Recovering signals within limited dynamic range (DR) constraints remains a central challenge for analog-to-digital converters (ADCs). To prevent data loss, an ADCs DR typically must exceed that of the input signal. Modulo sampling has recently gained attention as a promising approach for addressing DR limitations across various signal classes. However, existing methods often rely on ideal ADCs capable of capturing the high frequencies introduced by the modulo operator, which is impractical in real-world hardware applications. This paper introduces an innovative hardware-based sampling approach that addresses these high-frequency components using an analog mixer followed by a Low-Pass Filter (LPF). This allows the use of realistic ADCs, which do not need to handle frequencies beyond the intended sampling rate. Our method eliminates the requirement for high-specification ADCs and demonstrates that the resulting samples are equivalent to those from an ideal high-spec ADC. Consequently, any existing modulo recovery algorithm can be applied effectively. We present a practical hardware prototype of this approach, validated through both simulations and hardware recovery experiments. Using a recovery method designed to handle quantization noise, we show that our approach effectively manages high-frequency artifacts, enabling reliable modulo recovery with realistic ADCs. These findings confirm that our hardware solution not only outperforms conventional methods in high-precision settings but also demonstrates significant real-world applicability.
In massive MIMO systems, fully digital precoding offers high performance but has significant implementation complexity and energy consumption, particularly at millimeter frequencies and beyond. Hybrid analog-digital architectures provide a practical alternative by reducing the number of radio frequency (RF) chains while retaining performance in spatially sparse multipath scenarios. However, most hybrid precoder designs assume ideal, infinite-resolution analog phase shifters, which are impractical in real-world scenarios. Another practical constraint is the limited fronthaul capacity between the baseband processor and array, implying that each entry of the digital precoder must be picked from a finite set of quantization labels. To minimize the sum rate degradation caused by quantized analog and digital precoders, we propose novel designs inspired by the sphere decoding (SD) algorithm. We demonstrate numerically that our proposed designs outperform traditional methods, ensuring minimal sum rate loss in hybrid precoding systems with low-resolution phase shifters and limited fronthaul capacity.
Large-scale language models (LLMs) have demonstrated outstanding performance in language processing tasks, yet their deployment is often hindered by high memory demands and computational complexity. Although low-bit quantization techniques, such as 4-bit quantization, present a potential solution, they frequently lead to significant accuracy degradation or require substantial effort for such aggressive quantization approaches. To overcome these challenges, we introduce QRazor, a reliable and effortless quantization scheme designed to enable 4-bit quantization for weights, activations, and KV cache in transformer-based LLMs. The scheme involves two main stages: quantization and compression. During the quantization stage, weights, activations, and KV cache values are quantized with wider 8 or 16-bit integers as a basis to achieve nearly identical accuracy to the original full-precision LLM models, using the absolute max scaling. Subsequently, all data are compressed to 4-bit using our proposed significant data razoring (SDR) technique, which retains only the four most salient bits while discarding the others. Furthermore, we present an integer-based arithmetic unit dedicated to QRazor, enabling direct low-precision arithmetic operations without decompressing the SDR data. Despite the reduced quantization effort, QRazor achieves LLM accuracies better or comparable to state-of-the-art 4-bit methods. By also validating the hardware efficiency, our decompression-free arithmetic unit achieves 61.2% and 57.8% reduction in area and power consumption, respectively.
State Space Models (SSMs), as key components of Mamaba, have gained increasing attention for vision models recently, thanks to their efficient long sequence modeling capability. Given the computational cost of deploying SSMs on resource-limited edge devices, Post-Training Quantization (PTQ) is a technique with the potential for efficient deployment of SSMs. In this work, we propose QMamba, one of the first PTQ frameworks to our knowledge, designed for vision SSMs based on the analysis of the activation distributions in SSMs. We reveal that the distribution of discrete parameters exhibits long-tailed skewness and the distribution of the hidden state sequence exhibits highly dynamic variations. Correspondingly, we design Long-tailed Skewness Quantization (LtSQ) to quantize discrete parameters and Temporal Group Quantization (TGQ) to quantize hidden states, which reduces the quantization errors. Extensive experiments demonstrate that QMamba outperforms advanced PTQ methods on vision models across multiple model sizes and architectures. Notably, QMamba surpasses existing methods by 21.0% on ImageNet classification with 4-bit activations.
Recently, foundational diffusion models have attracted considerable attention in image compression tasks, whereas their application to video compression remains largely unexplored. In this article, we introduce DiffVC, a diffusion-based perceptual neural video compression framework that effectively integrates foundational diffusion model with the video conditional coding paradigm. This framework uses temporal context from previously decoded frame and the reconstructed latent representation of the current frame to guide the diffusion model in generating high-quality results. To accelerate the iterative inference process of diffusion model, we propose the Temporal Diffusion Information Reuse (TDIR) strategy, which significantly enhances inference efficiency with minimal performance loss by reusing the diffusion information from previous frames. Additionally, to address the challenges posed by distortion differences across various bitrates, we propose the Quantization Parameter-based Prompting (QPP) mechanism, which utilizes quantization parameters as prompts fed into the foundational diffusion model to explicitly modulate intermediate features, thereby enabling a robust variable bitrate diffusion-based neural compression framework. Experimental results demonstrate that our proposed solution delivers excellent performance in both perception metrics and visual quality.
Data2vec is a self-supervised learning (SSL) approach that employs a teacher-student architecture for contextual representation learning via masked prediction, demonstrating remarkable performance in monolingual ASR. Previous studies have revealed that data2vec's shallow layers capture speaker and language information, middle layers encode phoneme and word features, while deep layers are responsible for reconstruction. Language and phoneme features are crucial for multilingual ASR. However, data2vec's masked representation generation relies on multi-layer averaging, inevitably coupling these features. To address this limitation, we propose a decoupling quantization based data2vec (DQ-Data2vec) for multilingual ASR, which includes a data2vec backbone and two improved online K-means quantizers. Our core idea is using the K-means quantizer with specified cluster numbers to decouple language and phoneme information for masked prediction. Specifically, in the language quantization, considering that the number of languages is significantly different from other irrelevant features (e.g., speakers), we assign the cluster number to match the number of languages, explicitly decoupling shallow layers' language-related information from irrelevant features. This strategy is also applied to decoupling middle layers' phoneme and word features. In a self-supervised scenario, experiments on the CommonVoice dataset demonstrate that DQ-Data2vec achieves a relative reduction of 9.51% in phoneme error rate (PER) and 11.58% in word error rate (WER) compared to data2vec and UniData2vec. Moreover, in a weakly-supervised scenario incorporating language labels and high-resource language text labels, the relative reduction is 18.09% and 1.55%, respectively.
Spiking neural networks are emerging as a promising energy-efficient alternative to traditional artificial neural networks due to their spike-driven paradigm. However, recent research in the SNN domain has mainly focused on enhancing accuracy by designing large-scale Transformer structures, which typically rely on substantial computational resources, limiting their deployment on resource-constrained devices. To overcome this challenge, we propose a quantized spike-driven Transformer baseline (QSD-Transformer), which achieves reduced resource demands by utilizing a low bit-width parameter. Regrettably, the QSD-Transformer often suffers from severe performance degradation. In this paper, we first conduct empirical analysis and find that the bimodal distribution of quantized spike-driven self-attention (Q-SDSA) leads to spike information distortion (SID) during quantization, causing significant performance degradation. To mitigate this issue, we take inspiration from mutual information entropy and propose a bi-level optimization strategy to rectify the information distribution in Q-SDSA. Specifically, at the lower level, we introduce an information-enhanced LIF to rectify the information distribution in Q-SDSA. At the upper level, we propose a fine-grained distillation scheme for the QSD-Transformer to align the distribution in Q-SDSA with that in the counterpart ANN. By integrating the bi-level optimization strategy, the QSD-Transformer can attain enhanced energy efficiency without sacrificing its high-performance advantage.For instance, when compared to the prior SNN benchmark on ImageNet, the QSD-Transformer achieves 80.3% top-1 accuracy, accompanied by significant reductions of 6.0$\times$ and 8.1$\times$ in power consumption and model size, respectively. Code is available at https://github.com/bollossom/QSD-Transformer.
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.
In this paper, we propose a novel optimizer state compression algorithm, namely $\pi$-Quant, which leverages the properties of irrational numbers (e.g., $\pi$) for memory-efficient training. The core idea is based on our mathematical findings, which show that a pair of parameters can be represented by a single rotation angle using the complex rotation scheme. Building on this insight, we map the parameters into a complex space and perform quantization using the corresponding rotation angles. To efficiently integrate it into optimization process, we develop an efficient system of geometric equations that computes the precise rotation angles with linear complexity. We evaluate $\pi$-Quant on a wide range of tasks. Our experiments show that it can reduce the bit-width of parameters to 3.32-bit, achieving a 75% reduction in parameter scale and a 40% decrease in GPU memory usage, all while maintaining full accuracy.
3D Gaussian Splatting (3DGS) has emerged as a promising representation for photorealistic rendering of 3D scenes. However, its high storage requirements pose significant challenges for practical applications. We observe that Gaussians exhibit distinct roles and characteristics that are analogous to traditional artistic techniques -- Like how artists first sketch outlines before filling in broader areas with color, some Gaussians capture high-frequency features like edges and contours; While other Gaussians represent broader, smoother regions, that are analogous to broader brush strokes that add volume and depth to a painting. Based on this observation, we propose a novel hybrid representation that categorizes Gaussians into (i) Sketch Gaussians, which define scene boundaries, and (ii) Patch Gaussians, which cover smooth regions. Sketch Gaussians are efficiently encoded using parametric models, leveraging their geometric coherence, while Patch Gaussians undergo optimized pruning, retraining, and vector quantization to maintain volumetric consistency and storage efficiency. Our comprehensive evaluation across diverse indoor and outdoor scenes demonstrates that this structure-aware approach achieves up to 32.62% improvement in PSNR, 19.12% in SSIM, and 45.41% in LPIPS at equivalent model sizes, and correspondingly, for an indoor scene, our model maintains the visual quality with 2.3% of the original model size.
Large Language Models (LLMs) face significant deployment challenges due to their substantial resource requirements. While low-bit quantized weights can reduce memory usage and improve inference efficiency, current hardware lacks native support for mixed-precision General Matrix Multiplication (mpGEMM), resulting in inefficient dequantization-based implementations. Moreover, uniform quantization methods often fail to capture weight distributions adequately, leading to performance degradation. We propose GANQ (GPU-Adaptive Non-Uniform Quantization), a layer-wise post-training non-uniform quantization framework optimized for hardware-efficient lookup table-based mpGEMM. GANQ achieves superior quantization performance by utilizing a training-free, GPU-adaptive optimization algorithm to efficiently reduce layer-wise quantization errors. Extensive experiments demonstrate GANQ's ability to reduce the perplexity gap from the FP16 baseline compared to state-of-the-art methods for both 3-bit and 4-bit quantization. Furthermore, when deployed on a single NVIDIA RTX 4090 GPU, GANQ's quantized models achieve up to 2.57$\times$ speedup over the baseline, advancing memory and inference efficiency in LLM deployment.
This paper introduces HEPPO, an FPGA-based accelerator designed to optimize the Generalized Advantage Estimation (GAE) stage in Proximal Policy Optimization (PPO). Unlike previous approaches that focused on trajectory collection and actor-critic updates, HEPPO addresses GAE's computational demands with a parallel, pipelined architecture implemented on a single System-on-Chip (SoC). This design allows for the adaptation of various hardware accelerators tailored for different PPO phases. A key innovation is our strategic standardization technique, which combines dynamic reward standardization and block standardization for values, followed by 8-bit uniform quantization. This method stabilizes learning, enhances performance, and manages memory bottlenecks, achieving a 4x reduction in memory usage and a 1.5x increase in cumulative rewards. We propose a solution on a single SoC device with programmable logic and embedded processors, delivering throughput orders of magnitude higher than traditional CPU-GPU systems. Our single-chip solution minimizes communication latency and throughput bottlenecks, significantly boosting PPO training efficiency. Experimental results show a 30% increase in PPO speed and a substantial reduction in memory access time, underscoring HEPPO's potential for broad applicability in hardware-efficient reinforcement learning algorithms.
Quantization for deep neural networks (DNNs) is the process of mapping the parameter values of DNNs from original data types to other data types of lower precision to reduce model sizes and make inference faster. Quantization often maps different original values to a single quantized value because the range of the original values is larger than the range of the quantized values. This leads to the degradation of the accuracy of the quantized DNNs. Outliers are a main cause of the degradation of quantization resolution because they enlarge the range of original values. To solve the problem, the percentile method is often used to clip outliers. However, clipping the outliers has another problem of removing the important and strong signals in the DNNs. This paper proposes SplitQuant to keep the outliers and improve the quantization resolution at the same time. SplitQuant narrows down the range of the original values and mitigates the effect of outliers by splitting each quantizable layer into three mathematically equivalent layers and applies different scaling factors. Especially, weights and biases are clustered into lower, middle and upper clusters for optimized split. By preprocessing DNNs with SplitQuant, quantization algorithms can achieve better results. SplitQuant was applied on two BERT-Tiny models and improved the accuracy of INT2 quantization by 3.3%p and 2.1%p, achieving accuracies comparable to those of the original FP32 models.
Federated learning (FL) has been recognized as a viable solution for local-privacy-aware collaborative model training in wireless edge networks, but its practical deployment is hindered by the high communication overhead caused by frequent and costly server-device synchronization. Notably, most existing communication-efficient FL algorithms fail to reduce the significant inter-device variance resulting from the prevalent issue of device heterogeneity. This variance severely decelerates algorithm convergence, increasing communication overhead and making it more challenging to achieve a well-performed model. In this paper, we propose a novel communication-efficient FL algorithm, named FedQVR, which relies on a sophisticated variance-reduced scheme to achieve heterogeneity-robustness in the presence of quantized transmission and heterogeneous local updates among active edge devices. Comprehensive theoretical analysis justifies that FedQVR is inherently resilient to device heterogeneity and has a comparable convergence rate even with a small number of quantization bits, yielding significant communication savings. Besides, considering non-ideal wireless channels, we propose FedQVR-E which enhances the convergence of FedQVR by performing joint allocation of bandwidth and quantization bits across devices under constrained transmission delays. Extensive experimental results are also presented to demonstrate the superior performance of the proposed algorithms over their counterparts in terms of both communication efficiency and application performance.
Approximate nearest neighbor search for vectors relies on indexes that are most often accessed from RAM. Therefore, storage is the factor limiting the size of the database that can be served from a machine. Lossy vector compression, i.e., embedding quantization, has been applied extensively to reduce the size of indexes. However, for inverted file and graph-based indices, auxiliary data such as vector ids and links (edges) can represent most of the storage cost. We introduce and evaluate lossless compression schemes for these cases. These approaches are based on asymmetric numeral systems or wavelet trees that exploit the fact that the ordering of ids is irrelevant within the data structures. In some settings, we are able to compress the vector ids by a factor 7, with no impact on accuracy or search runtime. On billion-scale datasets, this results in a reduction of 30% of the index size. Furthermore, we show that for some datasets, these methods can also compress the quantized vector codes losslessly, by exploiting sub-optimalities in the original quantization algorithm. The source code for our approach available at https://github.com/facebookresearch/vector_db_id_compression.
One of the most fundamental, and yet relatively less explored, goals in transfer learning is the efficient means of selecting top candidates from a large number of previously trained models (optimized for various "source" tasks) that would perform the best for a new "target" task with a limited amount of data. In this paper, we undertake this goal by developing a novel task-similarity metric (BeST) and an associated method that consistently performs well in identifying the most transferrable source(s) for a given task. In particular, our design employs an innovative quantization-level optimization procedure in the context of classification tasks that yields a measure of similarity between a source model and the given target data. The procedure uses a concept similar to early stopping (usually implemented to train deep neural networks (DNNs) to ensure generalization) to derive a function that approximates the transfer learning mapping without training. The advantage of our metric is that it can be quickly computed to identify the top candidate(s) for a given target task before a computationally intensive transfer operation (typically using DNNs) can be implemented between the selected source and the target task. As such, our metric can provide significant computational savings for transfer learning from a selection of a large number of possible source models. Through extensive experimental evaluations, we establish that our metric performs well over different datasets and varying numbers of data samples.
Delivering an immersive experience to virtual reality (VR) users through wireless connectivity offers the freedom to engage from anywhere at any time. Nevertheless, it is challenging to ensure seamless wireless connectivity that delivers real-time and high-quality videos to the VR users. This paper proposes a field of view (FoV) aware caching for mobile edge computing (MEC)-enabled wireless VR network. In particular, the FoV of each VR user is cached/prefetched at the base stations (BSs) based on the caching strategies tailored to each BS. Specifically, decentralized and personalized federated learning (DP-FL) based caching strategies with guarantees are presented. Considering VR systems composed of multiple VR devices and BSs, a DP-FL caching algorithm is implemented at each BS to personalize content delivery for VR users. The utilized DP-FL algorithm guarantees a probably approximately correct (PAC) bound on the conditional average cache hit. Further, to reduce the cost of communicating gradients, one-bit quantization of the stochastic gradient descent (OBSGD) is proposed, and a convergence guarantee of $\mathcal{O}(1/\sqrt{T})$ is obtained for the proposed algorithm, where $T$ is the number of iterations. Additionally, to better account for the wireless channel dynamics, the FoVs are grouped into multicast or unicast groups based on the number of requesting VR users. The performance of the proposed DP-FL algorithm is validated through realistic VR head-tracking dataset, and the proposed algorithm is shown to have better performance in terms of average delay and cache hit as compared to baseline algorithms.
This paper presents LiFT, a lightweight, fully quantized 3D object detection algorithm for LiDAR data, optimized for real-time inference on FPGA platforms. Through an in-depth analysis of FPGA-specific limitations, we identify a set of FPGA-induced constraints that shape the algorithm's design. These include a computational complexity limit of 30 GMACs (billion multiply-accumulate operations), INT8 quantization for weights and activations, 2D cell-based processing instead of 3D voxels, and minimal use of skip connections. To meet these constraints while maximizing performance, LiFT combines novel mechanisms with state-of-the-art techniques such as reparameterizable convolutions and fully sparse architecture. Key innovations include the Dual-bound Pillar Feature Net, which boosts performance without increasing complexity, and an efficient scheme for INT8 quantization of input features. With a computational cost of just 20.73 GMACs, LiFT stands out as one of the few algorithms targeting minimal-complexity 3D object detection. Among comparable methods, LiFT ranks first, achieving an mAP of 51.84% and an NDS of 61.01% on the challenging NuScenes validation dataset. The code will be available at https://github.com/vision-agh/lift.
3D Gaussian Splatting (3DGS) has emerged as a promising framework for novel view synthesis, boasting rapid rendering speed with high fidelity. However, the substantial Gaussians and their associated attributes necessitate effective compression techniques. Nevertheless, the sparse and unorganized nature of the point cloud of Gaussians (or anchors in our paper) presents challenges for compression. To achieve a compact size, we propose HAC++, which leverages the relationships between unorganized anchors and a structured hash grid, utilizing their mutual information for context modeling. Additionally, HAC++ captures intra-anchor contextual relationships to further enhance compression performance. To facilitate entropy coding, we utilize Gaussian distributions to precisely estimate the probability of each quantized attribute, where an adaptive quantization module is proposed to enable high-precision quantization of these attributes for improved fidelity restoration. Moreover, we incorporate an adaptive masking strategy to eliminate invalid Gaussians and anchors. Overall, HAC++ achieves a remarkable size reduction of over 100X compared to vanilla 3DGS when averaged on all datasets, while simultaneously improving fidelity. It also delivers more than 20X size reduction compared to Scaffold-GS. Our code is available at https://github.com/YihangChen-ee/HAC-plus.
Video encoders optimize compression for human perception by minimizing reconstruction error under bit-rate constraints. In many modern applications such as autonomous driving, an overwhelming majority of videos serve as input for AI systems performing tasks like object recognition or segmentation, rather than being watched by humans. It is therefore useful to optimize the encoder for a downstream task instead of for perceptual image quality. However, a major challenge is how to combine such downstream optimization with existing standard video encoders, which are highly efficient and popular. Here, we address this challenge by controlling the Quantization Parameters (QPs) at the macro-block level to optimize the downstream task. This granular control allows us to prioritize encoding for task-relevant regions within each frame. We formulate this optimization problem as a Reinforcement Learning (RL) task, where the agent learns to balance long-term implications of choosing QPs on both task performance and bit-rate constraints. Notably, our policy does not require the downstream task as an input during inference, making it suitable for streaming applications and edge devices such as vehicles. We demonstrate significant improvements in two tasks, car detection, and ROI (saliency) encoding. Our approach improves task performance for a given bit rate compared to traditional task agnostic encoding methods, paving the way for more efficient task-aware video compression.