight pruning, quantization, and distillation have been embraced to compress LLMs, targeting memory reduction and inference acceleration. Most model compression techniques concentrate on weight optimization, overlooking the exploration of optimal architectures.
  arXiv  Detail & Related papers  (2024-09-25T21:32:12Z)
• Compute Better Spent: Replacing Dense Layers with Structured Matrices [77.61728033234233]
  We identify more efficient alternatives to dense matrices, as exemplified by the success of convolutional networks in the image domain. We show that different structures often require drastically different initialization scales and learning rates, which are crucial to performance. We propose a novel matrix family containing Monarch matrices, the Block-Train, which we show performs better than dense for the same compute on multiple tasks.
  arXiv  Detail & Related papers  (2024-06-10T13:25:43Z)
• Differentiable Learning of Generalized Structured Matrices for Efficient Deep Neural Networks [16.546708806547137]
  This paper investigates efficient deep neural networks (DNNs) to replace dense unstructured weight matrices with structured ones that possess desired properties. The challenge arises because the optimal weight matrix structure in popular neural network models is obscure in most cases and may vary from layer to layer even in the same network. We propose a generalized and differentiable framework to learn efficient structures of weight matrices by gradient descent.
  arXiv  Detail & Related papers  (2023-10-29T03:07:30Z)
• Monarch: Expressive Structured Matrices for Efficient and Accurate Training [64.6871423399431]
  Large neural networks excel in many domains, but they are expensive to train and fine-tune. A popular approach to reduce their compute or memory requirements is to replace dense weight matrices with structured ones. We propose a class of matrices (Monarch) that is hardware-efficient.
  arXiv  Detail & Related papers  (2022-04-01T17:37:29Z)
• Multi-View Spectral Clustering with High-Order Optimal Neighborhood Laplacian Matrix [57.11971786407279]
  Multi-view spectral clustering can effectively reveal the intrinsic cluster structure among data. This paper proposes a multi-view spectral clustering algorithm that learns a high-order optimal neighborhood Laplacian matrix. Our proposed algorithm generates the optimal Laplacian matrix by searching the neighborhood of the linear combination of both the first-order and high-order base.
  arXiv  Detail & Related papers  (2020-08-31T12:28:40Z)
• Block-encoding based quantum algorithm for linear systems with displacement structures [4.145426157018113]
  We present efficient and memory-reduced quantum algorithms for solving linear systems with displacement structures. The proposed block-encodings provide a quadratic speedup with respect to the dimension over classical algorithms. One of the quantum linear system solvers is applied to the linear prediction of time series.
  arXiv  Detail & Related papers  (2019-12-27T16:10:13Z)

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.