An official website of the United States government
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.
more »
« less
This content will become publicly available on March 1, 2026
An Efficient Training Algorithm for Models with Block-wise Sparsity
Large-scale machine learning (ML) models are increasingly being used in critical domains like education, lending, recruitment, healthcare, criminal justice, etc. However, the training, deployment, and utilization of these models demand substantial computational resources. To decrease computation and memory costs, machine learning models with sparse weight matrices are widely used in the literature. Among sparse models, those with special sparse structures (e.g., models with block-wise sparse weight matrices) fit better with the hardware accelerators and can decrease the memory and computation costs during the inference. Unfortunately, while there are several efficient training methods, none of them are designed to train a block-wise sparse model efficiently. As a result, the current methods for training block-wise sparse models start with full and dense models leading to inefficient training. In this work, we focus on training models with block-wise sparse matrices and propose an efficient training algorithm to decrease both computation and memory costs during training and inference. In addition, we will show that our proposed method enables us to efficiently find the right block size for the sparsity pattern during the training process. Our extensive empirical and theoretical analyses show that our algorithms can decrease the computation and memory costs significantly without a performance drop compared to baselines.
more »
« less
- PAR ID:
- 10632118
- Publisher / Repository:
- https://openreview.net/forum?id=nay3Kvw8BD
- Date Published:
- Journal Name:
- Transactions on machine learning research
- ISSN:
- 2835-8856
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Current PEFT methods for LLMs can achieve either high quality, efficient training, or scalable serving, but not all three simultaneously. To address this limitation, we investigate sparse fine-tuning and observe a remarkable improvement in generalization ability. Utilizing this key insight, we propose a family of \underline{S}tructured \underline{S}parse \underline{F}ine-\underline{T}uning (\textbf{\model}) methods for LLMs, which \textit{concurrently achieve state-of-the-art fine-tuning performance, training efficiency, and inference scalability}. \model \mbox{accomplishes this by ``selecting sparsely and computing densely". It selects a few} heads and channels in the MHA and FFN modules for each Transformer block, respectively. Next, it co-permutes weight matrices on both sides of the coupled structures in LLMs to connect the selected components in each layer into a dense submatrix. Finally, \model performs in-place gradient updates on all submatrices. Through theoretical analysis and empirical results, our method prevents overfitting and forgetting, delivers SOTA performance on both commonsense and arithmetic reasoning with 4.6$$\%$$ and 1.3$$\%$$ average improvements compared to LoRA, and surpasses full FT by 11.5$$\%$$ when generalizing to various domains after instruction tuning. Using our partial backpropagation algorithm, \model saves training memory up to 3$$\times$$ and improves latency by 1.5-2.7$$\times$$ compared to full FT, while delivering an average 10\% improvement over LoRA on both metrics. We further demonstrate that the weight updates in \model can be decoupled into adapters, enabling effective fusion, fast switch, and efficient parallelism for serving multiple fine-tuned models.more » « less
-
Quasi-Newton methods still face significant challenges in training large-scale neural networks due to additional compute costs in the Hessian related computations and instability issues in stochastic training. A well-known method, L-BFGS that efficiently approximates the Hessian using history parameter and gradient changes, suffers convergence instability in stochastic training. So far, attempts that adapt L-BFGS to large-scale stochastic training incur considerable extra overhead, which offsets its convergence benefits in wall-clock time. In this paper, we propose mL-BFGS, a lightweight momentum-based L-BFGS algorithm that paves the way for quasi-Newton (QN) methods in large-scale distributed deep neural network (DNN) optimization. mL-BFGS introduces a nearly cost-free momentum scheme into L-BFGS update and greatly reduces stochastic noise in the Hessian, therefore stabilizing convergence during stochastic optimization. For model training at a large scale, mL-BFGS approximates a block-wise Hessian, thus enabling distributing compute and memory costs across all computing nodes. We provide a supporting convergence analysis for mL-BFGS in stochastic settings. To investigate mL-BFGS’s potential in large-scale DNN training, we train benchmark neural models using mL-BFGS and compare performance with baselines (SGD, Adam, and other quasi-Newton methods). Results show that mL-BFGS achieves both noticeable iteration-wise and wall-clock speedup.more » « less
-
The record-breaking performance of deep neural networks (DNNs) comes with heavy parameter budgets, which leads to external dynamic random access memory (DRAM) for storage. The prohibitive energy of DRAM accesses makes it nontrivial for DNN deployment on resource-constrained devices, calling for minimizing the movements of weights and data in order to improve the energy efficiency. Driven by this critical bottleneck, we present SmartDeal, a hardware-friendly algorithm framework to trade higher-cost memory storage/access for lower-cost computation, in order to aggressively boost the storage and energy efficiency, for both DNN inference and training. The core technique of SmartDeal is a novel DNN weight matrix decomposition framework with respective structural constraints on each matrix factor, carefully crafted to unleash the hardware-aware efficiency potential. Specifically, we decompose each weight tensor as the product of a small basis matrix and a large structurally sparse coefficient matrix whose nonzero elements are readily quantized to the power-of-2. The resulting sparse and readily quantized DNNs enjoy greatly reduced energy consumption in data movement as well as weight storage, while incurring minimal overhead to recover the original weights thanks to the required sparse bit-operations and cost-favorable computations. Beyond inference, we take another leap to embrace energy-efficient training, by introducing several customized techniques to address the unique roadblocks arising in training while preserving the SmartDeal structures. We also design a dedicated hardware accelerator to fully utilize the new weight structure to improve the real energy efficiency and latency performance. We conduct experiments on both vision and language tasks, with nine models, four datasets, and three settings (inference-only, adaptation, and fine-tuning). Our extensive results show that 1) being applied to inference, SmartDeal achieves up to 2.44x improvement in energy efficiency as evaluated using real hardware implementations and 2) being applied to training, SmartDeal can lead to 10.56x and 4.48x reduction in the storage and the training energy cost, respectively, with usually negligible accuracy loss, compared to state-of-the-art training baselines. Our source codes are available at: https://github.com/VITA-Group/SmartDeal.more » « less
-
Binary classification is a fundamental machine learning task defined as correctly assigning new objects to one of two groups based on a set of training objects. Driven by the practical importance of binary classification, numerous machine learning techniques have been developed and refined over the last three decades. Among the most popular techniques are artificial neural networks, decision trees, ensemble methods, logistic regression, and support vector machines. We present here machine learning and pattern recognition algorithms that, unlike the commonly used techniques, are based on combinatorial optimization and make use of information on pairwise relations between the objects of the data set, whether training objects or not. These algorithms solve the respective problems optimally and efficiently, in contrast to the primarily heuristic approaches currently used for intractable problem models in pattern recognition and machine learning. The algorithms described solve efficiently the classification problem as a network flow problem on a graph. The technical tools used in the algorithm are the parametric cut procedure and a process called sparse computation that computes only the pairwise similarities that are “relevant.” Sparse computation enables the scalability of any algorithm that uses pairwise similarities. We present evidence on the effectiveness of the approaches, measured in terms of accuracy and running time, in pattern recognition, image segmentation, and general data mining.more » « less
