This paper introduces MotionLLaMA, a unified framework for motion synthesis and comprehension, along with a novel full-body motion tokenizer called the HoMi Tokenizer. MotionLLaMA is developed based on three core principles. First, it establishes a powerful unified representation space through the HoMi Tokenizer. Using a single codebook, the HoMi Tokenizer in MotionLLaMA achieves reconstruction accuracy comparable to residual vector quantization tokenizers utilizing six codebooks, outperforming all existing single-codebook tokenizers. Second, MotionLLaMA integrates a large language model to tackle various motion-related tasks. This integration bridges various modalities, facilitating both comprehensive and intricate motion synthesis and comprehension. Third, MotionLLaMA introduces the MotionHub dataset, currently the most extensive multimodal, multitask motion dataset, which enables fine-tuning of large language models. Extensive experimental results demonstrate that MotionLLaMA not only covers the widest range of motion-related tasks but also achieves state-of-the-art (SOTA) performance in motion completion, interaction dual-person text-to-motion, and all comprehension tasks while reaching performance comparable to SOTA in the remaining tasks. The code and MotionHub dataset are publicly available.
The article discusses the use of low cost System-On-Module (SOM) platforms for the implementation of efficient hyperspectral imaging (HSI) processors for application in autonomous driving. The work addresses the challenges of shaping and deploying multiple layer fully convolutional networks (FCN) for low-latency, on-board image semantic segmentation using resource- and power-constrained processing devices. The paper describes in detail the steps followed to redesign and customize a successfully trained HSI segmentation lightweight FCN that was previously tested on a high-end heterogeneous multiprocessing system-on-chip (MPSoC) to accommodate it to the constraints imposed by a low-cost SOM. This SOM features a lower-end but much cheaper MPSoC suitable for the deployment of automatic driving systems (ADS). In particular the article reports the data- and hardware-specific quantization techniques utilized to fit the FCN into a commercial fixed-point programmable AI coprocessor IP, and proposes a full customized post-training quantization scheme to reduce computation and storage costs without compromising segmentation accuracy.
Communication overhead is a key challenge in distributed deep learning, especially on slower Ethernet interconnects, and given current hardware trends, communication is likely to become a major bottleneck. While gradient compression techniques have been explored for SGD and Adam, the Lion optimizer has the distinct advantage that its update vectors are the output of a sign operation, enabling straightforward quantization. However, simply compressing updates for communication and using techniques like majority voting fails to lead to end-to-end speedups due to inefficient communication algorithms and reduced convergence. We analyze three factors critical to distributed learning with Lion: optimizing communication methods, identifying effective quantization methods, and assessing the necessity of momentum synchronization. Our findings show that quantization techniques adapted to Lion and selective momentum synchronization can significantly reduce communication costs while maintaining convergence. We combine these into Lion Cub, which enables up to 5x speedups in end-to-end training compared to Lion. This highlights Lion's potential as a communication-efficient solution for distributed training.
Time series are ubiquitous in numerous science and engineering domains, e.g., signal processing, bioinformatics, and astronomy. Previous work has verified the efficacy of symbolic time series representation in a variety of engineering applications due to its storage efficiency and numerosity reduction. The most recent symbolic aggregate approximation technique, ABBA, has been shown to preserve essential shape information of time series and improve downstream applications, e.g., neural network inference regarding prediction and anomaly detection in time series. Motivated by the emergence of high-performance hardware which enables efficient computation for low bit-width representations, we present a new quantization-based ABBA symbolic approximation technique, QABBA, which exhibits improved storage efficiency while retaining the original speed and accuracy of symbolic reconstruction. We prove an upper bound for the error arising from quantization and discuss how the number of bits should be chosen to balance this with other errors. An application of QABBA with large language models (LLMs) for time series regression is also presented, and its utility is investigated. By representing the symbolic chain of patterns on time series, QABBA not only avoids the training of embedding from scratch, but also achieves a new state-of-the-art on Monash regression dataset. The symbolic approximation to the time series offers a more efficient way to fine-tune LLMs on the time series regression task which contains various application domains. We further present a set of extensive experiments performed across various well-established datasets to demonstrate the advantages of the QABBA method for symbolic approximation.
Pretrained models have revolutionized deep learning by enabling significant performance improvements across a wide range of tasks, leveraging large-scale, pre-learned knowledge representations. However, deploying these models in real-world multi-task learning (MTL) scenarios poses substantial challenges, primarily due to high computational costs and inefficiencies in inference. Traditional approaches such as pruning, quantization, and knowledge distillation have been explored to mitigate these issues, but they often fall short in fully addressing the complexities of multi-task environments. This paper introduces \textbf{\underline{S}}elective \textbf{\underline{T}}ask \textbf{\underline{A}}rithmetic \underline{\textbf{(STA)}}, a training-free framework designed to enhance multi-task performance through task-specific parameter fusion. STA addresses three key challenges: (i) \textbf{Parameter importance diversity: } Recognizing that different tasks relie on distinct parameters, STA employs a loss-sensitive parameter importance metric derived from a first-order Taylor expansion to accurately measure the importance of parameters for each task. (ii) \textbf{Over-reliance on hyperparameter tuning: }By enhancing the sparsity of task vectors through parameter importance metrics, STA reduces the need for extensive hyperparameter tuning, thereby improving the generalization and robustness of the model. (iii) \textbf{Neglect of other abilities in task arithmetic: } Previous works have largely overlooked the potential for more precise task forgetting. STA leverages its parameter importance metric to achieve more controlled and effective task forgetting, minimizing the impact of noisy elements that can degrade model performance. Experimental results demonstrate that STA achieves superior multi-task performance across benchmarks and excellent performance in task forgetting.
In frequency division duplex (FDD) massive MIMO systems, a major challenge lies in acquiring the downlink channel state information}\ (CSI) at the base station (BS) from limited feedback sent by the user equipment (UE). To tackle this fundamental task, our contribution is twofold: First, a simple feedback framework is proposed, where a compression and Gaussian dithering-based quantization strategy is adopted at the UE side, and then a maximum likelihood estimator (MLE) is formulated at the BS side. Recoverability of the MIMO channel under the widely used double directional model is established. Specifically, analyses are presented for two compression schemes -- showing one being more overhead-economical and the other computationally lighter at the UE side. Second, to realize the MLE, an alternating direction method of multipliers (ADMM) algorithm is proposed. The algorithm is carefully designed to integrate a sophisticated harmonic retrieval (HR) solver as subroutine, which turns out to be the key of effectively tackling this hard MLE problem.Extensive numerical experiments are conducted to validate the efficacy of our approach.
The cost of deploying vision transformers increasingly represents a barrier to wider industrial adoption. Existing compression requires additional end-to-end fine-tuning or incurs a significant drawback to runtime, thus making them ill-suited for online inference. We introduce the $\textbf{Visual Word Tokenizer}$ (VWT), a training-free method for reducing energy costs while retaining performance and runtime. The VWT groups patches (visual subwords) that are frequently used into visual words while infrequent ones remain intact. To do so, intra-image or inter-image statistics are leveraged to identify similar visual concepts for compression. Experimentally, we demonstrate a reduction in wattage of up to 19% with only a 20% increase in runtime at most. Comparative approaches of 8-bit quantization and token merging achieve a lower or similar energy efficiency but exact a higher toll on runtime (up to $2\times$ or more). Our results indicate that VWTs are well-suited for efficient online inference with a marginal compromise on performance.
Visual tokenizers are fundamental to image generation. They convert visual data into discrete tokens, enabling transformer-based models to excel at image generation. Despite their success, VQ-based tokenizers like VQGAN face significant limitations due to constrained vocabulary sizes. Simply expanding the codebook often leads to training instability and diminishing performance gains, making scalability a critical challenge. In this work, we introduce Factorized Quantization (FQ), a novel approach that revitalizes VQ-based tokenizers by decomposing a large codebook into multiple independent sub-codebooks. This factorization reduces the lookup complexity of large codebooks, enabling more efficient and scalable visual tokenization. To ensure each sub-codebook captures distinct and complementary information, we propose a disentanglement regularization that explicitly reduces redundancy, promoting diversity across the sub-codebooks. Furthermore, we integrate representation learning into the training process, leveraging pretrained vision models like CLIP and DINO to infuse semantic richness into the learned representations. This design ensures our tokenizer captures diverse semantic levels, leading to more expressive and disentangled representations. Experiments show that the proposed FQGAN model substantially improves the reconstruction quality of visual tokenizers, achieving state-of-the-art performance. We further demonstrate that this tokenizer can be effectively adapted into auto-regressive image generation. https://showlab.github.io/FQGAN
Since 2023, Vector Quantization (VQ)-based discrete generation methods have rapidly dominated human motion generation, primarily surpassing diffusion-based continuous generation methods in standard performance metrics. However, VQ-based methods have inherent limitations. Representing continuous motion data as limited discrete tokens leads to inevitable information loss, reduces the diversity of generated motions, and restricts their ability to function effectively as motion priors or generation guidance. In contrast, the continuous space generation nature of diffusion-based methods makes them well-suited to address these limitations and with even potential for model scalability. In this work, we systematically investigate why current VQ-based methods perform well and explore the limitations of existing diffusion-based methods from the perspective of motion data representation and distribution. Drawing on these insights, we preserve the inherent strengths of a diffusion-based human motion generation model and gradually optimize it with inspiration from VQ-based approaches. Our approach introduces a human motion diffusion model enabled to perform bidirectional masked autoregression, optimized with a reformed data representation and distribution. Additionally, we also propose more robust evaluation methods to fairly assess different-based methods. Extensive experiments on benchmark human motion generation datasets demonstrate that our method excels previous methods and achieves state-of-the-art performances.
Large Multimodal Models (LMMs) have demonstrated impressive capabilities in visual-language tasks but face significant deployment challenges due to their high computational demands. While recent token reduction methods show promise for accelerating LMMs, they typically require extensive retraining or fine-tuning, making them impractical for many state-of-the-art models, especially those with proprietary training data. We propose freePruner, a training-free token reduction approach that can be directly applied to any open-source LMM without additional training. Unlike existing methods that rely heavily on token merging operations, freePruner employs a two-stage token selection strategy: (1) identifying pivotal tokens that capture high-level semantic information using our designed contribution degree metric, and (2) selecting complementary tokens that preserve essential low-level visual details through attention pattern analysis. Extensive experiments demonstrate that freePruner achieves 2x acceleration while maintaining comparable performance across mainstream visual question-answering benchmarks in the training-free setting. Moreover, freePruner is orthogonal to and can be combined with other post-training acceleration techniques, such as post-training quantization, providing a practical solution for efficient LMM deployment.
Vector quantization is a technique in machine learning that discretizes continuous representations into a set of discrete vectors. It is widely employed in tokenizing data representations for large language models, diffusion models, and other generative models. Despite its prevalence, the characteristics and behaviors of vector quantization in generative models remain largely underexplored. In this study, we investigate representation collapse in vector quantization - a critical degradation where codebook tokens or latent embeddings lose their discriminative power by converging to a limited subset of values. This collapse fundamentally compromises the model's ability to capture diverse data patterns. By leveraging both synthetic and real datasets, we identify the severity of each type of collapses and triggering conditions. Our analysis reveals that restricted initialization and limited encoder capacity result in tokens collapse and embeddings collapse. Building on these findings, we propose potential solutions aimed at mitigating each collapse. To the best of our knowledge, this is the first comprehensive study examining representation collapsing problems in vector quantization.
As a very common type of video, face videos often appear in movies, talk shows, live broadcasts, and other scenes. Real-world online videos are often plagued by degradations such as blurring and quantization noise, due to the high compression ratio caused by high communication costs and limited transmission bandwidth. These degradations have a particularly serious impact on face videos because the human visual system is highly sensitive to facial details. Despite the significant advancement in video face enhancement, current methods still suffer from $i)$ long processing time and $ii)$ inconsistent spatial-temporal visual effects (e.g., flickering). This study proposes a novel and efficient blind video face enhancement method to overcome the above two challenges, restoring high-quality videos from their compressed low-quality versions with an effective de-flickering mechanism. In particular, the proposed method develops upon a 3D-VQGAN backbone associated with spatial-temporal codebooks recording high-quality portrait features and residual-based temporal information. We develop a two-stage learning framework for the model. In Stage \Rmnum{1}, we learn the model with a regularizer mitigating the codebook collapse problem. In Stage \Rmnum{2}, we learn two transformers to lookup code from the codebooks and further update the encoder of low-quality videos. Experiments conducted on the VFHQ-Test dataset demonstrate that our method surpasses the current state-of-the-art blind face video restoration and de-flickering methods on both efficiency and effectiveness. Code is available at \url{https://github.com/Dixin-Lab/BFVR-STC}.
Transformer-based large language models (LLMs) have achieved remarkable success as model sizes continue to grow, yet their deployment remains challenging due to significant computational and memory demands. Quantization has emerged as a promising solution, and state-of-the-art quantization algorithms for LLMs introduce the need for mixed-precision matrix multiplication (mpGEMM), where lower-precision weights are multiplied with higher-precision activations. Despite its benefits, current hardware accelerators such as GPUs and TPUs lack native support for efficient mpGEMM, leading to inefficient dequantization operations in the main sequential loop. To address this limitation, we introduce MixPE, a specialized mixed-precision processing element designed for efficient low-bit quantization in LLM inference. MixPE leverages two key innovations to minimize dequantization overhead and unlock the full potential of low-bit quantization. First, recognizing that scale and zero point are shared within each quantization group, we propose performing dequantization after per-group mpGEMM, significantly reducing dequantization overhead. Second, instead of relying on conventional multipliers, MixPE utilizes efficient shift\&add operations for multiplication, optimizing both computation and energy efficiency. Our experimental results demonstrate that MixPE surpasses the state-of-the-art quantization accelerators by $2.6\times$ speedup and $1.4\times$ energy reduction.
One-shot voice conversion (VC) is a method that enables the transformation between any two speakers using only a single target speaker utterance. Existing methods often rely on complex architectures and pre-trained speaker verification (SV) models to improve the fidelity of converted speech. Recent works utilizing K-means quantization (KQ) with self-supervised learning (SSL) features have proven capable of capturing content information from speech. However, they often struggle to preserve speaking variation, such as prosodic detail and phonetic variation, particularly with smaller codebooks. In this work, we propose a simple yet effective one-shot VC model that utilizes the characteristics of SSL features and speech attributes. Our approach addresses the issue of losing speaking variation, enabling high-fidelity voice conversion trained with only reconstruction losses, without requiring external speaker embeddings. We demonstrate the performance of our model across 6 evaluation metrics, with results highlighting the benefits of the speaking variation compensation method.
Encoder-based transformers, powered by self-attention layers, have revolutionized machine learning with their context-aware representations. However, their quadratic growth in computational and memory demands presents significant bottlenecks. Analog-Mixed-Signal Process-in-Memory (AMS-PiM) architectures address these challenges by enabling efficient on-chip processing. Traditionally, AMS-PiM relies on Quantization-Aware Training (QAT), which is hardware-efficient but requires extensive retraining to adapt models to AMS-PiMs, making it increasingly impractical for transformer models. Post-Training Quantization (PTQ) mitigates this training overhead but introduces significant hardware inefficiencies. PTQ relies on dequantization-quantization (DQ-Q) processes, floating-point units (FPUs), and high-ENOB (Effective Number of Bits) analog-to-digital converters (ADCs). Particularly, High-ENOB ADCs scale exponentially in area and energy ($2^{ENOB}$), reduce sensing margins, and increase susceptibility to process, voltage, and temperature (PVT) variations, further compounding PTQ's challenges in AMS-PiM systems. To overcome these limitations, we propose RAP, an AMS-PiM architecture that eliminates DQ-Q processes, introduces FPU- and division-free nonlinear processing, and employs a low-ENOB-ADC-based sparse Matrix Vector multiplication technique. Using the proposed techniques, RAP improves error resiliency, area/energy efficiency, and computational speed while preserving numerical stability. Experimental results demonstrate that RAP outperforms state-of-the-art GPUs and conventional PiM architectures in energy efficiency, latency, and accuracy, making it a scalable solution for the efficient deployment of transformers.
Transformer-based diffusion models, dubbed Diffusion Transformers (DiTs), have achieved state-of-the-art performance in image and video generation tasks. However, their large model size and slow inference speed limit their practical applications, calling for model compression methods such as quantization. Unfortunately, existing DiT quantization methods overlook (1) the impact of reconstruction and (2) the varying quantization sensitivities across different layers, which hinder their achievable performance. To tackle these issues, we propose innovative time-aware quantization for DiTs (TaQ-DiT). Specifically, (1) we observe a non-convergence issue when reconstructing weights and activations separately during quantization and introduce a joint reconstruction method to resolve this problem. (2) We discover that Post-GELU activations are particularly sensitive to quantization due to their significant variability across different denoising steps as well as extreme asymmetries and variations within each step. To address this, we propose time-variance-aware transformations to facilitate more effective quantization. Experimental results show that when quantizing DiTs' weights to 4-bit and activations to 8-bit (W4A8), our method significantly surpasses previous quantization methods.
Deep neural networks, while achieving remarkable success across diverse tasks, demand significant resources, including computation, GPU memory, bandwidth, storage, and energy. Network quantization, as a standard compression and acceleration technique, reduces storage costs and enables potential inference acceleration by discretizing network weights and activations into a finite set of integer values. However, current quantization methods are often complex and sensitive, requiring extensive task-specific hyperparameters, where even a single misconfiguration can impair model performance, limiting generality across different models and tasks. In this paper, we propose Quantization without Tears (QwT), a method that simultaneously achieves quantization speed, accuracy, simplicity, and generality. The key insight of QwT is to incorporate a lightweight additional structure into the quantized network to mitigate information loss during quantization. This structure consists solely of a small set of linear layers, keeping the method simple and efficient. More importantly, it provides a closed-form solution, allowing us to improve accuracy effortlessly under 2 minutes. Extensive experiments across various vision, language, and multimodal tasks demonstrate that QwT is both highly effective and versatile. In fact, our approach offers a robust solution for network quantization that combines simplicity, accuracy, and adaptability, which provides new insights for the design of novel quantization paradigms.
Fine-tuning large language models (LLMs) under resource constraints is a significant challenge in deep learning. Low-Rank Adaptation (LoRA), pruning, and quantization are all effective methods for improving resource efficiency. However, combining them directly often results in suboptimal performance, especially with uniform quantization across all model layers. This is due to the complex, uneven interlayer relationships introduced by pruning, necessitating more refined quantization strategies. To address this, we propose AutoMixQ, an end-to-end optimization framework that selects optimal quantization configurations for each LLM layer. AutoMixQ leverages lightweight performance models to guide the selection process, significantly reducing time and computational resources compared to exhaustive search methods. By incorporating Pareto optimality, AutoMixQ balances memory usage and performance, approaching the upper bounds of model capability under strict resource constraints. Our experiments on widely used benchmarks show that AutoMixQ reduces memory consumption while achieving superior performance. For example, at a 30\% pruning rate in LLaMA-7B, AutoMixQ achieved 66.21\% on BoolQ compared to 62.45\% for LoRA and 58.96\% for LoftQ, while reducing memory consumption by 35.5\% compared to LoRA and 27.5\% compared to LoftQ.
Intelligent omni-surfaces (IOSs) with 360-degree electromagnetic radiation significantly improves the performance of wireless systems, while an adversarial IOS also poses a significant potential risk for physical layer security. In this paper, we propose a "DISCO" IOS (DIOS) based fully-passive jammer (FPJ) that can launch omnidirectional fully-passive jamming attacks. In the proposed DIOS-based FPJ, the interrelated refractive and reflective (R&R) coefficients of the adversarial IOS are randomly generated, acting like a "DISCO" that distributes wireless energy radiated by the base station. By introducing active channel aging (ACA) during channel coherence time, the DIOS-based FPJ can perform omnidirectional fully-passive jamming without neither jamming power nor channel knowledge of legitimate users (LUs). To characterize the impact of the DIOS-based PFJ, we derive the statistical characteristics of DIOS-jammed channels based on two widely-used IOS models, i.e., the constant-amplitude model and the variable-amplitude model. Consequently, the asymptotic analysis of the ergodic achievable sum rates under the DIOS-based omnidirectional fully-passive jamming is given based on the derived stochastic characteristics for both the two IOS models. Based on the derived analysis, the omnidirectional jamming impact of the proposed DIOS-based FPJ implemented by a constant-amplitude IOS does not depend on either the quantization number or the stochastic distribution of the DIOS coefficients, while the conclusion does not hold on when a variable-amplitude IOS is used. Numerical results based on one-bit quantization of the IOS phase shifts are provided to verify the effectiveness of the derived theoretical analysis. The proposed DIOS-based FPJ can not only launch omnidirectional fully-passive jamming, but also improve the jamming impact by about 55% at 10 dBm transmit power per LU.
Co-channel interference cancellation (CCI) is the process used to reduce interference from other signals using the same frequency channel, thereby enhancing the performance of wireless communication systems. An improvement to this approach is blind CCI, which reduces interference without relying on prior knowledge of the interfering signal characteristics. Recent work suggested using machine learning (ML) models for this purpose, but high-throughput ML solutions are still lacking, especially for edge devices with limited resources. This work explores the adaptation of U-Net Convolutional Neural Network models for high-throughput blind source separation. Our approach is established on architectural modifications, notably through quantization and the incorporation of depthwise separable convolution, to achieve a balance between computational efficiency and performance. Our results demonstrate that the proposed models achieve superior MSE scores when removing unknown interference sources from the signals while maintaining significantly lower computational complexity compared to baseline models. One of our proposed models is deeper and fully convolutional, while the other is shallower with a convolutional structure incorporating an LSTM. Depthwise separable convolution and quantization further reduce the memory footprint and computational demands, albeit with some performance trade-offs. Specifically, applying depthwise separable convolutions to the model with the LSTM results in only a 0.72% degradation in MSE score while reducing MACs by 58.66%. For the fully convolutional model, we observe a 0.63% improvement in MSE score with even 61.10% fewer MACs. Overall, our findings underscore the feasibility of using optimized machine-learning models for interference cancellation in devices with limited resources.