Large language models (LLMs) propel the prosperity of interactive AI applications showcased by ChatGPT that demand timely response of inference services. However, LLM inference is computation intensive and memory intensive, and improper parameter configuration at LLM platforms may exacerbate the inference time. In this paper, we analyze the impact of LLM output token distribution on the inference queueing delay, where the max-token clipping and the batched inference are considered. By formulating an M/G/1 model, we observe that enforcing a maximum output token limit on a very small fraction of inference requests can significantly reduce the queueing delay, and our model facilitates the selection of the optimal limit. For the batch inference, we model the service process as a bulk queue in which the batch processing time is affected by the batch size and the maximum token size inside this batch jointly. The queueing delays of the batching of all buffered requests (dynamic batching), the batching of constant number of requests (fixed batching), and the batching without intra-batch waiting (elastic batching) are derived. Experimental results show that our mathematical models coincide with the event-driven simulations well.
相關內容
Large language models (LLMs) have made remarkable progress in various natural language processing tasks as a benefit of their capability to comprehend and reason with factual knowledge. However, a significant amount of factual knowledge is stored in structured data, which possesses unique characteristics that differ from the unstructured texts used for pretraining. This difference can introduce imperceptible inference parameter deviations, posing challenges for LLMs in effectively utilizing and reasoning with structured data to accurately infer factual knowledge. To this end, we propose a benchmark named StructFact, to evaluate the structural reasoning capabilities of LLMs in inferring factual knowledge. StructFact comprises 8,340 factual questions encompassing various tasks, domains, timelines, and regions. This benchmark allows us to investigate the capability of LLMs across five factual tasks derived from the unique characteristics of structural facts. Extensive experiments on a set of LLMs with different training strategies reveal the limitations of current LLMs in inferring factual knowledge from structured data. We present this benchmark as a compass to navigate the strengths and weaknesses of LLMs in reasoning with structured data for knowledge-sensitive tasks, and to encourage advancements in related real-world applications. Please find our code at //github.com/EganGu/StructFact.
Vision-language models (VLMs) like CLIP have demonstrated impressive zero-shot ability in image classification tasks by aligning text and images but suffer inferior performance compared with task-specific expert models. On the contrary, expert models excel in their specialized domains but lack zero-shot ability for new tasks. How to obtain both the high performance of expert models and zero-shot ability is an important research direction. In this paper, we attempt to demonstrate that by constructing a model hub and aligning models with their functionalities using model labels, new tasks can be solved in a zero-shot manner by effectively selecting and reusing models in the hub. We introduce a novel paradigm, Model Label Learning (MLL), which bridges the gap between models and their functionalities through a Semantic Directed Acyclic Graph (SDAG) and leverages an algorithm, Classification Head Combination Optimization (CHCO), to select capable models for new tasks. Compared with the foundation model paradigm, it is less costly and more scalable, i.e., the zero-shot ability grows with the sizes of the model hub. Experiments on seven real-world datasets validate the effectiveness and efficiency of MLL, demonstrating that expert models can be effectively reused for zero-shot tasks. Our code will be released publicly.
Understanding emotions is a fundamental aspect of human communication. Integrating audio and video signals offers a more comprehensive understanding of emotional states compared to traditional methods that rely on a single data source, such as speech or facial expressions. Despite its potential, multimodal emotion recognition faces significant challenges, particularly in synchronization, feature extraction, and fusion of diverse data sources. To address these issues, this paper introduces a novel transformer-based model named Audio-Video Transformer Fusion with Cross Attention (AVT-CA). The AVT-CA model employs a transformer fusion approach to effectively capture and synchronize interlinked features from both audio and video inputs, thereby resolving synchronization problems. Additionally, the Cross Attention mechanism within AVT-CA selectively extracts and emphasizes critical features while discarding irrelevant ones from both modalities, addressing feature extraction and fusion challenges. Extensive experimental analysis conducted on the CMU-MOSEI, RAVDESS and CREMA-D datasets demonstrates the efficacy of the proposed model. The results underscore the importance of AVT-CA in developing precise and reliable multimodal emotion recognition systems for practical applications.
Multimodal large language models (MLLMs) demand considerable computations for inference due to the extensive parameters and the additional input tokens needed for visual information representation. Herein, we introduce Visual Tokens Withdrawal (VTW), a plug-and-play module to boost MLLMs for rapid inference. Our approach is inspired by two intriguing phenomena we have observed: (1) the attention sink phenomenon that is prevalent in LLMs also persists in MLLMs, suggesting that initial tokens and nearest tokens receive the majority of attention, while middle vision tokens garner minimal attention in deep layers; (2) the presence of information migration, which implies that visual information is transferred to subsequent text tokens within the first few layers of MLLMs. As per our findings, we conclude that vision tokens are unnecessary in the deep layers of MLLMs. Thus, we strategically withdraw them at a certain layer, enabling only text tokens to engage in subsequent layers. To pinpoint the ideal layer for VTW, we initially analyze a limited set of tiny datasets and choose the first layer that meets the Kullback-Leibler divergence criterion. Our VTW approach can cut computational overhead by over 40\% across diverse multimodal tasks while maintaining performance. Our code is released at \url{//github.com/lzhxmu/VTW}.
The recent trend in the Large Vision and Language model has brought a new change in how information extraction systems are built. VLMs have set a new benchmark with their State-of-the-art techniques in understanding documents and building question-answering systems across various industries. They are significantly better at generating text from document images and providing accurate answers to questions. However, there are still some challenges in effectively utilizing these models to build a precise conversational system. General prompting techniques used with large language models are often not suitable for these specially designed vision language models. The output generated by such generic input prompts is ordinary and may contain information gaps when compared with the actual content of the document. To obtain more accurate and specific answers, a well-targeted prompt is required by the vision language model, along with the document image. In this paper, a technique is discussed called Target prompting, which focuses on explicitly targeting parts of document images and generating related answers from those specific regions only. The paper also covers the evaluation of response for each prompting technique using different user queries and input prompts.
Piecewise deterministic Markov processes (PDMPs) are a class of continuous-time Markov processes that were recently used to develop a new class of Markov chain Monte Carlo algorithms. However, the implementation of the processes is challenging due to the continuous-time aspect and the necessity of integrating the rate function. Recently, Corbella, Spencer, and Roberts (2022) proposed a new algorithm to automate the implementation of the Zig-Zag sampler. However, the efficiency of the algorithm highly depends on a hyperparameter ($t_{\text{max}}$) that is fixed all along the run of the algorithm and needs preliminary runs to tune. In this work, we relax this assumption and propose a new variant of their algorithm that let this parameter change over time and automatically adapt to the target distribution. We also replace the Brent optimization algorithm by a grid-based method to compute the upper bound of the rate function. This method is more robust to the regularity of the function and gives a tighter upper bound while being quicker to compute. We also extend the algorithm to other PDMPs and provide a Python implementation of the algorithm based on JAX.
The ever-growing popularity of large language models (LLMs) has resulted in their increasing adoption for hardware design and verification. Prior research has attempted to assess the capability of LLMs to automate digital hardware design by producing superior-quality Register Transfer Logic (RTL) descriptions, particularly in Verilog. However, these tests have revealed that Verilog code production using LLMs at current state-of-the-art lack sufficient functional correctness to be practically viable, compared to automatic generation of programs in general-purpose programming languages such as C, C++, Python, etc. With this as the key insight, in this paper we assess the performance of a two-stage software pipeline for automated Verilog RTL generation: LLM based automatic generation of annotated C++ code suitable for high-level synthesis (HLS), followed by HLS to generate Verilog RTL. We have benchmarked the performance of our proposed scheme using the open-source VerilogEval dataset, for four different industry-scale LLMs, and the Vitis HLS tool. Our experimental results demonstrate that our two-step technique substantially outperforms previous proposed techniques of direct Verilog RTL generation by LLMs in terms of average functional correctness rates, reaching score of 0.86 in pass@1 metric.
Posterior sampling is a task of central importance in Bayesian inference. For many applications in Bayesian meta-analysis and Bayesian transfer learning, the prior distribution is unknown and needs to be estimated from samples. In practice, the prior distribution can be high-dimensional, adding to the difficulty of efficient posterior inference. In this paper, we propose a novel Markov chain Monte Carlo algorithm, which we term graph-enabled MCMC, for posterior sampling with unknown and potentially high-dimensional prior distributions. The algorithm is based on constructing a geometric graph from prior samples and subsequently uses the graph structure to guide the transition of the Markov chain. Through extensive theoretical and numerical studies, we demonstrate that our graph-enabled MCMC algorithm provides reliable approximation to the posterior distribution and is highly computationally efficient.
Many studies have demonstrated that large language models (LLMs) can produce harmful responses, exposing users to unexpected risks when LLMs are deployed. Previous studies have proposed comprehensive taxonomies of the risks posed by LLMs, as well as corresponding prompts that can be used to examine the safety mechanisms of LLMs. However, the focus has been almost exclusively on English, and little has been explored for other languages. Here we aim to bridge this gap. We first introduce a dataset for the safety evaluation of Chinese LLMs, and then extend it to two other scenarios that can be used to better identify false negative and false positive examples in terms of risky prompt rejections. We further present a set of fine-grained safety assessment criteria for each risk type, facilitating both manual annotation and automatic evaluation in terms of LLM response harmfulness. Our experiments on five LLMs show that region-specific risks are the prevalent type of risk, presenting the major issue with all Chinese LLMs we experimented with. Our data is available at //github.com/Libr-AI/do-not-answer. Warning: this paper contains example data that may be offensive, harmful, or biased.
Translational distance-based knowledge graph embedding has shown progressive improvements on the link prediction task, from TransE to the latest state-of-the-art RotatE. However, N-1, 1-N and N-N predictions still remain challenging. In this work, we propose a novel translational distance-based approach for knowledge graph link prediction. The proposed method includes two-folds, first we extend the RotatE from 2D complex domain to high dimension space with orthogonal transforms to model relations for better modeling capacity. Second, the graph context is explicitly modeled via two directed context representations. These context representations are used as part of the distance scoring function to measure the plausibility of the triples during training and inference. The proposed approach effectively improves prediction accuracy on the difficult N-1, 1-N and N-N cases for knowledge graph link prediction task. The experimental results show that it achieves better performance on two benchmark data sets compared to the baseline RotatE, especially on data set (FB15k-237) with many high in-degree connection nodes.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191