Nowadays, data is represented by vectors. Retrieving those vectors, among millions and billions, that are similar to a given query is a ubiquitous problem, known as similarity search, of relevance for a wide range of applications. Graph-based indices are currently the best performing techniques for billion-scale similarity search. However, their random-access memory pattern presents challenges to realize their full potential. In this work, we present new techniques and systems for creating faster and smaller graph-based indices. To this end, we introduce a novel vector compression method, Locally-adaptive Vector Quantization (LVQ), that uses per-vector scaling and scalar quantization to improve search performance with fast similarity computations and a reduced effective bandwidth, while decreasing memory footprint and barely impacting accuracy. LVQ, when combined with a new high-performance computing system for graph-based similarity search, establishes the new state of the art in terms of performance and memory footprint. For billions of vectors, LVQ outcompetes the second-best alternatives: (1) in the low-memory regime, by up to 20.7x in throughput with up to a 3x memory footprint reduction, and (2) in the high-throughput regime by 5.8x with 1.4x less memory.
Transformers' compute-intensive operations pose enormous challenges for their deployment in resource-constrained EdgeAI / tinyML devices. As an established neural network compression technique, quantization reduces the hardware computational and memory resources. In particular, fixed-point quantization is desirable to ease the computations using lightweight blocks, like adders and multipliers, of the underlying hardware. However, deploying fully-quantized Transformers on existing general-purpose hardware, generic AI accelerators, or specialized architectures for Transformers with floating-point units might be infeasible and/or inefficient. Towards this, we propose SwiftTron, an efficient specialized hardware accelerator designed for Quantized Transformers. SwiftTron supports the execution of different types of Transformers' operations (like Attention, Softmax, GELU, and Layer Normalization) and accounts for diverse scaling factors to perform correct computations. We synthesize the complete SwiftTron architecture in a $65$ nm CMOS technology with the ASIC design flow. Our Accelerator executes the RoBERTa-base model in 1.83 ns, while consuming 33.64 mW power, and occupying an area of 273 mm^2. To ease the reproducibility, the RTL of our SwiftTron architecture is released at https://github.com/albertomarchisio/SwiftTron.
Recently, self-supervised learning (SSL) has achieved tremendous success in learning image representation. Despite the empirical success, most self-supervised learning methods are rather "inefficient" learners, typically taking hundreds of training epochs to fully converge. In this work, we show that the key towards efficient self-supervised learning is to increase the number of crops from each image instance. Leveraging one of the state-of-the-art SSL method, we introduce a simplistic form of self-supervised learning method called Extreme-Multi-Patch Self-Supervised-Learning (EMP-SSL) that does not rely on many heuristic techniques for SSL such as weight sharing between the branches, feature-wise normalization, output quantization, and stop gradient, etc, and reduces the training epochs by two orders of magnitude. We show that the proposed method is able to converge to 85.1% on CIFAR-10, 58.5% on CIFAR-100, 38.1% on Tiny ImageNet and 58.5% on ImageNet-100 in just one epoch. Furthermore, the proposed method achieves 91.5% on CIFAR-10, 70.1% on CIFAR-100, 51.5% on Tiny ImageNet and 78.9% on ImageNet-100 with linear probing in less than ten training epochs. In addition, we show that EMP-SSL shows significantly better transferability to out-of-domain datasets compared to baseline SSL methods. We will release the code in https://github.com/tsb0601/EMP-SSL.
With the advent of general-purpose speech representations from large-scale self-supervised models, applying a single model to multiple downstream tasks is becoming a de-facto approach. However, the pooling problem remains; the length of speech representations is inherently variable. The naive average pooling is often used, even though it ignores the characteristics of speech, such as differently lengthed phonemes. Hence, we design a novel pooling method to squash acoustically similar representations via vector quantization, which does not require additional training, unlike attention-based pooling. Further, we evaluate various unsupervised pooling methods on various self-supervised models. We gather diverse methods scattered around speech and text to evaluate on various tasks: keyword spotting, speaker identification, intent classification, and emotion recognition. Finally, we quantitatively and qualitatively analyze our method, comparing it with supervised pooling methods.
Learning-based video compression has been extensively studied over the past years, but it still has limitations in adapting to various motion patterns and entropy models. In this paper, we propose multi-mode video compression (MMVC), a block wise mode ensemble deep video compression framework that selects the optimal mode for feature domain prediction adapting to different motion patterns. Proposed multi-modes include ConvLSTM-based feature domain prediction, optical flow conditioned feature domain prediction, and feature propagation to address a wide range of cases from static scenes without apparent motions to dynamic scenes with a moving camera. We partition the feature space into blocks for temporal prediction in spatial block-based representations. For entropy coding, we consider both dense and sparse post-quantization residual blocks, and apply optional run-length coding to sparse residuals to improve the compression rate. In this sense, our method uses a dual-mode entropy coding scheme guided by a binary density map, which offers significant rate reduction surpassing the extra cost of transmitting the binary selection map. We validate our scheme with some of the most popular benchmarking datasets. Compared with state-of-the-art video compression schemes and standard codecs, our method yields better or competitive results measured with PSNR and MS-SSIM.
The recent advances in machine learning and deep neural networks have made them attractive candidates for wireless communications functions such as channel estimation, decoding, and downlink channel state information (CSI) compression. However, most of these neural networks are large and inefficient making it a barrier for deployment in practical wireless systems that require low-latency and low memory footprints for individual network functions. To mitigate these limitations, we propose accelerated and compressed efficient neural networks for massive MIMO CSI feedback. Specifically, we have thoroughly investigated the adoption of network pruning, post-training dynamic range quantization, and weight clustering to optimize CSI feedback compression for massive MIMO systems. Furthermore, we have deployed the proposed model compression techniques on commodity hardware and demonstrated that in order to achieve inference gains, specialized libraries that accelerate computations for sparse neural networks are required. Our findings indicate that there is remarkable value in applying these model compression techniques and the proposed joint pruning and quantization approach reduced model size by 86.5% and inference time by 76.2% with minimal impact to model accuracy. These compression methods are crucial to pave the way for practical adoption and deployments of deep learning-based techniques in commercial wireless systems.
Large-scale language models (LLMs) have demonstrated impressive performance, but their deployment presents challenges due to their significant memory usage. This issue can be alleviated through quantization. In this paper, we identify that the challenge in quantizing activations in LLMs arises from varying ranges across channels, rather than solely the presence of outliers. To address this challenge, we introduce a quantization method called RPTQ, which utilizes a reorder-based approach. By rearranging the channels and quantizing them in clusters, RPTQ effectively mitigates the impact of range differences between channels. To minimize the overhead of the reorder operation, we fuse it into the layer norm operation and weights in linear layers. In our experiments, RPTQ achieved a significant breakthrough by utilizing 3-bit activation in LLMs for the first time, resulting in a substantial reduction in memory usage. For instance, quantizing OPT-175b can lead to a memory consumption reduction of up to 80%.
The recent detection transformer (DETR) has advanced object detection, but its application on resource-constrained devices requires massive computation and memory resources. Quantization stands out as a solution by representing the network in low-bit parameters and operations. However, there is a significant performance drop when performing low-bit quantized DETR (Q-DETR) with existing quantization methods. We find that the bottlenecks of Q-DETR come from the query information distortion through our empirical analyses. This paper addresses this problem based on a distribution rectification distillation (DRD). We formulate our DRD as a bi-level optimization problem, which can be derived by generalizing the information bottleneck (IB) principle to the learning of Q-DETR. At the inner level, we conduct a distribution alignment for the queries to maximize the self-information entropy. At the upper level, we introduce a new foreground-aware query matching scheme to effectively transfer the teacher information to distillation-desired features to minimize the conditional information entropy. Extensive experimental results show that our method performs much better than prior arts. For example, the 4-bit Q-DETR can theoretically accelerate DETR with ResNet-50 backbone by 6.6x and achieve 39.4% AP, with only 2.6% performance gaps than its real-valued counterpart on the COCO dataset.
Contemporary image rescaling aims at embedding a high-resolution (HR) image into a low-resolution (LR) thumbnail image that contains embedded information for HR image reconstruction. Unlike traditional image super-resolution, this enables high-fidelity HR image restoration faithful to the original one, given the embedded information in the LR thumbnail. However, state-of-the-art image rescaling methods do not optimize the LR image file size for efficient sharing and fall short of real-time performance for ultra-high-resolution (e.g., 6K) image reconstruction. To address these two challenges, we propose a novel framework (HyperThumbnail) for real-time 6K rate-distortion-aware image rescaling. Our framework first embeds an HR image into a JPEG LR thumbnail by an encoder with our proposed quantization prediction module, which minimizes the file size of the embedding LR JPEG thumbnail while maximizing HR reconstruction quality. Then, an efficient frequency-aware decoder reconstructs a high-fidelity HR image from the LR one in real time. Extensive experiments demonstrate that our framework outperforms previous image rescaling baselines in rate-distortion performance and can perform 6K image reconstruction in real time.
Learning rich data representations from unlabeled data is a key challenge towards applying deep learning algorithms in downstream tasks. Several variants of variational autoencoders (VAEs) have been proposed to learn compact data representations by encoding high-dimensional data in a lower dimensional space. Two main classes of VAEs methods may be distinguished depending on the characteristics of the meta-priors that are enforced in the representation learning step. The first class of methods derives a continuous encoding by assuming a static prior distribution in the latent space. The second class of methods learns instead a discrete latent representation using vector quantization (VQ) along with a codebook. However, both classes of methods suffer from certain challenges, which may lead to suboptimal image reconstruction results. The first class suffers from posterior collapse, whereas the second class suffers from codebook collapse. To address these challenges, we introduce a new VAE variant, termed sparse coding-based VAE with learned ISTA (SC-VAE), which integrates sparse coding within variational autoencoder framework. The proposed method learns sparse data representations that consist of a linear combination of a small number of predetermined orthogonal atoms. The sparse coding problem is solved using a learnable version of the iterative shrinkage thresholding algorithm (ISTA). Experiments on two image datasets demonstrate that our model achieves improved image reconstruction results compared to state-of-the-art methods. Moreover, we demonstrate that the use of learned sparse code vectors allows us to perform downstream tasks like image generation and unsupervised image segmentation through clustering image patches.
Object discovery -- separating objects from the background without manual labels -- is a fundamental open challenge in computer vision. Previous methods struggle to go beyond clustering of low-level cues, whether handcrafted (e.g., color, texture) or learned (e.g., from auto-encoders). In this work, we augment the auto-encoder representation learning framework with two key components: motion-guidance and mid-level feature tokenization. Although both have been separately investigated, we introduce a new transformer decoder showing that their benefits can compound thanks to motion-guided vector quantization. We show that our architecture effectively leverages the synergy between motion and tokenization, improving upon the state of the art on both synthetic and real datasets. Our approach enables the emergence of interpretable object-specific mid-level features, demonstrating the benefits of motion-guidance (no labeling) and quantization (interpretability, memory efficiency).
Zero-shot quantization (ZSQ) is promising for compressing and accelerating deep neural networks when the data for training full-precision models are inaccessible. In ZSQ, network quantization is performed using synthetic samples, thus, the performance of quantized models depends heavily on the quality of synthetic samples. Nonetheless, we find that the synthetic samples constructed in existing ZSQ methods can be easily fitted by models. Accordingly, quantized models obtained by these methods suffer from significant performance degradation on hard samples. To address this issue, we propose HArd sample Synthesizing and Training (HAST). Specifically, HAST pays more attention to hard samples when synthesizing samples and makes synthetic samples hard to fit when training quantized models. HAST aligns features extracted by full-precision and quantized models to ensure the similarity between features extracted by these two models. Extensive experiments show that HAST significantly outperforms existing ZSQ methods, achieving performance comparable to models that are quantized with real data.
Mixed-signal neuromorphic processors provide extremely low-power operation for edge inference workloads, taking advantage of sparse asynchronous computation within Spiking Neural Networks (SNNs). However, deploying robust applications to these devices is complicated by limited controllability over analog hardware parameters, as well as unintended parameter and dynamical variations of analog circuits due to fabrication non-idealities. Here we demonstrate a novel methodology for ofDine training and deployment of spiking neural networks (SNNs) to the mixed-signal neuromorphic processor DYNAP-SE2. The methodology utilizes gradient-based training using a differentiable simulation of the mixed-signal device, coupled with an unsupervised weight quantization method to optimize the network's parameters. Parameter noise injection during training provides robustness to the effects of quantization and device mismatch, making the method a promising candidate for real-world applications under hardware constraints and non-idealities. This work extends Rockpool, an open-source deep-learning library for SNNs, with support for accurate simulation of mixed-signal SNN dynamics. Our approach simplifies the development and deployment process for the neuromorphic community, making mixed-signal neuromorphic processors more accessible to researchers and developers.
In this paper, we introduce a new approach, called Posthoc Interpretation via Quantization (PIQ), for interpreting decisions made by trained classifiers. Our method utilizes vector quantization to transform the representations of a classifier into a discrete, class-specific latent space. The class-specific codebooks act as a bottleneck that forces the interpreter to focus on the parts of the input data deemed relevant by the classifier for making a prediction. Our model formulation also enables learning concepts by incorporating the supervision of pretrained annotation models such as state-of-the-art image segmentation models. We evaluated our method through quantitative and qualitative studies involving black-and-white images, color images, and audio. As a result of these studies we found that PIQ generates interpretations that are more easily understood by participants to our user studies when compared to several other interpretation methods in the literature.
Recently, vision transformers (ViTs) have superseded convolutional neural networks in numerous applications, including classification, detection, and segmentation. However, the high computational requirements of ViTs hinder their widespread implementation. To address this issue, researchers have proposed efficient hybrid transformer architectures that combine convolutional and transformer layers with optimized attention computation of linear complexity. Additionally, post-training quantization has been proposed as a means of mitigating computational demands. For mobile devices, achieving optimal acceleration for ViTs necessitates the strategic integration of quantization techniques and efficient hybrid transformer structures. However, no prior investigation has applied quantization to efficient hybrid transformers. In this paper, we discover that applying existing post-training quantization (PTQ) methods for ViTs to efficient hybrid transformers leads to a drastic accuracy drop, attributed to the four following challenges: (i) highly dynamic ranges, (ii) zero-point overflow, (iii) diverse normalization, and (iv) limited model parameters ($<$5M). To overcome these challenges, we propose a new post-training quantization method, which is the first to quantize efficient hybrid ViTs (MobileViTv1, MobileViTv2, Mobile-Former, EfficientFormerV1, EfficientFormerV2). We achieve a significant improvement of 17.73% for 8-bit and 29.75% for 6-bit on average, respectively, compared with existing PTQ methods (EasyQuant, FQ-ViT, PTQ4ViT, and RepQ-ViT)}. We plan to release our code at https://gitlab.com/ones-ai/q-hyvit.
Post-training quantization (PTQ) is widely regarded as one of the most efficient compression methods practically, benefitting from its data privacy and low computation costs. We argue that an overlooked problem of oscillation is in the PTQ methods. In this paper, we take the initiative to explore and present a theoretical proof to explain why such a problem is essential in PTQ. And then, we try to solve this problem by introducing a principled and generalized framework theoretically. In particular, we first formulate the oscillation in PTQ and prove the problem is caused by the difference in module capacity. To this end, we define the module capacity (ModCap) under data-dependent and data-free scenarios, where the differentials between adjacent modules are used to measure the degree of oscillation. The problem is then solved by selecting top-k differentials, in which the corresponding modules are jointly optimized and quantized. Extensive experiments demonstrate that our method successfully reduces the performance drop and is generalized to different neural networks and PTQ methods. For example, with 2/4 bit ResNet-50 quantization, our method surpasses the previous state-of-the-art method by 1.9%. It becomes more significant on small model quantization, e.g. surpasses BRECQ method by 6.61% on MobileNetV2*0.5.
Post-training quantization (PTQ) has emerged as a promising technique for mitigating memory consumption and computational costs in large language models (LLMs). However, a systematic examination of various quantization schemes, model families, and quantization bit precision has been absent from the literature. In this paper, we conduct a comprehensive analysis of these factors by investigating the effects of PTQ on weight-only, activation-only, and weight-and-activation quantization using diverse methods such as round-to-nearest (RTN), GPTQ, ZeroQuant, and their variants. We apply these methods to two distinct model families with parameters ranging from 125M to 176B. Our contributions include: (1) a sensitivity analysis revealing that activation quantization is generally more susceptible to weight quantization, with smaller models often outperforming larger models in terms of activation quantization; (2) an evaluation and comparison of existing PTQ methods to optimize model size reduction while minimizing the impact on accuracy, revealing that none of the current methods can achieve the original model quality for quantization with either INT4-weight or INT4-weight-and-INT8-activation; (3) based on these insights, we propose an optimized method called Low-Rank Compensation (LoRC), which employs low-rank matrices to enhance model quality recovery with a minimal increase in model size.
The Victoria Amazonica plant, often known as the Giant Water Lily, has the largest floating spherical leaf in the world, with a maximum leaf diameter of 3 meters. It spreads its leaves by the force of its spines and creates a large shadow underneath, killing any plants that require sunlight. These water tyrants use their formidable spines to compel each other to the surface and increase their strength to grab more space from the surface. As they spread throughout the pond or basin, with the earliest-growing leaves having more room to grow, each leaf gains a unique size. Its flowers are transsexual and when they bloom, Cyclocephala beetles are responsible for the pollination process, being attracted to the scent of the female flower. After entering the flower, the beetle becomes covered with pollen and transfers it to another flower for fertilization. After the beetle leaves, the flower turns into a male and changes color from white to pink. The male flower dies and sinks into the water, releasing its seed to help create a new generation. In this paper, the mathematical life cycle of this magnificent plant is introduced, and each leaf and blossom are treated as a single entity. The proposed bio-inspired algorithm is tested with 24 benchmark optimization test functions, such as Ackley, and compared to ten other famous algorithms, including the Genetic Algorithm. The proposed algorithm is tested on 10 optimization problems: Minimum Spanning Tree, Hub Location Allocation, Quadratic Assignment, Clustering, Feature Selection, Regression, Economic Dispatching, Parallel Machine Scheduling, Color Quantization, and Image Segmentation and compared to traditional and bio-inspired algorithms. Overall, the performance of the algorithm in all tasks is satisfactory.
Data-free quantization (DFQ) recovers the performance of quantized network (Q) without the original data, but generates the fake sample via a generator (G) by learning from full-precision network (P), which, however, is totally independent of Q, overlooking the adaptability of the knowledge from generated samples, i.e., informative or not to the learning process of Q, resulting into the overflow of generalization error. Building on this, several critical questions -- how to measure the sample adaptability to Q under varied bit-width scenarios? whether the largest adaptability is the best? how to generate the samples with adaptive adaptability to improve Q's generalization? To answer the above questions, in this paper, we propose an Adaptive Data-Free Quantization (AdaDFQ) method, which revisits DFQ from a zero-sum game perspective upon the sample adaptability between two players -- a generator and a quantized network. Following this viewpoint, we further define the disagreement and agreement samples to form two boundaries, where the margin is optimized to adaptively regulate the adaptability of generated samples to Q, so as to address the over-and-under fitting issues. Our AdaDFQ reveals: 1) the largest adaptability is NOT the best for sample generation to benefit Q's generalization; 2) the knowledge of the generated sample should not be informative to Q only, but also related to the category and distribution information of the training data for P. The theoretical and empirical analysis validate the advantages of AdaDFQ over the state-of-the-arts. Our code is available at https://github.com/hfutqian/AdaDFQ.
Edge applications, such as collaborative robotics and spacecraft rendezvous, demand efficient 6D object pose estimation on resource-constrained embedded platforms. Existing 6D pose estimation networks are often too large for such deployments, necessitating compression while maintaining reliable performance. To address this challenge, we introduce Modular Quantization-Aware Training (MQAT), an adaptive and mixed-precision quantization-aware training strategy that exploits the modular structure of modern 6D pose estimation architectures. MQAT guides a systematic gradated modular quantization sequence and determines module-specific bit precisions, leading to quantized models that outperform those produced by state-of-the-art uniform and mixed-precision quantization techniques. Our experiments showcase the generality of MQAT across datasets, architectures, and quantization algorithms. Remarkably, MQAT-trained quantized models achieve a significant accuracy boost (>7%) over the baseline full-precision network while reducing model size by a factor of 4x or more. Our project website is at: https://saqibjaved1.github.io/MQAT_/