Resistive random access memory (ReRAM)-based processing-in-memory (PIM) architectures have demonstrated great potential to accelerate Deep Neural Network (DNN) training/inference. However, the computational accuracy of analog PIM is compromised due to the non-idealities, such as the conductance variation of ReRAM cells. The impact of these non-idealities worsens as the number of concurrently activated wordlines and bitlines increases. To guarantee computational accuracy, only a limited number of wordlines and bitlines of the crossbar array can be turned on concurrently, significantly reducing the achievable parallelism of the architecture. While the constraints on parallelism limit the efficiency of the accelerators, they also provide a new opportunity for fine-grained mixed-precision quantization. To enable efficient DNN inference on practical ReRAM-based accelerators, we propose an algorithm-architecture co-design framework called \underline{B}lock-\underline{W}ise mixed-precision \underline{Q}uantization (BWQ). At the algorithm level, BWQ-A introduces a mixed-precision quantization scheme at the block level, which achieves a high weight and activation compression ratio with negligible accuracy degradation. We also present the hardware architecture design BWQ-H, which leverages the low-bit-width models achieved by BWQ-A to perform high-efficiency DNN inference on ReRAM devices. BWQ-H also adopts a novel precision-aware weight mapping method to increase the ReRAM crossbar's throughput. Our evaluation demonstrates the effectiveness of BWQ, which achieves a 6.08x speedup and a 17.47x energy saving on average compared to existing ReRAM-based architectures.
相關內容
Entity alignment seeks identical entities in different knowledge graphs, which is a long-standing task in the database research. Recent work leverages deep learning to embed entities in vector space and align them via nearest neighbor search. Although embedding-based entity alignment has gained marked success in recent years, it lacks explanations for alignment decisions. In this paper, we present the first framework that can generate explanations for understanding and repairing embedding-based entity alignment results. Given an entity alignment pair produced by an embedding model, we first compare its neighbor entities and relations to build a matching subgraph as a local explanation. We then construct an alignment dependency graph to understand the pair from an abstract perspective. Finally, we repair the pair by resolving three types of alignment conflicts based on dependency graphs. Experiments on five datasets demonstrate the effectiveness and generalization of our framework in explaining and repairing embedding-based entity alignment results.
Protecting the copyright of large language models (LLMs) has become crucial due to their resource-intensive training and accompanying carefully designed licenses. However, identifying the original base model of an LLM is challenging due to potential parameter alterations through fine-tuning or continued pretraining. In this study, we introduce HuRef, a human-readable fingerprint for LLMs that uniquely identifies the base model without exposing model parameters or interfering with training. We first observe that the vector direction of LLM parameters remains stable after the model has converged during pretraining, showing negligible perturbations through subsequent training steps, including continued pretraining, supervised fine-tuning (SFT), and RLHF, which makes it a sufficient condition to identify the base model. The necessity is validated by continuing to train an LLM with an extra term to drive away the model parameters' direction and the model becomes damaged. However, this direction is vulnerable to simple attacks like dimension permutation or matrix rotation, which significantly change it without affecting performance. To address this, leveraging the Transformer structure, we systematically analyze potential attacks and define three invariant terms that identify an LLM's base model. We make these invariant terms human-readable by mapping them to a Gaussian vector using a convolutional encoder and then converting it into a natural image with StyleGAN2. Our method generates a dog image as an identity fingerprint for an LLM, where the dog's appearance strongly indicates the LLM's base model. Experimental results across various LLMs demonstrate the effectiveness of our method, the generated dog image remains invariant to different training steps, including SFT, RLHF, or even continued pretraining with augmented vocabulary in a new language.
Recent work has shown that energy-based language modeling is an effective framework for controllable text generation because it enables flexible integration of arbitrary discriminators. However, because energy-based LMs are globally normalized, approximate techniques like Metropolis-Hastings (MH) are required for inference. Past work has largely explored simple proposal distributions that modify a single token at a time, like in Gibbs sampling. In this paper, we develop a novel MH sampler that, in contrast, proposes re-writes of the entire sequence in each step via iterative prompting of a large language model. Our new sampler (a) allows for more efficient and accurate sampling from a target distribution and (b) allows generation length to be determined through the sampling procedure rather than fixed in advance, as past work has required. We perform experiments on two controlled generation tasks, showing both downstream performance gains and more accurate target distribution sampling in comparison with single-token proposal techniques.
In recent years, knowledge distillation methods based on contrastive learning have achieved promising results on image classification and object detection tasks. However, in this line of research, we note that less attention is paid to semantic segmentation. Existing methods heavily rely on data augmentation and memory buffer, which entail high computational resource demands when applying them to handle semantic segmentation that requires to preserve high-resolution feature maps for making dense pixel-wise predictions. In order to address this problem, we present Augmentation-free Dense Contrastive Knowledge Distillation (Af-DCD), a new contrastive distillation learning paradigm to train compact and accurate deep neural networks for semantic segmentation applications. Af-DCD leverages a masked feature mimicking strategy, and formulates a novel contrastive learning loss via taking advantage of tactful feature partitions across both channel and spatial dimensions, allowing to effectively transfer dense and structured local knowledge learnt by the teacher model to a target student model while maintaining training efficiency. Extensive experiments on five mainstream benchmarks with various teacher-student network pairs demonstrate the effectiveness of our approach. For instance, the DeepLabV3-Res18|DeepLabV3-MBV2 model trained by Af-DCD reaches 77.03%|76.38% mIOU on Cityscapes dataset when choosing DeepLabV3-Res101 as the teacher, setting new performance records. Besides that, Af-DCD achieves an absolute mIOU improvement of 3.26%|3.04%|2.75%|2.30%|1.42% compared with individually trained counterpart on Cityscapes|Pascal VOC|Camvid|ADE20K|COCO-Stuff-164K. Code is available at //github.com/OSVAI/Af-DCD
Entity resolution (ER) is an important data integration task with a wide spectrum of applications. The state-of-the-art solutions on ER rely on pre-trained language models (PLMs), which require fine-tuning on a lot of labeled matching/non-matching entity pairs. Recently, large languages models (LLMs), such as GPT-4, have shown the ability to perform many tasks without tuning model parameters, which is known as in-context learning (ICL) that facilitates effective learning from a few labeled input context demonstrations. However, existing ICL approaches to ER typically necessitate providing a task description and a set of demonstrations for each entity pair and thus have limitations on the monetary cost of interfacing LLMs. To address the problem, in this paper, we provide a comprehensive study to investigate how to develop a cost-effective batch prompting approach to ER. We introduce a framework BATCHER consisting of demonstration selection and question batching and explore different design choices that support batch prompting for ER. We also devise a covering-based demonstration selection strategy that achieves an effective balance between matching accuracy and monetary cost. We conduct a thorough evaluation to explore the design space and evaluate our proposed strategies. Through extensive experiments, we find that batch prompting is very cost-effective for ER, compared with not only PLM-based methods fine-tuned with extensive labeled data but also LLM-based methods with manually designed prompting. We also provide guidance for selecting appropriate design choices for batch prompting.
As Internet censors rapidly evolve new blocking techniques, circumvention tools must also adapt and roll out new strategies to remain unblocked. But new strategies can be time consuming for circumventors to develop and deploy, and usually an update to one tool often requires significant additional effort to be ported to others. Moreover, distributing the updated application across different platforms poses its own set of challenges. In this paper, we introduce $\textit{WATER}$ (WebAssembly Transport Executables Runtime), a novel design that enables applications to use a WebAssembly-based application-layer to wrap network transports (e.g., TLS). Deploying a new circumvention technique with $\textit{WATER}$ only requires distributing the WebAssembly Transport Module(WATM) binary and any transport-specific configuration, allowing dynamic transport updates without any change to the application itself. WATMs are also designed to be generic such that different applications using $\textit{WATER}$ can use the same WATM to rapidly deploy successful circumvention techniques to their own users, facilitating rapid interoperability between independent circumvention tools.
Although advancements in machine learning have driven the development of malicious URL detection technology, current techniques still face significant challenges in their capacity to generalize and their resilience against evolving threats. In this paper, we propose PyraTrans, a novel method that integrates pretrained Transformers with pyramid feature learning to detect malicious URL. PyraTrans utilizes a pretrained CharBERT as its foundation and is augmented with three interconnected feature modules: 1) Encoder Feature Extraction, extracting multi-order feature matrices from each CharBERT encoder layer; 2) Multi-Scale Feature Learning, capturing local contextual insights at various scales and aggregating information across encoder layers; and 3) Spatial Pyramid Attention, focusing on regional-level attention to emphasize areas rich in expressive information. The proposed approach addresses the limitations of the Transformer in local feature learning and regional relational awareness, which are vital for capturing URL-specific word patterns, character combinations, or structural anomalies. In several challenging experimental scenarios, the proposed method has shown significant improvements in accuracy, generalization, and robustness in malicious URL detection. For instance, it achieved a peak F1-score improvement of 40% in class-imbalanced scenarios, and exceeded the best baseline result by 14.13% in accuracy in adversarial attack scenarios. Additionally, we conduct a case study where our method accurately identifies all 30 active malicious web pages, whereas two pior SOTA methods miss 4 and 7 malicious web pages respectively. Codes and data are available at://github.com/Alixyvtte/PyraTrans.
Graph convolutional networks (GCNs) have been widely used and achieved remarkable results in skeleton-based action recognition. We think the key to skeleton-based action recognition is a skeleton hanging in frames, so we focus on how the Graph Convolutional Convolution networks learn different topologies and effectively aggregate joint features in the global temporal and local temporal. In this work, we propose three Channel-wise Tolopogy Graph Convolution based on Channel-wise Topology Refinement Graph Convolution (CTR-GCN). Combining CTR-GCN with two joint cross-attention modules can capture the upper-lower body part and hand-foot relationship skeleton features. After that, to capture features of human skeletons changing in frames we design the Temporal Attention Transformers to extract skeletons effectively. The Temporal Attention Transformers can learn the temporal features of human skeleton sequences. Finally, we fuse the temporal features output scale with MLP and classification. We develop a powerful graph convolutional network named Spatial Temporal Effective Body-part Cross Attention Transformer which notably high-performance on the NTU RGB+D, NTU RGB+D 120 datasets. Our code and models are available at //github.com/maclong01/STEP-CATFormer
Transformer architectures have facilitated the development of large-scale and general-purpose sequence models for prediction tasks in natural language processing and computer vision, e.g., GPT-3 and Swin Transformer. Although originally designed for prediction problems, it is natural to inquire about their suitability for sequential decision-making and reinforcement learning problems, which are typically beset by long-standing issues involving sample efficiency, credit assignment, and partial observability. In recent years, sequence models, especially the Transformer, have attracted increasing interest in the RL communities, spawning numerous approaches with notable effectiveness and generalizability. This survey presents a comprehensive overview of recent works aimed at solving sequential decision-making tasks with sequence models such as the Transformer, by discussing the connection between sequential decision-making and sequence modeling, and categorizing them based on the way they utilize the Transformer. Moreover, this paper puts forth various potential avenues for future research intending to improve the effectiveness of large sequence models for sequential decision-making, encompassing theoretical foundations, network architectures, algorithms, and efficient training systems. As this article has been accepted by the Frontiers of Computer Science, here is an early version, and the most up-to-date version can be found at //journal.hep.com.cn/fcs/EN/10.1007/s11704-023-2689-5
Learning with limited data is a key challenge for visual recognition. Few-shot learning methods address this challenge by learning an instance embedding function from seen classes and apply the function to instances from unseen classes with limited labels. This style of transfer learning is task-agnostic: the embedding function is not learned optimally discriminative with respect to the unseen classes, where discerning among them is the target task. In this paper, we propose a novel approach to adapt the embedding model to the target classification task, yielding embeddings that are task-specific and are discriminative. To this end, we employ a type of self-attention mechanism called Transformer to transform the embeddings from task-agnostic to task-specific by focusing on relating instances from the test instances to the training instances in both seen and unseen classes. Our approach also extends to both transductive and generalized few-shot classification, two important settings that have essential use cases. We verify the effectiveness of our model on two standard benchmark few-shot classification datasets --- MiniImageNet and CUB, where our approach demonstrates state-of-the-art empirical performance.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191