With the development of hardware for machine learning, newer models often come at the cost of both increased sizes and computational complexity. In effort to improve the efficiency for these models, we apply and investigate recent quantization techniques on speech representation learning models. The quantization techniques were evaluated on the SUPERB benchmark. On the ASR task, with aggressive quantization to 1 bit, we achieved 86.32% storage reduction (184.42 -> 25.23), 88% estimated runtime reduction (1.00 -> 0.12) with increased word error rate (7.06 -> 15.96). In comparison with DistillHuBERT which also aims for model compression, the 2-bit configuration yielded slightly smaller storage (35.84 vs. 46.98), better word error rate (12.68 vs. 13.37) and more efficient estimated runtime (0.15 vs. 0.73).
We address the challenge of producing trustworthy and accurate compact models for edge devices. While Knowledge Distillation (KD) has improved model compression in terms of achieving high accuracy performance, calibration of these compact models has been overlooked. We introduce BD-KD (Balanced Divergence Knowledge Distillation), a framework for logit-based online KD. BD-KD enhances both accuracy and model calibration simultaneously, eliminating the need for post-hoc recalibration techniques, which add computational overhead to the overall training pipeline and degrade performance. Our method encourages student-centered training by adjusting the conventional online distillation loss on both the student and teacher losses, employing sample-wise weighting of forward and reverse Kullback-Leibler divergence. This strategy balances student network confidence and boosts performance. Experiments across CIFAR10, CIFAR100, TinyImageNet, and ImageNet datasets, and various architectures demonstrate improved calibration and accuracy compared to recent online KD methods.
Neural fields, also known as coordinate-based or implicit neural representations, have shown a remarkable capability of representing, generating, and manipulating various forms of signals. For video representations, however, mapping pixel-wise coordinates to RGB colors has shown relatively low compression performance and slow convergence and inference speed. Frame-wise video representation, which maps a temporal coordinate to its entire frame, has recently emerged as an alternative method to represent videos, improving compression rates and encoding speed. While promising, it has still failed to reach the performance of state-of-the-art video compression algorithms. In this work, we propose FFNeRV, a novel method for incorporating flow information into frame-wise representations to exploit the temporal redundancy across the frames in videos inspired by the standard video codecs. Furthermore, we introduce a fully convolutional architecture, enabled by one-dimensional temporal grids, improving the continuity of spatial features. Experimental results show that FFNeRV yields the best performance for video compression and frame interpolation among the methods using frame-wise representations or neural fields. To reduce the model size even further, we devise a more compact convolutional architecture using the group and pointwise convolutions. With model compression techniques, including quantization-aware training and entropy coding, FFNeRV outperforms widely-used standard video codecs (H.264 and HEVC) and performs on par with state-of-the-art video compression algorithms.
Convolution neural networks (CNNs) have achieved remarkable success, but typically accompany high computation cost and numerous redundant weight parameters. To reduce the FLOPs, structure pruning is a popular approach to remove the entire hidden structures via introducing coarse-grained sparsity. Meanwhile, plentiful pruning works leverage fine-grained sparsity instead (sparsity are randomly distributed), whereas their sparse models lack special designed computing library for potential speedup. In this technical report, we study and present an efficient convolution neural network inference system to accelerate its forward pass by utilizing the fine-grained sparsity of compressed CNNs. Our developed FSCNN is established based on a set of specialized designed sparse data structures, operators and associated algorithms. Experimentally, we validate that FSCNN outperforms standard deep learning library PyTorch on popular CNN architectures such as VGG16 if sufficiently high sparsity exhibits. However, due to the contiguity issue of sparse operators, FSCNN is typically not comparable with highly optimized dense operator. Therefore, coarse-grained (structured) sparsity is our recommendation for generic model compression.
Yes. In this paper, we investigate strong lottery tickets in generative models, the subnetworks that achieve good generative performance without any weight update. Neural network pruning is considered the main cornerstone of model compression for reducing the costs of computation and memory. Unfortunately, pruning a generative model has not been extensively explored, and all existing pruning algorithms suffer from excessive weight-training costs, performance degradation, limited generalizability, or complicated training. To address these problems, we propose to find a strong lottery ticket via moment-matching scores. Our experimental results show that the discovered subnetwork can perform similarly or better than the trained dense model even when only 10% of the weights remain. To the best of our knowledge, we are the first to show the existence of strong lottery tickets in generative models and provide an algorithm to find it stably. Our code and supplementary materials are publicly available.
Post-training quantization (PTQ), which only requires a tiny dataset for calibration without end-to-end retraining, is a light and practical model compression technique. Recently, several PTQ schemes for vision transformers (ViTs) have been presented; unfortunately, they typically suffer from non-trivial accuracy degradation, especially in low-bit cases. In this paper, we propose RepQ-ViT, a novel PTQ framework for ViTs based on quantization scale reparameterization, to address the above issues. RepQ-ViT decouples the quantization and inference processes, where the former employs complex quantizers and the latter employs scale-reparameterized simplified quantizers. This ensures both accurate quantization and efficient inference, which distinguishes it from existing approaches that sacrifice quantization performance to meet the target hardware. More specifically, we focus on two components with extreme distributions: post-LayerNorm activations with severe inter-channel variation and post-Softmax activations with power-law features, and initially apply channel-wise quantization and log$\sqrt{2}$ quantization, respectively. Then, we reparameterize the scales to hardware-friendly layer-wise quantization and log2 quantization for inference, with only slight accuracy or computational costs. Extensive experiments are conducted on multiple vision tasks with different model variants, proving that RepQ-ViT, without hyperparameters and expensive reconstruction procedures, can outperform existing strong baselines and encouragingly improve the accuracy of 4-bit PTQ of ViTs to a usable level. Code is available at https://github.com/zkkli/RepQ-ViT.
Knowledge distillation (KD) has been widely used for model compression and knowledge transfer. Typically, a big teacher model trained on sufficient data transfers knowledge to a small student model. However, despite the success of KD, little effort has been made to study whether KD leaks the training data of the teacher model. In this paper, we experimentally reveal that KD suffers from the risk of privacy leakage. To alleviate this issue, we propose a novel knowledge distillation method, swing distillation, which can effectively protect the private information of the teacher model from flowing to the student model. In our framework, the temperature coefficient is dynamically and adaptively adjusted according to the degree of private information contained in the data, rather than a predefined constant hyperparameter. It assigns different temperatures to tokens according to the likelihood that a token in a position contains private information. In addition, we inject noise into soft targets provided to the student model, in order to avoid unshielded knowledge transfer. Experiments on multiple datasets and tasks demonstrate that the proposed swing distillation can significantly reduce (by over 80% in terms of canary exposure) the risk of privacy leakage in comparison to KD with competitive or better performance. Furthermore, swing distillation is robust against the increasing privacy budget.
Quantization has become a predominant approach for model compression, enabling deployment of large models trained on GPUs onto smaller form-factor devices for inference. Quantization-aware training (QAT) optimizes model parameters with respect to the end task while simulating quantization error, leading to better performance than post-training quantization. Approximation of gradients through the non-differentiable quantization operator is typically achieved using the straight-through estimator (STE) or additive noise. However, STE-based methods suffer from instability due to biased gradients, whereas existing noise-based methods cannot reduce the resulting variance. In this work, we incorporate exponentially decaying quantization-error-aware noise together with a learnable scale of task loss gradient to approximate the effect of a quantization operator. We show this method combines gradient scale and quantization noise in a better optimized way, providing finer-grained estimation of gradients at each weight and activation layer's quantizer bin size. Our controlled noise also contains an implicit curvature term that could encourage flatter minima, which we show is indeed the case in our experiments. Experiments training ResNet architectures on the CIFAR-10, CIFAR-100 and ImageNet benchmarks show that our method obtains state-of-the-art top-1 classification accuracy for uniform (non mixed-precision) quantization, out-performing previous methods by 0.5-1.2% absolute.
Learning style refers to a type of training mechanism adopted by an individual to gain new knowledge. As suggested by the VARK model, humans have different learning preferences, like Visual (V), Auditory (A), Read/Write (R), and Kinesthetic (K), for acquiring and effectively processing information. Our work endeavors to leverage this concept of knowledge diversification to improve the performance of model compression techniques like Knowledge Distillation (KD) and Mutual Learning (ML). Consequently, we use a single-teacher and two-student network in a unified framework that not only allows for the transfer of knowledge from teacher to students (KD) but also encourages collaborative learning between students (ML). Unlike the conventional approach, where the teacher shares the same knowledge in the form of predictions or feature representations with the student network, our proposed approach employs a more diversified strategy by training one student with predictions and the other with feature maps from the teacher. We further extend this knowledge diversification by facilitating the exchange of predictions and feature maps between the two student networks, enriching their learning experiences. We have conducted comprehensive experiments with three benchmark datasets for both classification and segmentation tasks using two different network architecture combinations. These experimental results demonstrate that knowledge diversification in a combined KD and ML framework outperforms conventional KD or ML techniques (with similar network configuration) that only use predictions with an average improvement of 2%. Furthermore, consistent improvement in performance across different tasks, with various network architectures, and over state-of-the-art techniques establishes the robustness and generalizability of the proposed model
Federated learning (FL) is an effective technique to directly involve edge devices in machine learning training while preserving client privacy. However, the substantial communication overhead of FL makes training challenging when edge devices have limited network bandwidth. Existing work to optimize FL bandwidth overlooks downstream transmission and does not account for FL client sampling. In this paper we propose GlueFL, a framework that incorporates new client sampling and model compression algorithms to mitigate low download bandwidths of FL clients. GlueFL prioritizes recently used clients and bounds the number of changed positions in compression masks in each round. Across three popular FL datasets and three state-of-the-art strategies, GlueFL reduces downstream client bandwidth by 27% on average and reduces training time by 29% on average.
Federated Deep Learning frameworks can be used strategically to monitor Land Use locally and infer environmental impacts globally. Distributed data from across the world would be needed to build a global model for Land Use classification. The need for a Federated approach in this application domain would be to avoid transfer of data from distributed locations and save network bandwidth to reduce communication cost. We use a Federated UNet model for Semantic Segmentation of satellite and street view images. The novelty of the proposed architecture is the integration of Knowledge Distillation to reduce communication cost and response time. The accuracy obtained was above 95% and we also brought in a significant model compression to over 17 times and 62 times for street View and satellite images respectively. Our proposed framework has the potential to be a game-changer in real-time tracking of climate change across the planet.
Model compression and model defense for deep neural networks (DNNs) have been extensively and individually studied. Considering the co-importance of model compactness and robustness in practical applications, several prior works have explored to improve the adversarial robustness of the sparse neural networks. However, the structured sparse models obtained by the exiting works suffer severe performance degradation for both benign and robust accuracy, thereby causing a challenging dilemma between robustness and structuredness of the compact DNNs. To address this problem, in this paper, we propose CSTAR, an efficient solution that can simultaneously impose the low-rankness-based Compactness, high STructuredness and high Adversarial Robustness on the target DNN models. By formulating the low-rankness and robustness requirement within the same framework and globally determining the ranks, the compressed DNNs can simultaneously achieve high compression performance and strong adversarial robustness. Evaluations for various DNN models on different datasets demonstrate the effectiveness of CSTAR. Compared with the state-of-the-art robust structured pruning methods, CSTAR shows consistently better performance. For instance, when compressing ResNet-18 on CIFAR-10, CSTAR can achieve up to 20.07% and 11.91% improvement for benign accuracy and robust accuracy, respectively. For compressing ResNet-18 with 16x compression ratio on Imagenet, CSTAR can obtain 8.58% benign accuracy gain and 4.27% robust accuracy gain compared to the existing robust structured pruning method.
Experience management is an emerging business area where organizations focus on understanding the feedback of customers and employees in order to improve their end-to-end experiences. This results in a unique set of machine learning problems to help understand how people feel, discover issues they care about, and find which actions need to be taken on data that are different in content and distribution from traditional NLP domains. In this paper, we present a case study of building text analysis applications that perform multiple classification tasks efficiently in 12 languages in the nascent business area of experience management. In order to scale up modern ML methods on experience data, we leverage cross lingual and multi-task modeling techniques to consolidate our models into a single deployment to avoid overhead. We also make use of model compression and model distillation to reduce overall inference latency and hardware cost to the level acceptable for business needs while maintaining model prediction quality. Our findings show that multi-task modeling improves task performance for a subset of experience management tasks in both XLM-R and mBert architectures. Among the compressed architectures we explored, we found that MiniLM achieved the best compression/performance tradeoff. Our case study demonstrates a speedup of up to 15.61x with 2.60% average task degradation (or 3.29x speedup with 1.71% degradation) and estimated savings of 44% over using the original full-size model. These results demonstrate a successful scaling up of text classification for the challenging new area of ML for experience management.
Approximating radiance fields with volumetric grids is one of promising directions for improving NeRF, represented by methods like Plenoxels and DVGO, which achieve super-fast training convergence and real-time rendering. However, these methods typically require a tremendous storage overhead, costing up to hundreds of megabytes of disk space and runtime memory for a single scene. We address this issue in this paper by introducing a simple yet effective framework, called vector quantized radiance fields (VQRF), for compressing these volume-grid-based radiance fields. We first present a robust and adaptive metric for estimating redundancy in grid models and performing voxel pruning by better exploring intermediate outputs of volumetric rendering. A trainable vector quantization is further proposed to improve the compactness of grid models. In combination with an efficient joint tuning strategy and post-processing, our method can achieve a compression ratio of 100$\times$ by reducing the overall model size to 1 MB with negligible loss on visual quality. Extensive experiments demonstrate that the proposed framework is capable of achieving unrivaled performance and well generalization across multiple methods with distinct volumetric structures, facilitating the wide use of volumetric radiance fields methods in real-world applications. Code Available at \url{https://github.com/AlgoHunt/VQRF}
The problem of broad practical interest in spatiotemporal data analysis, i.e., discovering interpretable dynamic patterns from spatiotemporal data, is studied in this paper. Towards this end, we develop a time-varying reduced-rank vector autoregression (VAR) model whose coefficient matrices are parameterized by low-rank tensor factorization. Benefiting from the tensor factorization structure, the proposed model can simultaneously achieve model compression and pattern discovery. In particular, the proposed model allows one to characterize nonstationarity and time-varying system behaviors underlying spatiotemporal data. To evaluate the proposed model, extensive experiments are conducted on various spatiotemporal data representing different nonlinear dynamical systems, including fluid dynamics, sea surface temperature, USA surface temperature, and NYC taxi trips. Experimental results demonstrate the effectiveness of modeling spatiotemporal data and characterizing spatial/temporal patterns with the proposed model. In the spatial context, the spatial patterns can be automatically extracted and intuitively characterized by the spatial modes. In the temporal context, the complex time-varying system behaviors can be revealed by the temporal modes in the proposed model. Thus, our model lays an insightful foundation for understanding complex spatiotemporal data in real-world dynamical systems. The adapted datasets and Python implementation are publicly available at https://github.com/xinychen/vars.
As a promising solution for model compression, knowledge distillation (KD) has been applied in recommender systems (RS) to reduce inference latency. Traditional solutions first train a full teacher model from the training data, and then transfer its knowledge (\ie \textit{soft labels}) to supervise the learning of a compact student model. However, we find such a standard distillation paradigm would incur serious bias issue -- popular items are more heavily recommended after the distillation. This effect prevents the student model from making accurate and fair recommendations, decreasing the effectiveness of RS. In this work, we identify the origin of the bias in KD -- it roots in the biased soft labels from the teacher, and is further propagated and intensified during the distillation. To rectify this, we propose a new KD method with a stratified distillation strategy. It first partitions items into multiple groups according to their popularity, and then extracts the ranking knowledge within each group to supervise the learning of the student. Our method is simple and teacher-agnostic -- it works on distillation stage without affecting the training of the teacher model. We conduct extensive theoretical and empirical studies to validate the effectiveness of our proposal. We release our code at: https://github.com/chengang95/UnKD.
For time-critical IoT applications using deep learning, inference acceleration through distributed computing is a promising approach to meet a stringent deadline. In this paper, we implement a working prototype of a new distributed inference acceleration method HALP using three raspberry Pi 4. HALP accelerates inference by designing a seamless collaboration among edge devices (EDs) in Edge Computing. We maximize the parallelization between communication and computation among the collaborative EDs by optimizing the task partitioning ratio based on the segment-based partitioning. Experimental results show that the distributed inference HALP achieves 1.7x inference acceleration for VGG-16. Then, we combine distributed inference with conventional neural network model compression by setting up different shrinking hyperparameters for MobileNet-V1. In this way, we can further accelerate inference but at the cost of inference accuracy loss. To strike a balance between latency and accuracy, we propose dynamic model selection to select a model which provides the highest accuracy within the latency constraint. It is shown that the model selection with distributed inference HALP can significantly improve service reliability compared to the conventional stand-alone computation.
Probabilistic circuits (PCs) are a tractable representation of probability distributions allowing for exact and efficient computation of likelihoods and marginals. There has been significant recent progress on improving the scale and expressiveness of PCs. However, PC training performance plateaus as model size increases. We discover that most capacity in existing large PC structures is wasted: fully-connected parameter layers are only sparsely used. We propose two operations: pruning and growing, that exploit the sparsity of PC structures. Specifically, the pruning operation removes unimportant sub-networks of the PC for model compression and comes with theoretical guarantees. The growing operation increases model capacity by increasing the size of the latent space. By alternatingly applying pruning and growing, we increase the capacity that is meaningfully used, allowing us to significantly scale up PC learning. Empirically, our learner achieves state-of-the-art likelihoods on MNIST-family image datasets and on Penn Tree Bank language data compared to other PC learners and less tractable deep generative models such as flow-based models and variational autoencoders (VAEs).
Vector Quantization (VQ) is an appealing model compression method to obtain a tiny model with less accuracy loss. While methods to obtain better codebooks and codes under fixed clustering dimensionality have been extensively studied, optimizations of the vectors in favour of clustering performance are not carefully considered, especially via the reduction of vector dimensionality. This paper reports our recent progress on the combination of dimensionality compression and vector quantization, proposing a Low-Rank Representation Vector Quantization ($\text{LR}^2\text{VQ}$) method that outperforms previous VQ algorithms in various tasks and architectures. $\text{LR}^2\text{VQ}$ joins low-rank representation with subvector clustering to construct a new kind of building block that is directly optimized through end-to-end training over the task loss. Our proposed design pattern introduces three hyper-parameters, the number of clusters $k$, the size of subvectors $m$ and the clustering dimensionality $\tilde{d}$. In our method, the compression ratio could be directly controlled by $m$, and the final accuracy is solely determined by $\tilde{d}$. We recognize $\tilde{d}$ as a trade-off between low-rank approximation error and clustering error and carry out both theoretical analysis and experimental observations that empower the estimation of the proper $\tilde{d}$ before fine-tunning. With a proper $\tilde{d}$, we evaluate $\text{LR}^2\text{VQ}$ with ResNet-18/ResNet-50 on ImageNet classification datasets, achieving 2.8\%/1.0\% top-1 accuracy improvements over the current state-of-the-art VQ-based compression algorithms with 43$\times$/31$\times$ compression factor.
Knowledge distillation (KD) has been a ubiquitous method for model compression to strengthen the capability of a lightweight model with the transferred knowledge from the teacher. In particular, KD has been employed in quantization-aware training (QAT) of Transformer encoders like BERT to improve the accuracy of the student model with the reduced-precision weight parameters. However, little is understood about which of the various KD approaches best fits the QAT of Transformers. In this work, we provide an in-depth analysis of the mechanism of KD on attention recovery of quantized large Transformers. In particular, we reveal that the previously adopted MSE loss on the attention score is insufficient for recovering the self-attention information. Therefore, we propose two KD methods; attention-map and attention-output losses. Furthermore, we explore the unification of both losses to address task-dependent preference between attention-map and output losses. The experimental results on various Transformer encoder models demonstrate that the proposed KD methods achieve state-of-the-art accuracy for QAT with sub-2-bit weight quantization.