Data-free quantization (DFQ), which facilitates model quantization without real data to address increasing concerns about data security, has garnered significant attention within the model compression community. Recently, the unique architecture of vision transformers (ViTs) has driven the development of specialized DFQ techniques. However, we observe that the synthetic images from existing methods suffer from the deficient semantics issue compared to real images, thereby compromising performance. Motivated by this, we propose SPDFQ, a Semantics Prompting Data-Free Quantization method for ViTs. First, SPDFQ incorporates Attention Priors Alignment (APA), which uses randomly generated attention priors to enhance the semantics of synthetic images. Second, SPDFQ introduces Multi-Semantic Reinforcement (MSR), which utilizes localized patch optimization to prompt efficient parameterization and diverse semantics in synthetic images. Finally, SPDFQ employs Softlabel Learning (SL), where soft learning targets are adapted to encourage more complex semantics and accommodate images augmented by MSR. Experimental results demonstrate that SPDFQ significantly outperforms existing methods. For instance, SPDFQ achieves a 15.52% increase in top-1 accuracy on ImageNet for W4A4 ViT-B
3D Gaussian Splatting (3DGS) has demonstrated impressive performance in scene reconstruction. However, most existing GS-based surface reconstruction methods focus on 3D objects or limited scenes. Directly applying these methods to large-scale scene reconstruction will pose challenges such as high memory costs, excessive time consumption, and lack of geometric detail, which makes it difficult to implement in practical applications. To address these issues, we propose a multi-agent collaborative fast 3DGS surface reconstruction framework based on distributed learning for large-scale surface reconstruction. Specifically, we develop local model compression (LMC) and model aggregation schemes (MAS) to achieve high-quality surface representation of large scenes while reducing GPU memory consumption. Extensive experiments on Urban3d, MegaNeRF, and BlendedMVS demonstrate that our proposed method can achieve fast and scalable high-fidelity surface reconstruction and photorealistic rendering. Our project page is available at \url{https://gyy456.github.io/CoSurfGS}.
Lightweight design, as a key approach to mitigate disparity between computational requirements of deep learning models and hardware performance, plays a pivotal role in advancing application of deep learning technologies on mobile and embedded devices, alongside rapid development of smart home, telemedicine, and autonomous driving. With its outstanding feature extracting capabilities, Deep Convolutional Neural Networks (DCNNs) have demonstrated superior performance in computer vision tasks. However, high computational costs and large network architectures severely limit the widespread application of DCNNs on resource-constrained hardware platforms such as smartphones, robots, and IoT devices. This paper reviews lightweight design strategies for DCNNs and examines recent research progress in both lightweight architectural design and model compression. Additionally, this paper discusses current limitations in this field of research and propose prospects for future directions, aiming to provide valuable guidance and reflection for lightweight design philosophy on deep neural networks in the field of computer vision.
While pruning methods effectively maintain model performance without extra training costs, they often focus solely on preserving crucial connections, overlooking the impact of pruned weights on subsequent fine-tuning or distillation, leading to inefficiencies. Moreover, most compression techniques for generative models have been developed primarily for GANs, tailored to specific architectures like StyleGAN, and research into compressing Diffusion models has just begun. Even more, these methods are often applicable only to GANs or Diffusion models, highlighting the need for approaches that work across both model types. In this paper, we introduce Singular Value Scaling (SVS), a versatile technique for refining pruned weights, applicable to both model types. Our analysis reveals that pruned weights often exhibit dominant singular vectors, hindering fine-tuning efficiency and leading to suboptimal performance compared to random initialization. Our method enhances weight initialization by minimizing the disparities between singular values of pruned weights, thereby improving the fine-tuning process. This approach not only guides the compressed model toward superior solutions but also significantly speeds up fine-tuning. Extensive experiments on StyleGAN2, StyleGAN3 and DDPM demonstrate that SVS improves compression performance across model types without additional training costs. Our code is available at: https://github.com/LAIT-CVLab/Singular-Value-Scaling.
Foundation model (FM) powered agent services are regarded as a promising solution to develop intelligent and personalized applications for advancing toward Artificial General Intelligence (AGI). To achieve high reliability and scalability in deploying these agent services, it is essential to collaboratively optimize computational and communication resources, thereby ensuring effective resource allocation and seamless service delivery. In pursuit of this vision, this paper proposes a unified framework aimed at providing a comprehensive survey on deploying FM-based agent services across heterogeneous devices, with the emphasis on the integration of model and resource optimization to establish a robust infrastructure for these services. Particularly, this paper begins with exploring various low-level optimization strategies during inference and studies approaches that enhance system scalability, such as parallelism techniques and resource scaling methods. The paper then discusses several prominent FMs and investigates research efforts focused on inference acceleration, including techniques such as model compression and token reduction. Moreover, the paper also investigates critical components for constructing agent services and highlights notable intelligent applications. Finally, the paper presents potential research directions for developing real-time agent services with high Quality of Service (QoS).
Large Language Models (LLMs) have achieved remarkable success, yet recent findings reveal that their deeper layers often contribute minimally and can be pruned without affecting overall performance. While some view this as an opportunity for model compression, we identify it as a training shortfall rooted in the widespread use of Pre-Layer Normalization (Pre-LN). We demonstrate that Pre-LN, commonly employed in models like GPT and LLaMA, leads to diminished gradient norms in its deeper layers, reducing their effectiveness. In contrast, Post-Layer Normalization (Post-LN) preserves larger gradient norms in deeper layers but suffers from vanishing gradients in earlier layers. To address this, we introduce Mix-LN, a novel normalization technique that combines the strengths of Pre-LN and Post-LN within the same model. Mix-LN applies Post-LN to the earlier layers and Pre-LN to the deeper layers, ensuring more uniform gradients across layers. This allows all parts of the network--both shallow and deep layers--to contribute effectively to training. Extensive experiments with various model sizes from 70M to 7B demonstrate that Mix-LN consistently outperforms both Pre-LN and Post-LN, promoting more balanced, healthier gradient norms throughout the network, and enhancing the overall quality of LLM pre-training. Furthermore, we demonstrate that models pre-trained with Mix-LN learn better compared to those using Pre-LN or Post-LN during supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), highlighting the critical importance of high-quality deep layers. By effectively addressing the inefficiencies of deep layers in current LLMs, Mix-LN unlocks their potential, enhancing model capacity without increasing model size. Our code is available at https://github.com/pixeli99/MixLN.
Since high resolution remote sensing image classification often requires a relatively high computation complexity, lightweight models tend to be practical and efficient. Model pruning is an effective method for model compression. However, existing methods rarely take into account the specificity of remote sensing images, resulting in significant accuracy loss after pruning. To this end, we propose an effective structural pruning approach for remote sensing image classification. Specifically, a pruning strategy that amplifies the differences in channel importance of the model is introduced. Then an adaptive mining loss function is designed for the fine-tuning process of the pruned model. Finally, we conducted experiments on two remote sensing classification datasets. The experimental results demonstrate that our method achieves minimal accuracy loss after compressing remote sensing classification models, achieving state-of-the-art (SoTA) performance.
Specializing large language models (LLMs) for local deployment in domain-specific use cases is necessary for strong performance while meeting latency and privacy constraints. However, conventional task-specific adaptation approaches do not show simultaneous memory saving and inference speedup at deployment time. Practical compression techniques like quantization and pruning require dedicated hardware or kernel support to achieve measured inference speedup. We develop TrimLLM based on the layer-wise specialization phenomenon we empirically observed and verified on contemporary LLMs. TrimLLM reduces the depth of LLMs via progressive layer dropping. We show it retains LLMs' capacity in specific domains and achieves inference speedup irrespective of hardware and deep learning frameworks. We evaluated TrimLLM on LLMs of various sizes for inference; models adapted on medical, legal, and financial datasets all demonstrate $2.1-5.7\times$ inference speedup on consumer GPUs and up to $3.1\times$ speedup on A100 when compared to state-of-the-art model compression algorithms, with no loss in accuracy at 50$\sim$60\% model compression ratio.
Knowledge distillation (KD) is a model compression technique that transfers knowledge from a large teacher model to a smaller student model to enhance its performance. Existing methods often assume that the student model is inherently inferior to the teacher model. However, we identify that the fundamental issue affecting student performance is the bias transferred by the teacher. Current KD frameworks transmit both right and wrong knowledge, introducing bias that misleads the student model. To address this issue, we propose a novel strategy to rectify bias and greatly improve the student model's performance. Our strategy involves three steps: First, we differentiate knowledge and design a bias elimination method to filter out biases, retaining only the right knowledge for the student model to learn. Next, we propose a bias rectification method to rectify the teacher model's wrong predictions, fundamentally addressing bias interference. The student model learns from both the right knowledge and the rectified biases, greatly improving its prediction accuracy. Additionally, we introduce a dynamic learning approach with a loss function that updates weights dynamically, allowing the student model to quickly learn right knowledge-based easy tasks initially and tackle hard tasks corresponding to biases later, greatly enhancing the student model's learning efficiency. To the best of our knowledge, this is the first strategy enabling the student model to surpass the teacher model. Experiments demonstrate that our strategy, as a plug-and-play module, is versatile across various mainstream KD frameworks. We will release our code after the paper is accepted.
This paper proposes small and efficient machine learning models (TinyML) for resource-constrained edge devices, specifically for on-device indoor localisation. Typical approaches for indoor localisation rely on centralised remote processing of data transmitted from lower powered devices such as wearables. However, there are several benefits for moving this to the edge device itself, including increased battery life, enhanced privacy, reduced latency and lowered operational costs, all of which are key for common applications such as health monitoring. The work focuses on model compression techniques, including quantization and knowledge distillation, to significantly reduce the model size while maintaining high predictive performance. We base our work on a large state-of-the-art transformer-based model and seek to deploy it within low-power MCUs. We also propose a state-space-based architecture using Mamba as a more compact alternative to the transformer. Our results show that the quantized transformer model performs well within a 64 KB RAM constraint, achieving an effective balance between model size and localisation precision. Additionally, the compact Mamba model has strong performance under even tighter constraints, such as a 32 KB of RAM, without the need for model compression, making it a viable option for more resource-limited environments. We demonstrate that, through our framework, it is feasible to deploy advanced indoor localisation models onto low-power MCUs with restricted memory limitations. The application of these TinyML models in healthcare has the potential to revolutionize patient monitoring by providing accurate, real-time location data while minimizing power consumption, increasing data privacy, improving latency and reducing infrastructure costs.
We consider the problem of model compression for Large Language Models (LLMs) at post-training time, where the task is to compress a well-trained model using only a small set of calibration input data. In this work, we introduce a new low-rank approach to correct for quantization errors of \emph{activations} in LLMs: we propose to add low-rank weight matrices in full precision that act on the \emph{unquantized} activations. We then solve a joint optimization problem over the quantized representation of the weights and additional low-rank weight matrices to quantize both weights and activations. We focus on the case of 4-bit weight-and-activation quantization (W4A4). Using ranks equivalent to 10\% of the original weight matrix size, our approach reduces the accuracy gap with the original model by more than 50\%. Using ranks equivalent to 30\% of the original weight matrix, the accuracy gap is closed completely. We demonstrate our results on four recent LLMs, namely Llama-2, Llama-3, Phi-3 and Mixtral models.
The rapid growth of the big neural network models puts forward new requirements for lightweight network representation methods. The traditional methods based on model compression have achieved great success, especially VQ technology which realizes the high compression ratio of models by sharing code words. However, because each layer of the network needs to build a code table, the traditional top-down compression technology lacks attention to the underlying commonalities, resulting in limited compression rate and frequent memory access. In this paper, we propose a bottom-up method to share the universal codebook among multiple neural networks, which not only effectively reduces the number of codebooks but also further reduces the memory access and chip area by storing static code tables in the built-in ROM. Specifically, we introduce VQ4ALL, a VQ-based method that utilizes codewords to enable the construction of various neural networks and achieve efficient representations. The core idea of our method is to adopt a kernel density estimation approach to extract a universal codebook and then progressively construct different low-bit networks by updating differentiable assignments. Experimental results demonstrate that VQ4ALL achieves compression rates exceeding 16 $\times$ while preserving high accuracy across multiple network architectures, highlighting its effectiveness and versatility.
This work focus on how to stabilize and lossless model compression, aiming to reduce model complexity and enhance efficiency without sacrificing performance due to compression errors. A key challenge is effectively leveraging compression errors and defining the boundaries for lossless compression to minimize model loss. i.e., compression for better. Currently, there is no systematic approach to determining this error boundary or understanding its specific impact on model performance. We propose a general \textbf{L}oss\textbf{L}ess \textbf{C}ompression theoretical framework (\textbf{LLC}), which further delineates the compression neighborhood and higher-order analysis boundaries through the total differential, thereby specifying the error range within which a model can be compressed without loss. To verify the effectiveness of LLC, we apply various compression techniques, including quantization and decomposition. Specifically, for quantization, we reformulate the classic quantization search problem as a grouped knapsack problem within the lossless neighborhood, achieving lossless quantization while improving computational efficiency. For decomposition, LLC addresses the approximation problem under low-rank constraints, automatically determining the rank for each layer and producing lossless low-rank models. We conduct extensive experiments on multiple neural network architectures on different datasets. The results show that without fancy tricks, LLC can effectively achieve lossless model compression. Our code will be made publicly.
Low-rank factorization is a popular model compression technique that minimizes the error $\delta$ between approximated and original weight matrices. Despite achieving performances close to the original models when $\delta$ is optimized, a performance discrepancy remains due to the separate optimization processes for low-rank factorization and model performance, resulting in unavoidable losses. We address this issue by introducing a novel joint optimization strategy for lossless low-rank weight factorization, which, for the first time, enhances the model's performance beyond the original. Our approach begins with a theoretical analysis of the relationship between low-rank factorization and model optimization objectives, establishing a precise perturbation range for matrix factorization errors on model performance. This challenge is then reformulated as a numerical rank deficiency problem with inequality constraints and develop a joint objective that simultaneously addresses factorization error and model performance. Based on the above analysis, we propose two optimization algorithms: \textbf{a lossless optimization algorithm} that maximizes model accuracy while ensuring compression, and \textbf{a compact optimization algorithm} that minimizes model size while preserving performance. These algorithms do not require fine-tuning and can directly compress numerous deep models to achieve lossless results. Our methods demonstrate robust efficacy across various vision and language tasks. For example, the compressed model reduced by 70\% on ResNext50 outperforms the original. Our code will be made public.
Spiking Neural Networks (SNNs) are amenable to deployment on edge devices and neuromorphic hardware due to their lower dissipation. Recently, SNN-based transformers have garnered significant interest, incorporating attention mechanisms akin to their counterparts in Artificial Neural Networks (ANNs) while demonstrating excellent performance. However, deploying large spiking transformer models on resource-constrained edge devices such as mobile phones, still poses significant challenges resulted from the high computational demands of large uncompressed high-precision models. In this work, we introduce a novel heterogeneous quantization method for compressing spiking transformers through layer-wise quantization. Our approach optimizes the quantization of each layer using one of two distinct quantization schemes, i.e., uniform or power-of-two quantification, with mixed bit resolutions. Our heterogeneous quantization demonstrates the feasibility of maintaining high performance for spiking transformers while utilizing an average effective resolution of 3.14-3.67 bits with less than a 1% accuracy drop on DVS Gesture and CIFAR10-DVS datasets. It attains a model compression rate of 8.71x-10.19x for standard floating-point spiking transformers. Moreover, the proposed approach achieves a significant energy reduction of 5.69x, 8.72x, and 10.2x while maintaining high accuracy levels of 85.3%, 97.57%, and 80.4% on N-Caltech101, DVS-Gesture, and CIFAR10-DVS datasets, respectively.
Large language models have transformed the comprehension and generation of natural language tasks, but they come with substantial memory and computational requirements. Quantization techniques have emerged as a promising avenue for addressing these challenges while preserving accuracy and making energy efficient. We propose CPTQuant, a comprehensive strategy that introduces correlation-based (CMPQ), pruning-based (PMPQ), and Taylor decomposition-based (TDMPQ) mixed precision techniques. CMPQ adapts the precision level based on canonical correlation analysis of different layers. PMPQ optimizes precision layer-wise based on their sensitivity to sparsity. TDMPQ modifies precision using Taylor decomposition to assess each layer's sensitivity to input perturbation. These strategies allocate higher precision to more sensitive layers while diminishing precision to robust layers. CPTQuant assesses the performance across BERT, OPT-125M, OPT-350M, OPT-1.3B, and OPT-2.7B. We demonstrate up to 4x compression and a 2x-fold increase in efficiency with minimal accuracy drop compared to Hugging Face FP16. PMPQ stands out for achieving a considerably higher model compression. Sensitivity analyses across various LLMs show that the initial and final 30% of layers exhibit higher sensitivities than the remaining layers. PMPQ demonstrates an 11% higher compression ratio than other methods for classification tasks, while TDMPQ achieves a 30% greater compression ratio for language modeling tasks.
In many domains, the most successful AI models tend to be the largest, indeed often too large to be handled by AI players with limited computational resources. To mitigate this, a number of compression methods have been developed, including methods that prune the network down to high sparsity whilst retaining performance. The best-performing pruning techniques are often those that use second-order curvature information (such as an estimate of the Fisher information matrix) to score the importance of each weight and to predict the optimal compensation for weight deletion. However, these methods are difficult to scale to high-dimensional parameter spaces without making heavy approximations. Here, we propose the FishLeg surgeon (FLS), a new second-order pruning method based on the Fisher-Legendre (FishLeg) optimizer. At the heart of FishLeg is a meta-learning approach to amortising the action of the inverse FIM, which brings a number of advantages. Firstly, the parameterisation enables the use of flexible tensor factorisation techniques to improve computational and memory efficiency without sacrificing much accuracy, alleviating challenges associated with scalability of most second-order pruning methods. Secondly, directly estimating the inverse FIM leads to less sensitivity to the amplification of stochasticity during inversion, thereby resulting in more precise estimates. Thirdly, our approach also allows for progressive assimilation of the curvature into the parameterisation. In the gradual pruning regime, this results in a more efficient estimate refinement as opposed to re-estimation. We find that FishLeg achieves higher or comparable performance against two common baselines in the area, most notably in the high sparsity regime when considering a ResNet18 model on CIFAR-10 (84% accuracy at 95% sparsity vs 60% for OBS) and TinyIM (53% accuracy at 80% sparsity vs 48% for OBS).
The high computational cost and slow inference time are major obstacles to deploying the video diffusion model (VDM) in practical applications. To overcome this, we introduce a new Video Diffusion Model Compression approach using individual content and motion dynamics preserved pruning and consistency loss. First, we empirically observe that deeper VDM layers are crucial for maintaining the quality of \textbf{motion dynamics} e.g., coherence of the entire video, while shallower layers are more focused on \textbf{individual content} e.g., individual frames. Therefore, we prune redundant blocks from the shallower layers while preserving more of the deeper layers, resulting in a lightweight VDM variant called VDMini. Additionally, we propose an \textbf{Individual Content and Motion Dynamics (ICMD)} Consistency Loss to gain comparable generation performance as larger VDM, i.e., the teacher to VDMini i.e., the student. Particularly, we first use the Individual Content Distillation (ICD) Loss to ensure consistency in the features of each generated frame between the teacher and student models. Next, we introduce a Multi-frame Content Adversarial (MCA) Loss to enhance the motion dynamics across the generated video as a whole. This method significantly accelerates inference time while maintaining high-quality video generation. Extensive experiments demonstrate the effectiveness of our VDMini on two important video generation tasks, Text-to-Video (T2V) and Image-to-Video (I2V), where we respectively achieve an average 2.5 $\times$ and 1.4 $\times$ speed up for the I2V method SF-V and the T2V method T2V-Turbo-v2, while maintaining the quality of the generated videos on two benchmarks, i.e., UCF101 and VBench.
Knowledge distillation approaches are model compression techniques, with the goal of training a highly performant student model by using a teacher network that is larger or contains a different inductive bias. These approaches are particularly useful when applied to large computer vision foundation models, which can be compressed into smaller variants that retain desirable properties such as improved robustness. This paper presents a label-free knowledge distillation approach called Teacher in the Middle (TinTeM), which improves on previous methods by learning an approximately orthogonal mapping from the latent space of the teacher to the student network. This produces a more faithful student, which better replicates the behavior of the teacher network across a range of benchmarks testing model robustness, generalisability and out-of-distribution detection. It is further shown that knowledge distillation with TinTeM on task specific datasets leads to more accurate models with greater generalisability and OOD detection performance, and that this technique provides a competitive pathway for training highly performant lightweight models on small datasets.
As text-to-image models grow increasingly powerful and complex, their burgeoning size presents a significant obstacle to widespread adoption, especially on resource-constrained devices. This paper presents a pioneering study on post-training pruning of Stable Diffusion 2, addressing the critical need for model compression in text-to-image domain. Our study tackles the pruning techniques for the previously unexplored multi-modal generation models, and particularly examines the pruning impact on the textual component and the image generation component separately. We conduct a comprehensive comparison on pruning the model or the single component of the model in various sparsities. Our results yield previously undocumented findings. For example, contrary to established trends in language model pruning, we discover that simple magnitude pruning outperforms more advanced techniques in text-to-image context. Furthermore, our results show that Stable Diffusion 2 can be pruned to 38.5% sparsity with minimal quality loss, achieving a significant reduction in model size. We propose an optimal pruning configuration that prunes the text encoder to 47.5% and the diffusion generator to 35%. This configuration maintains image generation quality while substantially reducing computational requirements. In addition, our work uncovers intriguing questions about information encoding in text-to-image models: we observe that pruning beyond certain thresholds leads to sudden performance drops (unreadable images), suggesting that specific weights encode critical semantics information. This finding opens new avenues for future research in model compression, interoperability, and bias identification in text-to-image models. By providing crucial insights into the pruning behavior of text-to-image models, our study lays the groundwork for developing more efficient and accessible AI-driven image generation systems