PyTorchKR Papers

최신 논문

총 3,892편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2405.13514

Joint Optimization of Streaming and Non-Streaming Automatic Speech Recognition with Multi-Decoder and Knowledge Distillation

End-to-end (E2E) automatic speech recognition (ASR) can operate in two modes: streaming and non-streaming, each with its pros and cons. Streaming ASR processes the speech frames in real-time as it is being received, while non-streaming ASR waits for the entire speech utterance; thus, professionals may have to operate in either mode to satisfy their application. In this work, we present joint optimization of streaming and non-streaming ASR based on multi-decoder and knowledge distillation. Primarily, we study 1) the encoder integration of these ASR modules, followed by 2) separate decoders to make the switching mode flexible, and enhancing performance by 3) incorporating similarity-preserving knowledge distillation between the two modular encoders and decoders. Evaluation results show 2.6%-5.3% relative character error rate reductions (CERR) on CSJ for streaming ASR, and 8.3%-9.7% relative CERRs for non-streaming ASR within a single model compared to multiple standalone modules.

arXiv:2405.13923공식 코드 · Papers with Code

Why Not Transform Chat Large Language Models to Non-English?

The scarcity of non-English data limits the development of non-English large language models (LLMs). Transforming English-centric LLMs to non-English has been identified as an effective and resource-efficient method. Previous works start from base LLMs and perform knowledge distillation (KD) with data generated by stronger LLMs, e.g. GPT-4. Compared to base LLMs, chat LLMs are further optimized for advanced abilities, e.g. multi-turn conversation and human preference alignment, and thus more powerful in both helpfulness and safety. However, transforming a chat LLM involves two critical issues: (1) How can we effectively transfer advanced abilities without their supervised data? (2) How can we prevent the original knowledge from catastrophic forgetting during transformation? We target these issues by introducing a simple framework called TransLLM. For the first issue, TransLLM divides the transfer problem into some common sub-tasks with the translation chain-of-thought, which uses the translation as the bridge between English and non-English step-by-step. We further enhance the performance of sub-tasks with publicly available data. For the second issue, we propose a method comprising two synergistic components: low-rank adaptation for training to maintain the original LLM parameters, and recovery KD, which utilizes data generated by the chat LLM itself to recover the original knowledge from the frozen parameters. In the experiments, we transform the LLaMA-2-chat-7B to the Thai language. Our method, using only single-turn data, outperforms strong baselines and ChatGPT on multi-turn benchmark MT-bench. Furthermore, our method, without safety data, rejects more harmful queries of safety benchmark AdvBench than both ChatGPT and GPT-4.

arXiv:2405.14136

Efficient Multitask Dense Predictor via Binarization

Multi-task learning for dense prediction has emerged as a pivotal area in computer vision, enabling simultaneous processing of diverse yet interrelated pixel-wise prediction tasks. However, the substantial computational demands of state-of-the-art (SoTA) models often limit their widespread deployment. This paper addresses this challenge by introducing network binarization to compress resource-intensive multi-task dense predictors. Specifically, our goal is to significantly accelerate multi-task dense prediction models via Binary Neural Networks (BNNs) while maintaining and even improving model performance at the same time. To reach this goal, we propose a Binary Multi-task Dense Predictor, Bi-MTDP, and several variants of Bi-MTDP, in which a multi-task dense predictor is constructed via specified binarized modules. Our systematical analysis of this predictor reveals that performance drop from binarization is primarily caused by severe information degradation. To address this issue, we introduce a deep information bottleneck layer that enforces representations for downstream tasks satisfying Gaussian distribution in forward propagation. Moreover, we introduce a knowledge distillation mechanism to correct the direction of information flow in backward propagation. Intriguingly, one variant of Bi-MTDP outperforms full-precision (FP) multi-task dense prediction SoTAs, ARTC (CNN-based) and InvPT (ViT-Based). This result indicates that Bi-MTDP is not merely a naive trade-off between performance and efficiency, but is rather a benefit of the redundant information flow thanks to the multi-task architecture. Code is available at https://github.com/42Shawn/BiMTDP.

arXiv:2405.14791공식 코드 · Papers with Code

Recurrent Early Exits for Federated Learning with Heterogeneous Clients

Federated learning (FL) has enabled distributed learning of a model across multiple clients in a privacy-preserving manner. One of the main challenges of FL is to accommodate clients with varying hardware capacities; clients have differing compute and memory requirements. To tackle this challenge, recent state-of-the-art approaches leverage the use of early exits. Nonetheless, these approaches fall short of mitigating the challenges of joint learning multiple exit classifiers, often relying on hand-picked heuristic solutions for knowledge distillation among classifiers and/or utilizing additional layers for weaker classifiers. In this work, instead of utilizing multiple classifiers, we propose a recurrent early exit approach named ReeFL that fuses features from different sub-models into a single shared classifier. Specifically, we use a transformer-based early-exit module shared among sub-models to i) better exploit multi-layer feature representations for task-specific prediction and ii) modulate the feature representation of the backbone model for subsequent predictions. We additionally present a per-client self-distillation approach where the best sub-model is automatically selected as the teacher of the other sub-models at each client. Our experiments on standard image and speech classification benchmarks across various emerging federated fine-tuning baselines demonstrate ReeFL's effectiveness over previous works.

arXiv:2405.14365공식 코드 · Papers with Code

JiuZhang3.0: Efficiently Improving Mathematical Reasoning by Training Small Data Synthesis Models

Mathematical reasoning is an important capability of large language models~(LLMs) for real-world applications. To enhance this capability, existing work either collects large-scale math-related texts for pre-training, or relies on stronger LLMs (\eg GPT-4) to synthesize massive math problems. Both types of work generally lead to large costs in training or synthesis. To reduce the cost, based on open-source available texts, we propose an efficient way that trains a small LLM for math problem synthesis, to efficiently generate sufficient high-quality pre-training data. To achieve it, we create a dataset using GPT-4 to distill its data synthesis capability into the small LLM. Concretely, we craft a set of prompts based on human education stages to guide GPT-4, to synthesize problems covering diverse math knowledge and difficulty levels. Besides, we adopt the gradient-based influence estimation method to select the most valuable math-related texts. The both are fed into GPT-4 for creating the knowledge distillation dataset to train the small LLM. We leverage it to synthesize 6 million math problems for pre-training our JiuZhang3.0 model, which only needs to invoke GPT-4 API 9.3k times and pre-train on 4.6B data. Experimental results have shown that JiuZhang3.0 achieves state-of-the-art performance on several mathematical reasoning datasets, under both natural language reasoning and tool manipulation settings. Our code and data will be publicly released in \url{https://github.com/RUCAIBox/JiuZhang3.0}.

arXiv:2405.14307공식 코드 · Papers with Code

AdaGMLP: AdaBoosting GNN-to-MLP Knowledge Distillation

Graph Neural Networks (GNNs) have revolutionized graph-based machine learning, but their heavy computational demands pose challenges for latency-sensitive edge devices in practical industrial applications. In response, a new wave of methods, collectively known as GNN-to-MLP Knowledge Distillation, has emerged. They aim to transfer GNN-learned knowledge to a more efficient MLP student, which offers faster, resource-efficient inference while maintaining competitive performance compared to GNNs. However, these methods face significant challenges in situations with insufficient training data and incomplete test data, limiting their applicability in real-world applications. To address these challenges, we propose AdaGMLP, an AdaBoosting GNN-to-MLP Knowledge Distillation framework. It leverages an ensemble of diverse MLP students trained on different subsets of labeled nodes, addressing the issue of insufficient training data. Additionally, it incorporates a Node Alignment technique for robust predictions on test data with missing or incomplete features. Our experiments on seven benchmark datasets with different settings demonstrate that AdaGMLP outperforms existing G2M methods, making it suitable for a wide range of latency-sensitive real-world applications. We have submitted our code to the GitHub repository (https://github.com/WeigangLu/AdaGMLP-KDD24).

arXiv:2405.14200공식 코드 · Papers with Code

Awesome Multi-modal Object Tracking

Multi-modal object tracking (MMOT) is an emerging field that combines data from various modalities, \eg vision (RGB), depth, thermal infrared, event, language and audio, to estimate the state of an arbitrary object in a video sequence. It is of great significance for many applications such as autonomous driving and intelligent surveillance. In recent years, MMOT has received more and more attention. However, existing MMOT algorithms mainly focus on two modalities (\eg RGB+depth, RGB+thermal infrared, and RGB+language). To leverage more modalities, some recent efforts have been made to learn a unified visual object tracking model for any modality. Additionally, some large-scale multi-modal tracking benchmarks have been established by simultaneously providing more than two modalities, such as vision-language-audio (\eg WebUAV-3M) and vision-depth-language (\eg UniMod1K). To track the latest progress in MMOT, we conduct a comprehensive investigation in this report. Specifically, we first divide existing MMOT tasks into five main categories, \ie RGBL tracking, RGBE tracking, RGBD tracking, RGBT tracking, and miscellaneous (RGB+X), where X can be any modality, such as language, depth, and event. Then, we analyze and summarize each MMOT task, focusing on widely used datasets and mainstream tracking algorithms based on their technical paradigms (\eg self-supervised learning, prompt learning, knowledge distillation, generative models, and state space models). Finally, we maintain a continuously updated paper list for MMOT at https://github.com/983632847/Awesome-Multimodal-Object-Tracking.

arXiv:2405.12353

TinyM$^2$Net-V3: Memory-Aware Compressed Multimodal Deep Neural Networks for Sustainable Edge Deployment

The advancement of sophisticated artificial intelligence (AI) algorithms has led to a notable increase in energy usage and carbon dioxide emissions, intensifying concerns about climate change. This growing problem has brought the environmental sustainability of AI technologies to the forefront, especially as they expand across various sectors. In response to these challenges, there is an urgent need for the development of sustainable AI solutions. These solutions must focus on energy-efficient embedded systems that are capable of handling diverse data types even in environments with limited resources, thereby ensuring both technological progress and environmental responsibility. Integrating complementary multimodal data into tiny machine learning models for edge devices is challenging due to increased complexity, latency, and power consumption. This work introduces TinyM$^2$Net-V3, a system that processes different modalities of complementary data, designs deep neural network (DNN) models, and employs model compression techniques including knowledge distillation and low bit-width quantization with memory-aware considerations to fit models within lower memory hierarchy levels, reducing latency and enhancing energy efficiency on resource-constrained devices. We evaluated TinyM$^2$Net-V3 in two multimodal case studies: COVID-19 detection using cough, speech, and breathing audios, and pose classification from depth and thermal images. With tiny inference models (6 KB and 58 KB), we achieved 92.95% and 90.7% accuracies, respectively. Our tiny machine learning models, deployed on resource limited hardware, demonstrated low latencies within milliseconds and very high power efficiency.

arXiv:2405.12509공식 코드 · Papers with Code

Active Object Detection with Knowledge Aggregation and Distillation from Large Models

Accurately detecting active objects undergoing state changes is essential for comprehending human interactions and facilitating decision-making. The existing methods for active object detection (AOD) primarily rely on visual appearance of the objects within input, such as changes in size, shape and relationship with hands. However, these visual changes can be subtle, posing challenges, particularly in scenarios with multiple distracting no-change instances of the same category. We observe that the state changes are often the result of an interaction being performed upon the object, thus propose to use informed priors about object related plausible interactions (including semantics and visual appearance) to provide more reliable cues for AOD. Specifically, we propose a knowledge aggregation procedure to integrate the aforementioned informed priors into oracle queries within the teacher decoder, offering more object affordance commonsense to locate the active object. To streamline the inference process and reduce extra knowledge inputs, we propose a knowledge distillation approach that encourages the student decoder to mimic the detection capabilities of the teacher decoder using the oracle query by replicating its predictions and attention. Our proposed framework achieves state-of-the-art performance on four datasets, namely Ego4D, Epic-Kitchens, MECCANO, and 100DOH, which demonstrates the effectiveness of our approach in improving AOD.

arXiv:2405.12503공식 코드 · Papers with Code

CLRKDNet: Speeding up Lane Detection with Knowledge Distillation

Road lanes are integral components of the visual perception systems in intelligent vehicles, playing a pivotal role in safe navigation. In lane detection tasks, balancing accuracy with real-time performance is essential, yet existing methods often sacrifice one for the other. To address this trade-off, we introduce CLRKDNet, a streamlined model that balances detection accuracy with real-time performance. The state-of-the-art model CLRNet has demonstrated exceptional performance across various datasets, yet its computational overhead is substantial due to its Feature Pyramid Network (FPN) and muti-layer detection head architecture. Our method simplifies both the FPN structure and detection heads, redesigning them to incorporate a novel teacher-student distillation process alongside a newly introduced series of distillation losses. This combination reduces inference time by up to 60% while maintaining detection accuracy comparable to CLRNet. This strategic balance of accuracy and speed makes CLRKDNet a viable solution for real-time lane detection tasks in autonomous driving applications.

arXiv:2405.12419

GeoMask3D: Geometrically Informed Mask Selection for Self-Supervised Point Cloud Learning in 3D

We introduce a pioneering approach to self-supervised learning for point clouds, employing a geometrically informed mask selection strategy called GeoMask3D (GM3D) to boost the efficiency of Masked Auto Encoders (MAE). Unlike the conventional method of random masking, our technique utilizes a teacher-student model to focus on intricate areas within the data, guiding the model's focus toward regions with higher geometric complexity. This strategy is grounded in the hypothesis that concentrating on harder patches yields a more robust feature representation, as evidenced by the improved performance on downstream tasks. Our method also presents a complete-to-partial feature-level knowledge distillation technique designed to guide the prediction of geometric complexity utilizing a comprehensive context from feature-level information. Extensive experiments confirm our method's superiority over State-Of-The-Art (SOTA) baselines, demonstrating marked improvements in classification, and few-shot tasks.

arXiv:2405.12944공식 코드 · Papers with Code

AMFD: Distillation via Adaptive Multimodal Fusion for Multispectral Pedestrian Detection

Multispectral pedestrian detection has been shown to be effective in improving performance within complex illumination scenarios. However, prevalent double-stream networks in multispectral detection employ two separate feature extraction branches for multi-modal data, leading to nearly double the inference time compared to single-stream networks utilizing only one feature extraction branch. This increased inference time has hindered the widespread employment of multispectral pedestrian detection in embedded devices for autonomous systems. To address this limitation, various knowledge distillation methods have been proposed. However, traditional distillation methods focus only on the fusion features and ignore the large amount of information in the original multi-modal features, thereby restricting the student network's performance. To tackle the challenge, we introduce the Adaptive Modal Fusion Distillation (AMFD) framework, which can fully utilize the original modal features of the teacher network. Specifically, a Modal Extraction Alignment (MEA) module is utilized to derive learning weights for student networks, integrating focal and global attention mechanisms. This methodology enables the student network to acquire optimal fusion strategies independent from that of teacher network without necessitating an additional feature fusion module. Furthermore, we present the SMOD dataset, a well-aligned challenging multispectral dataset for detection. Extensive experiments on the challenging KAIST, LLVIP and SMOD datasets are conducted to validate the effectiveness of AMFD. The results demonstrate that our method outperforms existing state-of-the-art methods in both reducing log-average Miss Rate and improving mean Average Precision. The code is available at https://github.com/bigD233/AMFD.git.

arXiv:2405.11828공식 코드 · Papers with Code

Federated Learning for Time-Series Healthcare Sensing with Incomplete Modalities

Many healthcare sensing applications utilize multimodal time-series data from sensors embedded in mobile and wearable devices. Federated Learning (FL), with its privacy-preserving advantages, is particularly well-suited for health applications. However, most multimodal FL methods assume the availability of complete modality data for local training, which is often unrealistic. Moreover, recent approaches tackling incomplete modalities scale poorly and become inefficient as the number of modalities increases. To address these limitations, we propose FLISM, an efficient FL training algorithm with incomplete sensing modalities while maintaining high accuracy. FLISM employs three key techniques: (1) modality-invariant representation learning to extract effective features from clients with a diverse set of modalities, (2) modality quality-aware aggregation to prioritize contributions from clients with higher-quality modality data, and (3) global-aligned knowledge distillation to reduce local update shifts caused by modality differences. Extensive experiments on real-world datasets show that FLISM not only achieves high accuracy but is also faster and more efficient compared with state-of-the-art methods handling incomplete modality problems in FL. We release the code as open-source at https://github.com/AdibaOrz/FLISM.

arXiv:2405.11448

Cross-Domain Knowledge Distillation for Low-Resolution Human Pose Estimation

In practical applications of human pose estimation, low-resolution inputs frequently occur, and existing state-of-the-art models perform poorly with low-resolution images. This work focuses on boosting the performance of low-resolution models by distilling knowledge from a high-resolution model. However, we face the challenge of feature size mismatch and class number mismatch when applying knowledge distillation to networks with different input resolutions. To address this issue, we propose a novel cross-domain knowledge distillation (CDKD) framework. In this framework, we construct a scale-adaptive projector ensemble (SAPE) module to spatially align feature maps between models of varying input resolutions. It adopts a projector ensemble to map low-resolution features into multiple common spaces and adaptively merges them based on multi-scale information to match high-resolution features. Additionally, we construct a cross-class alignment (CCA) module to solve the problem of the mismatch of class numbers. By combining an easy-to-hard training (ETHT) strategy, the CCA module further enhances the distillation performance. The effectiveness and efficiency of our approach are demonstrated by extensive experiments on two common benchmark datasets: MPII and COCO. The code is made available in supplementary material.

arXiv:2405.11852

Evolving Storytelling: Benchmarks and Methods for New Character Customization with Diffusion Models

Diffusion-based models for story visualization have shown promise in generating content-coherent images for storytelling tasks. However, how to effectively integrate new characters into existing narratives while maintaining character consistency remains an open problem, particularly with limited data. Two major limitations hinder the progress: (1) the absence of a suitable benchmark due to potential character leakage and inconsistent text labeling, and (2) the challenge of distinguishing between new and old characters, leading to ambiguous results. To address these challenges, we introduce the NewEpisode benchmark, comprising refined datasets designed to evaluate generative models' adaptability in generating new stories with fresh characters using just a single example story. The refined dataset involves refined text prompts and eliminates character leakage. Additionally, to mitigate the character confusion of generated results, we propose EpicEvo, a method that customizes a diffusion-based visual story generation model with a single story featuring the new characters seamlessly integrating them into established character dynamics. EpicEvo introduces a novel adversarial character alignment module to align the generated images progressively in the diffusive process, with exemplar images of new characters, while applying knowledge distillation to prevent forgetting of characters and background details. Our evaluation quantitatively demonstrates that EpicEvo outperforms existing baselines on the NewEpisode benchmark, and qualitative studies confirm its superior customization of visual story generation in diffusion models. In summary, EpicEvo provides an effective way to incorporate new characters using only one example story, unlocking new possibilities for applications such as serialized cartoons.

arXiv:2405.11823

Stereo-Knowledge Distillation from dpMV to Dual Pixels for Light Field Video Reconstruction

Dual pixels contain disparity cues arising from the defocus blur. This disparity information is useful for many vision tasks ranging from autonomous driving to 3D creative realism. However, directly estimating disparity from dual pixels is less accurate. This work hypothesizes that distilling high-precision dark stereo knowledge, implicitly or explicitly, to efficient dual-pixel student networks enables faithful reconstructions. This dark knowledge distillation should also alleviate stereo-synchronization setup and calibration costs while dramatically increasing parameter and inference time efficiency. We collect the first and largest 3-view dual-pixel video dataset, dpMV, to validate our explicit dark knowledge distillation hypothesis. We show that these methods outperform purely monocular solutions, especially in challenging foreground-background separation regions using faithful guidance from dual pixels. Finally, we demonstrate an unconventional use case unlocked by dpMV and implicit dark knowledge distillation from an ensemble of teachers for Light Field (LF) video reconstruction. Our LF video reconstruction method is the fastest and most temporally consistent to date. It remains competitive in reconstruction fidelity while offering many other essential properties like high parameter efficiency, implicit disocclusion handling, zero-shot cross-dataset transfer, geometrically consistent inference on higher spatial-angular resolutions, and adaptive baseline control. All source code is available at the anonymous repository https://github.com/Aryan-Garg.

arXiv:2405.11614

Nickel and Diming Your GAN: A Dual-Method Approach to Enhancing GAN Efficiency via Knowledge Distillation

In this paper, we address the challenge of compressing generative adversarial networks (GANs) for deployment in resource-constrained environments by proposing two novel methodologies: Distribution Matching for Efficient compression (DiME) and Network Interactive Compression via Knowledge Exchange and Learning (NICKEL). DiME employs foundation models as embedding kernels for efficient distribution matching, leveraging maximum mean discrepancy to facilitate effective knowledge distillation. Simultaneously, NICKEL employs an interactive compression method that enhances the communication between the student generator and discriminator, achieving a balanced and stable compression process. Our comprehensive evaluation on the StyleGAN2 architecture with the FFHQ dataset shows the effectiveness of our approach, with NICKEL & DiME achieving FID scores of 10.45 and 15.93 at compression rates of 95.73% and 98.92%, respectively. Remarkably, our methods sustain generative quality even at an extreme compression rate of 99.69%, surpassing the previous state-of-the-art performance by a large margin. These findings not only demonstrate our methodologies' capacity to significantly lower GANs' computational demands but also pave the way for deploying high-quality GAN models in settings with limited resources. Our code will be released soon.

arXiv:2405.11533

Hierarchical Selective Classification

Deploying deep neural networks for risk-sensitive tasks necessitates an uncertainty estimation mechanism. This paper introduces hierarchical selective classification, extending selective classification to a hierarchical setting. Our approach leverages the inherent structure of class relationships, enabling models to reduce the specificity of their predictions when faced with uncertainty. In this paper, we first formalize hierarchical risk and coverage, and introduce hierarchical risk-coverage curves. Next, we develop algorithms for hierarchical selective classification (which we refer to as "inference rules"), and propose an efficient algorithm that guarantees a target accuracy constraint with high probability. Lastly, we conduct extensive empirical studies on over a thousand ImageNet classifiers, revealing that training regimes such as CLIP, pretraining on ImageNet21k and knowledge distillation boost hierarchical selective performance.

arXiv:2405.11809

Distill-then-prune: An Efficient Compression Framework for Real-time Stereo Matching Network on Edge Devices

In recent years, numerous real-time stereo matching methods have been introduced, but they often lack accuracy. These methods attempt to improve accuracy by introducing new modules or integrating traditional methods. However, the improvements are only modest. In this paper, we propose a novel strategy by incorporating knowledge distillation and model pruning to overcome the inherent trade-off between speed and accuracy. As a result, we obtained a model that maintains real-time performance while delivering high accuracy on edge devices. Our proposed method involves three key steps. Firstly, we review state-of-the-art methods and design our lightweight model by removing redundant modules from those efficient models through a comparison of their contributions. Next, we leverage the efficient model as the teacher to distill knowledge into the lightweight model. Finally, we systematically prune the lightweight model to obtain the final model. Through extensive experiments conducted on two widely-used benchmarks, Sceneflow and KITTI, we perform ablation studies to analyze the effectiveness of each module and present our state-of-the-art results.

arXiv:2405.11704

Efficiency optimization of large-scale language models based on deep learning in natural language processing tasks

The internal structure and operation mechanism of large-scale language models are analyzed theoretically, especially how Transformer and its derivative architectures can restrict computing efficiency while capturing long-term dependencies. Further, we dig deep into the efficiency bottleneck of the training phase, and evaluate in detail the contribution of adaptive optimization algorithms (such as AdamW), massively parallel computing techniques, and mixed precision training strategies to accelerate convergence and reduce memory footprint. By analyzing the mathematical principles and implementation details of these algorithms, we reveal how they effectively improve training efficiency in practice. In terms of model deployment and inference optimization, this paper systematically reviews the latest advances in model compression techniques, focusing on strategies such as quantification, pruning, and knowledge distillation. By comparing the theoretical frameworks of these techniques and their effects in different application scenarios, we demonstrate their ability to significantly reduce model size and inference delay while maintaining model prediction accuracy. In addition, this paper critically examines the limitations of current efficiency optimization methods, such as the increased risk of overfitting, the control of performance loss after compression, and the problem of algorithm generality, and proposes some prospects for future research. In conclusion, this study provides a comprehensive theoretical framework for understanding the efficiency optimization of large-scale language models.