PyTorchKR Papers

최신 논문

총 4,526편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2401.17244공식 코드 · Papers with Code

LLaMP: Large Language Model Made Powerful for High-fidelity Materials Knowledge Retrieval and Distillation

Reducing hallucination of Large Language Models (LLMs) is imperative for use in the sciences, where reliability and reproducibility are crucial. However, LLMs inherently lack long-term memory, making it a nontrivial, ad hoc, and inevitably biased task to fine-tune them on domain-specific literature and data. Here we introduce LLaMP, a multimodal retrieval-augmented generation (RAG) framework of hierarchical reasoning-and-acting (ReAct) agents that can dynamically and recursively interact with computational and experimental data on Materials Project (MP) and run atomistic simulations via high-throughput workflow interface. Without fine-tuning, LLaMP demonstrates strong tool usage ability to comprehend and integrate various modalities of materials science concepts, fetch relevant data stores on the fly, process higher-order data (such as crystal structure and elastic tensor), and streamline complex tasks in computational materials and chemistry. We propose a simple metric combining uncertainty and confidence estimates to evaluate the self-consistency of responses by LLaMP and vanilla LLMs. Our benchmark shows that LLaMP effectively mitigates the intrinsic bias in LLMs, counteracting the errors on bulk moduli, electronic bandgaps, and formation energies that seem to derive from mixed data sources. We also demonstrate LLaMP's capability to edit crystal structures and run annealing molecular dynamics simulations using pre-trained machine-learning force fields. The framework offers an intuitive and nearly hallucination-free approach to exploring and scaling materials informatics, and establishes a pathway for knowledge distillation and fine-tuning other language models. Code and live demo are available at https://github.com/chiang-yuan/llamp

arXiv:2401.16055공식 코드 · Papers with Code

Stolen Subwords: Importance of Vocabularies for Machine Translation Model Stealing

In learning-based functionality stealing, the attacker is trying to build a local model based on the victim's outputs. The attacker has to make choices regarding the local model's architecture, optimization method and, specifically for NLP models, subword vocabulary, such as BPE. On the machine translation task, we explore (1) whether the choice of the vocabulary plays a role in model stealing scenarios and (2) if it is possible to extract the victim's vocabulary. We find that the vocabulary itself does not have a large effect on the local model's performance. Given gray-box model access, it is possible to collect the victim's vocabulary by collecting the outputs (detokenized subwords on the output). The results of the minimum effect of vocabulary choice are important more broadly for black-box knowledge distillation.

arXiv:2401.16367공식 코드 · Papers with Code

TQCompressor: improving tensor decomposition methods in neural networks via permutations

We introduce TQCompressor, a novel method for neural network model compression with improved tensor decompositions. We explore the challenges posed by the computational and storage demands of pre-trained language models in NLP tasks and propose a permutation-based enhancement to Kronecker decomposition. This enhancement makes it possible to reduce loss in model expressivity which is usually associated with factorization. We demonstrate this method applied to the GPT-2$_{small}$. The result of the compression is TQCompressedGPT-2 model, featuring 81 mln. parameters compared to 124 mln. in the GPT-2$_{small}$. We make TQCompressedGPT-2 publicly available. We further enhance the performance of the TQCompressedGPT-2 through a training strategy involving multi-step knowledge distillation, using only a 3.1% of the OpenWebText. TQCompressedGPT-2 surpasses DistilGPT-2 and KnGPT-2 in comparative evaluations, marking an advancement in the efficient and effective deployment of models in resource-constrained environments.

arXiv:2401.15366

Face to Cartoon Incremental Super-Resolution using Knowledge Distillation

Facial super-resolution/hallucination is an important area of research that seeks to enhance low-resolution facial images for a variety of applications. While Generative Adversarial Networks (GANs) have shown promise in this area, their ability to adapt to new, unseen data remains a challenge. This paper addresses this problem by proposing an incremental super-resolution using GANs with knowledge distillation (ISR-KD) for face to cartoon. Previous research in this area has not investigated incremental learning, which is critical for real-world applications where new data is continually being generated. The proposed ISR-KD aims to develop a novel unified framework for facial super-resolution that can handle different settings, including different types of faces such as cartoon face and various levels of detail. To achieve this, a GAN-based super-resolution network was pre-trained on the CelebA dataset and then incrementally trained on the iCartoonFace dataset, using knowledge distillation to retain performance on the CelebA test set while improving the performance on iCartoonFace test set. Our experiments demonstrate the effectiveness of knowledge distillation in incrementally adding capability to the model for cartoon face super-resolution while retaining the learned knowledge for facial hallucination tasks in GANs.

arXiv:2401.15275

Dynamic Transformer Architecture for Continual Learning of Multimodal Tasks

Transformer neural networks are increasingly replacing prior architectures in a wide range of applications in different data modalities. The increasing size and computational demands of fine-tuning large pre-trained transformer neural networks pose significant challenges for the widespread adoption of these models for applications that demand on-edge computing. To tackle this challenge, continual learning (CL) emerges as a solution by facilitating the transfer of knowledge across tasks that arrive sequentially for an autonomously learning agent. However, current CL methods mainly focus on learning tasks that are exclusively vision-based or language-based. We propose a transformer-based CL framework focusing on learning tasks that involve both vision and language, known as Vision-and-Language (VaL) tasks. Due to the success of transformers in other modalities, our architecture has the potential to be used in multimodal learning settings. In our framework, we benefit from introducing extra parameters to a base transformer to specialize the network for each task. As a result, we enable dynamic model expansion to learn several tasks in a sequence. We also use knowledge distillation to benefit from relevant past experiences to learn the current task more efficiently. Our proposed method, Task Attentive Multimodal Continual Learning (TAM-CL), allows for the exchange of information between tasks while mitigating the problem of catastrophic forgetting. Notably, our approach is scalable, incurring minimal memory and time overhead. TAM-CL achieves state-of-the-art (SOTA) performance on challenging multimodal tasks

arXiv:2401.15489

Distilling Privileged Multimodal Information for Expression Recognition using Optimal Transport

Deep learning models for multimodal expression recognition have reached remarkable performance in controlled laboratory environments because of their ability to learn complementary and redundant semantic information. However, these models struggle in the wild, mainly because of the unavailability and quality of modalities used for training. In practice, only a subset of the training-time modalities may be available at test time. Learning with privileged information enables models to exploit data from additional modalities that are only available during training. State-of-the-art knowledge distillation (KD) methods have been proposed to distill information from multiple teacher models (each trained on a modality) to a common student model. These privileged KD methods typically utilize point-to-point matching, yet have no explicit mechanism to capture the structural information in the teacher representation space formed by introducing the privileged modality. Experiments were performed on two challenging problems - pain estimation on the Biovid dataset (ordinal classification) and arousal-valance prediction on the Affwild2 dataset (regression). Results show that our proposed method can outperform state-of-the-art privileged KD methods on these problems. The diversity among modalities and fusion architectures indicates that PKDOT is modality- and model-agnostic.

arXiv:2401.15347

A Comprehensive Survey of Compression Algorithms for Language Models

How can we compress language models without sacrificing accuracy? The number of compression algorithms for language models is rapidly growing to benefit from remarkable advances of recent language models without side effects due to the gigantic size of language models, such as increased carbon emissions and expensive maintenance fees. While numerous compression algorithms have shown remarkable progress in compressing language models, it ironically becomes challenging to capture emerging trends and identify the fundamental concepts underlying them due to the excessive number of algorithms. In this paper, we survey and summarize diverse compression algorithms including pruning, quantization, knowledge distillation, low-rank approximation, parameter sharing, and efficient architecture design. We not only summarize the overall trend of diverse compression algorithms but also select representative algorithms and provide in-depth analyses of them. We discuss the value of each category of compression algorithms, and the desired properties of low-cost compression algorithms which have a significant impact due to the emergence of large language models. Finally, we introduce promising future research topics based on our survey results.

arXiv:2401.15123

Large Language Model Guided Knowledge Distillation for Time Series Anomaly Detection

Self-supervised methods have gained prominence in time series anomaly detection due to the scarcity of available annotations. Nevertheless, they typically demand extensive training data to acquire a generalizable representation map, which conflicts with scenarios of a few available samples, thereby limiting their performance. To overcome the limitation, we propose \textbf{AnomalyLLM}, a knowledge distillation-based time series anomaly detection approach where the student network is trained to mimic the features of the large language model (LLM)-based teacher network that is pretrained on large-scale datasets. During the testing phase, anomalies are detected when the discrepancy between the features of the teacher and student networks is large. To circumvent the student network from learning the teacher network's feature of anomalous samples, we devise two key strategies. 1) Prototypical signals are incorporated into the student network to consolidate the normal feature extraction. 2) We use synthetic anomalies to enlarge the representation gap between the two networks. AnomalyLLM demonstrates state-of-the-art performance on 15 datasets, improving accuracy by at least 14.5\% in the UCR dataset.

arXiv:2401.14211공식 코드 · Papers with Code

Communication-Efficient Federated Learning through Adaptive Weight Clustering and Server-Side Distillation

Federated Learning (FL) is a promising technique for the collaborative training of deep neural networks across multiple devices while preserving data privacy. Despite its potential benefits, FL is hindered by excessive communication costs due to repeated server-client communication during training. To address this challenge, model compression techniques, such as sparsification and weight clustering are applied, which often require modifying the underlying model aggregation schemes or involve cumbersome hyperparameter tuning, with the latter not only adjusts the model's compression rate but also limits model's potential for continuous improvement over growing data. In this paper, we propose FedCompress, a novel approach that combines dynamic weight clustering and server-side knowledge distillation to reduce communication costs while learning highly generalizable models. Through a comprehensive evaluation on diverse public datasets, we demonstrate the efficacy of our approach compared to baselines in terms of communication costs and inference speed.

arXiv:2401.14038공식 코드 · Papers with Code

Deep Clustering with Diffused Sampling and Hardness-aware Self-distillation

Deep clustering has gained significant attention due to its capability in learning clustering-friendly representations without labeled data. However, previous deep clustering methods tend to treat all samples equally, which neglect the variance in the latent distribution and the varying difficulty in classifying or clustering different samples. To address this, this paper proposes a novel end-to-end deep clustering method with diffused sampling and hardness-aware self-distillation (HaDis). Specifically, we first align one view of instances with another view via diffused sampling alignment (DSA), which helps improve the intra-cluster compactness. To alleviate the sampling bias, we present the hardness-aware self-distillation (HSD) mechanism to mine the hardest positive and negative samples and adaptively adjust their weights in a self-distillation fashion, which is able to deal with the potential imbalance in sample contributions during optimization. Further, the prototypical contrastive learning is incorporated to simultaneously enhance the inter-cluster separability and intra-cluster compactness. Experimental results on five challenging image datasets demonstrate the superior clustering performance of our HaDis method over the state-of-the-art. Source code is available at https://github.com/Regan-Zhang/HaDis.

arXiv:2401.13937

Self-supervised Video Object Segmentation with Distillation Learning of Deformable Attention

Video object segmentation is a fundamental research problem in computer vision. Recent techniques have often applied attention mechanism to object representation learning from video sequences. However, due to temporal changes in the video data, attention maps may not well align with the objects of interest across video frames, causing accumulated errors in long-term video processing. In addition, existing techniques have utilised complex architectures, requiring highly computational complexity and hence limiting the ability to integrate video object segmentation into low-powered devices. To address these issues, we propose a new method for self-supervised video object segmentation based on distillation learning of deformable attention. Specifically, we devise a lightweight architecture for video object segmentation that is effectively adapted to temporal changes. This is enabled by deformable attention mechanism, where the keys and values capturing the memory of a video sequence in the attention module have flexible locations updated across frames. The learnt object representations are thus adaptive to both the spatial and temporal dimensions. We train the proposed architecture in a self-supervised fashion through a new knowledge distillation paradigm where deformable attention maps are integrated into the distillation loss. We qualitatively and quantitatively evaluate our method and compare it with existing methods on benchmark datasets including DAVIS 2016/2017 and YouTube-VOS 2018/2019. Experimental results verify the superiority of our method via its achieved state-of-the-art performance and optimal memory usage.

arXiv:2401.12987공식 코드 · Papers with Code

TelME: Teacher-leading Multimodal Fusion Network for Emotion Recognition in Conversation

Emotion Recognition in Conversation (ERC) plays a crucial role in enabling dialogue systems to effectively respond to user requests. The emotions in a conversation can be identified by the representations from various modalities, such as audio, visual, and text. However, due to the weak contribution of non-verbal modalities to recognize emotions, multimodal ERC has always been considered a challenging task. In this paper, we propose Teacher-leading Multimodal fusion network for ERC (TelME). TelME incorporates cross-modal knowledge distillation to transfer information from a language model acting as the teacher to the non-verbal students, thereby optimizing the efficacy of the weak modalities. We then combine multimodal features using a shifting fusion approach in which student networks support the teacher. TelME achieves state-of-the-art performance in MELD, a multi-speaker conversation dataset for ERC. Finally, we demonstrate the effectiveness of our components through additional experiments.

arXiv:2401.13174

Towards Complementary Knowledge Distillation for Efficient Dense Image Prediction

It has been revealed that small efficient dense image prediction (EDIP) models, trained using the knowledge distillation (KD) framework, encounter two key challenges, including maintaining boundary region completeness and preserving target region connectivity, despite their favorable capacity to recognize main object regions. In this work, we propose a complementary boundary and context distillation (BCD) method within the KD framework for EDIPs, which facilitates the targeted knowledge transfer from large accurate teacher models to compact efficient student models. Specifically, the boundary distillation component focuses on extracting explicit object-level semantic boundaries from the hierarchical feature maps of the backbone network to enhance the student model's mask quality in boundary regions. Concurrently, the context distillation component leverages self-relations as a bridge to transfer implicit pixel-level contexts from the teacher model to the student model, ensuring strong connectivity in target regions. Our proposed BCD method is specifically designed for EDIP tasks and is characterized by its simplicity and efficiency. Extensive experimental results across semantic segmentation, object detection, and instance segmentation on various representative datasets demonstrate that our method can outperform existing methods without requiring extra supervisions or incurring increased inference costs, resulting in well-defined object boundaries and smooth connecting regions.

논문 ID: document/1030589

Applying Knowledge Distillation to Improve Weed Mapping With Drones

In precision agriculture, non-invasive remote sensing using UAVs can be employed to observe crops in visible and non-visible spectra. This paper investigates the effectiveness of state-of-the-art knowledge distillation techniques for mapping weeds with drones, an essential component of precision agriculture that employs remote sensing to monitor crops and weeds. The study introduces a lightweight Vision Transformer-based model that achieves optimal weed mapping capabilities while maintaining minimal computation time. The research shows that the student model effectively learns from the teacher model using the WeedMap dataset, achieving accurate results suitable for mobile platforms such as drones, with only 0.5 GMacs compared to 42.5 GMacs of the teacher model. The trained models obtained an F1 score of 0.863 and 0.631 on two data subsets, with a performance improvement of 2 and 7 points, respectively, over the undistilled model. The study results suggest that developing efficient computer vision algorithms on drones can significantly improve agricultural management practices, leading to greater profitability and environmental sustainability.

arXiv:2401.12472공식 코드 · Papers with Code

Contrastive Learning in Distilled Models

Natural Language Processing models like BERT can provide state-of-the-art word embeddings for downstream NLP tasks. However, these models yet to perform well on Semantic Textual Similarity, and may be too large to be deployed as lightweight edge applications. We seek to apply a suitable contrastive learning method based on the SimCSE paper, to a model architecture adapted from a knowledge distillation based model, DistilBERT, to address these two issues. Our final lightweight model DistilFace achieves an average of 72.1 in Spearman's correlation on STS tasks, a 34.2 percent improvement over BERT base.

arXiv:2401.12624

Knowledge Distillation from Language-Oriented to Emergent Communication for Multi-Agent Remote Control

In this work, we compare emergent communication (EC) built upon multi-agent deep reinforcement learning (MADRL) and language-oriented semantic communication (LSC) empowered by a pre-trained large language model (LLM) using human language. In a multi-agent remote navigation task, with multimodal input data comprising location and channel maps, it is shown that EC incurs high training cost and struggles when using multimodal data, whereas LSC yields high inference computing cost due to the LLM's large size. To address their respective bottlenecks, we propose a novel framework of language-guided EC (LEC) by guiding the EC training using LSC via knowledge distillation (KD). Simulations corroborate that LEC achieves faster travel time while avoiding areas with poor channel conditions, as well as speeding up the MADRL training convergence by up to 61.8% compared to EC.

arXiv:2401.12433

A Novel Garment Transfer Method Supervised by Distilled Knowledge of Virtual Try-on Model

This paper proposes a novel garment transfer method supervised with knowledge distillation from virtual try-on. Our method first reasons the transfer parsing to provide shape prior to downstream tasks. We employ a multi-phase teaching strategy to supervise the training of the transfer parsing reasoning model, learning the response and feature knowledge from the try-on parsing reasoning model. To correct the teaching error, it transfers the garment back to its owner to absorb the hard knowledge in the self-study phase. Guided by the transfer parsing, we adjust the position of the transferred garment via STN to prevent distortion. Afterward, we estimate a progressive flow to precisely warp the garment with shape and content correspondences. To ensure warping rationality, we supervise the training of the garment warping model using target shape and warping knowledge from virtual try-on. To better preserve body features in the transfer result, we propose a well-designed training strategy for the arm regrowth task to infer new exposure skin. Experiments demonstrate that our method has state-of-the-art performance compared with other virtual try-on and garment transfer methods in garment transfer, especially for preserving garment texture and body features.

arXiv:2401.11700

Keep Decoding Parallel with Effective Knowledge Distillation from Language Models to End-to-end Speech Recognisers

This study presents a novel approach for knowledge distillation (KD) from a BERT teacher model to an automatic speech recognition (ASR) model using intermediate layers. To distil the teacher's knowledge, we use an attention decoder that learns from BERT's token probabilities. Our method shows that language model (LM) information can be more effectively distilled into an ASR model using both the intermediate layers and the final layer. By using the intermediate layers as distillation target, we can more effectively distil LM knowledge into the lower network layers. Using our method, we achieve better recognition accuracy than with shallow fusion of an external LM, allowing us to maintain fast parallel decoding. Experiments on the LibriSpeech dataset demonstrate the effectiveness of our approach in enhancing greedy decoding with connectionist temporal classification (CTC).

arXiv:2401.11365

Confidence Preservation Property in Knowledge Distillation Abstractions

Social media platforms prevent malicious activities by detecting harmful content of posts and comments. To that end, they employ large-scale deep neural network language models for sentiment analysis and content understanding. Some models, like BERT, are complex, and have numerous parameters, which makes them expensive to operate and maintain. To overcome these deficiencies, industry experts employ a knowledge distillation compression technique, where a distilled model is trained to reproduce the classification behavior of the original model. The distillation processes terminates when the distillation loss function reaches the stopping criteria. This function is mainly designed to ensure that the original and the distilled models exhibit alike classification behaviors. However, besides classification accuracy, there are additional properties of the original model that the distilled model should preserve to be considered as an appropriate abstraction. In this work, we explore whether distilled TinyBERT models preserve confidence values of the original BERT models, and investigate how this confidence preservation property could guide tuning hyperparameters of the distillation process.

arXiv:2401.12019

Stereo-Matching Knowledge Distilled Monocular Depth Estimation Filtered by Multiple Disparity Consistency

In stereo-matching knowledge distillation methods of the self-supervised monocular depth estimation, the stereo-matching network's knowledge is distilled into a monocular depth network through pseudo-depth maps. In these methods, the learning-based stereo-confidence network is generally utilized to identify errors in the pseudo-depth maps to prevent transferring the errors. However, the learning-based stereo-confidence networks should be trained with ground truth (GT), which is not feasible in a self-supervised setting. In this paper, we propose a method to identify and filter errors in the pseudo-depth map using multiple disparity maps by checking their consistency without the need for GT and a training process. Experimental results show that the proposed method outperforms the previous methods and works well on various configurations by filtering out erroneous areas where the stereo-matching is vulnerable, especially such as textureless regions, occlusion boundaries, and reflective surfaces.