Generative tasks, such as text generation and question answering, hold a crucial position in the realm of mobile applications. Due to their sensitivity to privacy concerns, there is a growing demand for their execution directly on mobile devices. Currently, the execution of these generative tasks heavily depends on Large Language Models (LLMs). Nevertheless, the limited memory capacity of these devices presents a formidable challenge to the scalability of such models. In our research, we introduce LLMCad, an innovative on-device inference engine specifically designed for efficient generative Natural Language Processing (NLP) tasks. The core idea behind LLMCad revolves around model collaboration: a compact LLM, residing in memory, takes charge of generating the most straightforward tokens, while a high-precision LLM steps in to validate these tokens and rectify any identified errors. LLMCad incorporates three novel techniques: (1) Instead of generating candidate tokens in a sequential manner, LLMCad employs the smaller LLM to construct a token tree, encompassing a wider range of plausible token pathways. Subsequently, the larger LLM can efficiently validate all of these pathways simultaneously. (2) It employs a self-adjusting fallback strategy, swiftly initiating the verification process whenever the smaller LLM generates an erroneous token. (3) To ensure a continuous flow of token generation, LLMCad speculatively generates tokens during the verification process by implementing a compute-IO pipeline. Through an extensive series of experiments, LLMCad showcases an impressive token generation speed, achieving rates up to 9.3x faster than existing inference engines.
相關內容
Graph Neural Networks (GNNs) have proven to be quite versatile for a variety of applications, including recommendation systems, fake news detection, drug discovery, and even computer vision. Due to the expanding size of graph-structured data, GNN models have also increased in complexity, leading to substantial latency issues. This is primarily attributed to the irregular structure of graph data and its access pattern into memory. The natural solution to reduce latency is to compress large GNNs into small GNNs. One way to do this is via knowledge distillation (KD). However, most KD approaches for GNNs only consider the outputs of the last layers and do not consider the outputs of the intermediate layers of the GNNs; these layers may contain important inductive biases indicated by the graph structure. To address this shortcoming, we propose a novel KD approach to GNN compression that we call Attention-Based Knowledge Distillation (ABKD). ABKD is a KD approach that uses attention to identify important intermediate teacher-student layer pairs and focuses on aligning their outputs. ABKD enables higher compression of GNNs with a smaller accuracy dropoff compared to existing KD approaches. On average, we achieve a 1.79% increase in accuracy with a 32.3x compression ratio on OGBN-Mag, a large graph dataset, compared to state-of-the-art approaches.
In the absence of readily available labeled data for a given sequence labeling task and language, annotation projection has been proposed as one of the possible strategies to automatically generate annotated data. Annotation projection has often been formulated as the task of transporting, on parallel corpora, the labels pertaining to a given span in the source language into its corresponding span in the target language. In this paper we present T-Projection, a novel approach for annotation projection that leverages large pretrained text-to-text language models and state-of-the-art machine translation technology. T-Projection decomposes the label projection task into two subtasks: (i) A candidate generation step, in which a set of projection candidates using a multilingual T5 model is generated and, (ii) a candidate selection step, in which the generated candidates are ranked based on translation probabilities. We conducted experiments on intrinsic and extrinsic tasks in 5 Indo-European and 8 low-resource African languages. We demostrate that T-projection outperforms previous annotation projection methods by a wide margin. We believe that T-Projection can help to automatically alleviate the lack of high-quality training data for sequence labeling tasks. Code and data are publicly available.
Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages to instruction-tuned LLMs can result in catastrophic forgetting, which leads to the loss of multitasking ability. To address this issue, we propose InstructAlign which uses continual crosslingual instruction tuning to enable LLMs to align new unseen languages with previously learned high-resource languages. Our results demonstrate the effectiveness of InstructAlign in enabling the model to understand low-resource languages with limited parallel data while preventing catastrophic forgetting. Our work contributes to the advancement of language adaptation methods, particularly for adapting instruction-tuned LLMs to underrepresented languages. Our code is released on //github.com/HLTCHKUST/InstructAlign
Despite outstanding performance in many tasks, language models are notoriously inclined to make factual errors in tasks requiring arithmetic computation. We address this deficiency by creating Calc-X, a collection of datasets that demonstrates the appropriate use of a calculator in reasoning chains. Calc-X is suitable for teaching language models to offload computations to a symbolic system. We survey and unify several existing chain-of-thought datasets into a proposed format, resulting in a standard collection of over 300,000 samples requiring arithmetic reasoning. Finally, we use the new Calc-X collection to train open-source calculator-using models we call Calcformers and show that these models approximately double the accuracy of generating correct results compared to vanilla language model baselines. We make all Calc-X datasets, source code and Calcformers models publicly available.
Road user trajectory prediction in dynamic environments is a challenging but crucial task for various applications, such as autonomous driving. One of the main challenges in this domain is the multimodal nature of future trajectories stemming from the unknown yet diverse intentions of the agents. Diffusion models have shown to be very effective in capturing such stochasticity in prediction tasks. However, these models involve many computationally expensive denoising steps and sampling operations that make them a less desirable option for real-time safety-critical applications. To this end, we present a novel framework that leverages diffusion models for predicting future trajectories in a computationally efficient manner. To minimize the computational bottlenecks in iterative sampling, we employ an efficient sampling mechanism that allows us to maximize the number of sampled trajectories for improved accuracy while maintaining inference time in real time. Moreover, we propose a scoring mechanism to select the most plausible trajectories by assigning relative ranks. We show the effectiveness of our approach by conducting empirical evaluations on common pedestrian (UCY/ETH) and autonomous driving (nuScenes) benchmark datasets on which our model achieves state-of-the-art performance on several subsets and metrics.
Large language models (LLMs), such as ChatGPT, have simplified text generation tasks, yet their inherent privacy risks are increasingly garnering attention. While differential privacy techniques have been successfully applied to text classification tasks, the resultant semantic bias makes them unsuitable for text generation. Homomorphic encryption inference methods have also been introduced. However, the significant computational and communication costs limit their viability. Furthermore, closed-source, black-box models such as GPT-4 withhold their architecture, thwarting certain privacy-enhancing strategies such as splitting inference into local and remote and then adding noise when communicating. To overcome these challenges, we introduce PrivInfer, the first practical privacy-preserving inference framework for black-box LLMs in text generation. PrivInfer employs differential privacy methods to generate perturbed prompts for remote LLMs inference and extracts the meaningful response from the remote perturbed results. We also introduce RANTEXT, a differential privacy mechanism within the perturbation module of PrivInfer specifically for LLMs that leverages random adjacency in text perturbations. Experimental results indicate that PrivInfer is comparable to GPT-4 in terms of text generation quality while protecting privacy, and RANTEXT provides enhanced privacy protection against three types of differential privacy attacks, including our newly introduced GPT inference attack, compared to baseline methods.
Quantum networks serve as the means to transmit information, encoded in quantum bits or qubits, between quantum processors that are physically separated. Given the instability of qubits, the design of such networks is challenging, necessitating a careful balance between reliability and efficiency. Typically, quantum networks fall into two categories: those utilize quantum entanglements for quantum teleportation, and those directly transfer quantum message. In this paper, we present SurfaceNet, a quantum network in the second category that employs surface codes as logical qubits for preserving and transferring message. Our approach of using surface codes can fault-tolerantly correct both operational and photon loss errors within the network. We propose a novel one-way quantum communication procedure, designed to better integrate surface codes into our network architecture. We also propose an efficient routing protocol that optimizes resource utilization for our communication procedure. Simulation results demonstrate that SurfaceNet significantly enhances the overall communication fidelity.
Question answering over hybrid contexts is a complex task, which requires the combination of information extracted from unstructured texts and structured tables in various ways. Recently, In-Context Learning demonstrated significant performance advances for reasoning tasks. In this paradigm, a large language model performs predictions based on a small set of supporting exemplars. The performance of In-Context Learning depends heavily on the selection procedure of the supporting exemplars, particularly in the case of HybridQA, where considering the diversity of reasoning chains and the large size of the hybrid contexts becomes crucial. In this work, we present Selection of ExEmplars for hybrid Reasoning (SEER), a novel method for selecting a set of exemplars that is both representative and diverse. The key novelty of SEER is that it formulates exemplar selection as a Knapsack Integer Linear Program. The Knapsack framework provides the flexibility to incorporate diversity constraints that prioritize exemplars with desirable attributes, and capacity constraints that ensure that the prompt size respects the provided capacity budgets. The effectiveness of SEER is demonstrated on FinQA and TAT-QA, two real-world benchmarks for HybridQA, where it outperforms previous exemplar selection methods.
Deep neural models in recent years have been successful in almost every field, including extremely complex problem statements. However, these models are huge in size, with millions (and even billions) of parameters, thus demanding more heavy computation power and failing to be deployed on edge devices. Besides, the performance boost is highly dependent on redundant labeled data. To achieve faster speeds and to handle the problems caused by the lack of data, knowledge distillation (KD) has been proposed to transfer information learned from one model to another. KD is often characterized by the so-called `Student-Teacher' (S-T) learning framework and has been broadly applied in model compression and knowledge transfer. This paper is about KD and S-T learning, which are being actively studied in recent years. First, we aim to provide explanations of what KD is and how/why it works. Then, we provide a comprehensive survey on the recent progress of KD methods together with S-T frameworks typically for vision tasks. In general, we consider some fundamental questions that have been driving this research area and thoroughly generalize the research progress and technical details. Additionally, we systematically analyze the research status of KD in vision applications. Finally, we discuss the potentials and open challenges of existing methods and prospect the future directions of KD and S-T learning.
Many tasks in natural language processing can be viewed as multi-label classification problems. However, most of the existing models are trained with the standard cross-entropy loss function and use a fixed prediction policy (e.g., a threshold of 0.5) for all the labels, which completely ignores the complexity and dependencies among different labels. In this paper, we propose a meta-learning method to capture these complex label dependencies. More specifically, our method utilizes a meta-learner to jointly learn the training policies and prediction policies for different labels. The training policies are then used to train the classifier with the cross-entropy loss function, and the prediction policies are further implemented for prediction. Experimental results on fine-grained entity typing and text classification demonstrate that our proposed method can obtain more accurate multi-label classification results.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191