Chunjie Luo

Zhengxin Yang, Wanling Gao, Chunjie Luo et al.

Machine learning inference should be subject to stringent inference time constraints while ensuring high inference quality, especially in safety-critical (e.g., autonomous driving) and mission-critical (e.g., emotion recognition) contexts. Neglecting either aspect can lead to severe consequences, such as loss of life and property damage. Many studies lack a comprehensive consideration of these metrics, leading to incomplete or misleading evaluations. The study unveils a counterintuitive revelation: deep learning inference quality exhibits fluctuations due to inference time. To depict this phenomenon, the authors coin a new term, "tail quality," providing a more comprehensive evaluation, and overcoming conventional metric limitations. Moreover, the research proposes an initial evaluation framework to analyze factors affecting quality fluctuations, facilitating the prediction of the potential distribution of inference quality. The effectiveness of the evaluation framework is validated through experiments conducted on deep learning models for three different tasks across four systems.

Fanda Fan, Chaoxu Guo, Litong Gong et al.

Video outpainting aims to adequately complete missing areas at the edges of video frames. Compared to image outpainting, it presents an additional challenge as the model should maintain the temporal consistency of the filled area. In this paper, we introduce a masked 3D diffusion model for video outpainting. We use the technique of mask modeling to train the 3D diffusion model. This allows us to use multiple guide frames to connect the results of multiple video clip inferences, thus ensuring temporal consistency and reducing jitter between adjacent frames. Meanwhile, we extract the global frames of the video as prompts and guide the model to obtain information other than the current video clip using cross-attention. We also introduce a hybrid coarse-to-fine inference pipeline to alleviate the artifact accumulation problem. The existing coarse-to-fine pipeline only uses the infilling strategy, which brings degradation because the time interval of the sparse frames is too large. Our pipeline benefits from bidirectional learning of the mask modeling and thus can employ a hybrid strategy of infilling and interpolation when generating sparse frames. Experiments show that our method achieves state-of-the-art results in video outpainting tasks. More results and codes are provided at our https://fanfanda.github.io/M3DDM/.

Fanda Fan, Chunjie Luo, Wanling Gao et al.

The burgeoning field of Artificial Intelligence Generated Content (AIGC) is witnessing rapid advancements, particularly in video generation. This paper introduces AIGCBench, a pioneering comprehensive and scalable benchmark designed to evaluate a variety of video generation tasks, with a primary focus on Image-to-Video (I2V) generation. AIGCBench tackles the limitations of existing benchmarks, which suffer from a lack of diverse datasets, by including a varied and open-domain image-text dataset that evaluates different state-of-the-art algorithms under equivalent conditions. We employ a novel text combiner and GPT-4 to create rich text prompts, which are then used to generate images via advanced Text-to-Image models. To establish a unified evaluation framework for video generation tasks, our benchmark includes 11 metrics spanning four dimensions to assess algorithm performance. These dimensions are control-video alignment, motion effects, temporal consistency, and video quality. These metrics are both reference video-dependent and video-free, ensuring a comprehensive evaluation strategy. The evaluation standard proposed correlates well with human judgment, providing insights into the strengths and weaknesses of current I2V algorithms. The findings from our extensive experiments aim to stimulate further research and development in the I2V field. AIGCBench represents a significant step toward creating standardized benchmarks for the broader AIGC landscape, proposing an adaptable and equitable framework for future assessments of video generation tasks. We have open-sourced the dataset and evaluation code on the project website: https://www.benchcouncil.org/AIGCBench.

Yuan Liang, Yange Guo, Yanxia Gong et al.

Federated learning is a new machine learning paradigm. The goal is to build a machine learning model from the data sets distributed on multiple devices so-called an isolated data island, while keeping their data secure and private. Most existing federated learning benchmarks work manually splits commonly used public datasets into partitions to simulate real world isolated data island scenarios. Still, this simulation fails to capture real world isolated data island intrinsic characteristics. This paper presents a federated learning (FL) benchmark suite named FLBench. FLBench contains three domains: medical, financial, and AIoT. By configuring various domains, FLBench is qualified to evaluate federated learning systems and algorithms essential aspects, like communication, scenario transformation, privacy-preserving, data distribution heterogeneity, and cooperation strategy. Hence, it becomes a promising platform for developing novel federated learning algorithms. Currently, FLBench is open sourced and in fast evolution. We package it as an automated deployment tool. The benchmark suite is available from https://www.benchcouncil.org/flbench.html.

Zihan Jiang, Wanling Gao, Lei Wang et al.

In recent years, with the trend of applying deep learning (DL) in high performance scientific computing, the unique characteristics of emerging DL workloads in HPC raise great challenges in designing, implementing HPC AI systems. The community needs a new yard stick for evaluating the future HPC systems. In this paper, we propose HPC AI500 --- a benchmark suite for evaluating HPC systems that running scientific DL workloads. Covering the most representative scientific fields, each workload from HPC AI500 is based on real-world scientific DL applications. Currently, we choose 14 scientific DL benchmarks from perspectives of application scenarios, data sets, and software stack. We propose a set of metrics for comprehensively evaluating the HPC AI systems, considering both accuracy, performance as well as power and cost. We provide a scalable reference implementation of HPC AI500. HPC AI500 is a part of the open-source AIBench project, the specification and source code are publicly available from \url{http://www.benchcouncil.org/AIBench/index.html}.

Wanling Gao, Yuqing Zhu, Zhen Jia et al.

This paper presents our joint research efforts on big data benchmarking with several industrial partners. Considering the complexity, diversity, workload churns, and rapid evolution of big data systems, we take an incremental approach in big data benchmarking. For the first step, we pay attention to search engines, which are the most important domain in Internet services in terms of the number of page views and daily visitors. However, search engine service providers treat data, applications, and web access logs as business confidentiality, which prevents us from building benchmarks. To overcome those difficulties, with several industry partners, we widely investigated the open source solutions in search engines, and obtained the permission of using anonymous Web access logs. Moreover, with two years' great efforts, we created a sematic search engine named ProfSearch (available from http://prof.ict.ac.cn). These efforts pave the path for our big data benchmark suite from search engines---BigDataBench, which is released on the web page (http://prof.ict.ac.cn/BigDataBench). We report our detailed analysis of search engine workloads, and present our benchmarking methodology. An innovative data generation methodology and tool are proposed to generate scalable volumes of big data from a small seed of real data, preserving semantics and locality of data. Also, we preliminarily report two case studies using BigDataBench for both system and architecture researches.

Chunjie Luo, Fei Luo, Yusen Wang et al.

Reconstructing a dynamic human with loose clothing is an important but difficult task. To address this challenge, we propose a method named DLCA-Recon to create human avatars from monocular videos. The distance from loose clothing to the underlying body rapidly changes in every frame when the human freely moves and acts. Previous methods lack effective geometric initialization and constraints for guiding the optimization of deformation to explain this dramatic change, resulting in the discontinuous and incomplete reconstruction surface. To model the deformation more accurately, we propose to initialize an estimated 3D clothed human in the canonical space, as it is easier for deformation fields to learn from the clothed human than from SMPL. With both representations of explicit mesh and implicit SDF, we utilize the physical connection information between consecutive frames and propose a dynamic deformation field (DDF) to optimize deformation fields. DDF accounts for contributive forces on loose clothing to enhance the interpretability of deformations and effectively capture the free movement of loose clothing. Moreover, we propagate SMPL skinning weights to each individual and refine pose and skinning weights during the optimization to improve skinning transformation. Based on more reasonable initialization and DDF, we can simulate real-world physics more accurately. Extensive experiments on public and our own datasets validate that our method can produce superior results for humans with loose clothing compared to the SOTA methods.

Kuiye Ding, Fanda Fan, Yao Wang et al.

Multivariate Time Series Forecasting plays a key role in many applications. Recent works have explored using Large Language Models for MTSF to take advantage of their reasoning abilities. However, many methods treat LLMs as end-to-end forecasters, which often leads to a loss of numerical precision and forces LLMs to handle patterns beyond their intended design. Alternatively, methods that attempt to align textual and time series modalities within latent space frequently encounter alignment difficulty. In this paper, we propose to treat LLMs not as standalone forecasters, but as semantic guidance modules within a dual-stream framework. We propose DualSG, a dual-stream framework that provides explicit semantic guidance, where LLMs act as Semantic Guides to refine rather than replace traditional predictions. As part of DualSG, we introduce Time Series Caption, an explicit prompt format that summarizes trend patterns in natural language and provides interpretable context for LLMs, rather than relying on implicit alignment between text and time series in the latent space. We also design a caption-guided fusion module that explicitly models inter-variable relationships while reducing noise and computation. Experiments on real-world datasets from diverse domains show that DualSG consistently outperforms 15 state-of-the-art baselines, demonstrating the value of explicitly combining numerical forecasting with semantic guidance.

Kuiye Ding, Fanda Fan, Zheya Wang et al.

In the World Wide Web, reliable time series forecasts provide the forward-looking signals that drive resource planning, cache placement, and anomaly response, enabling platforms to operate efficiently as user behavior and content distributions evolve. Compared with other domains, time series forecasting for Web applications requires much faster responsiveness to support real-time decision making. We present KAIROS, a non-autoregressive time series forecasting framework that directly models segment-level multi-peak distributions. Unlike autoregressive approaches, KAIROS avoids error accumulation and achieves just-in-time inference, while improving over existing non-autoregressive models that collapse to over-smoothed predictions. Trained on the large-scale corpus, KAIROS demonstrates strong zero-shot generalization on six widely used benchmarks, delivering forecasting performance comparable to state-of-the-art foundation models with similar scale, at a fraction of their inference cost. Beyond empirical results, KAIROS highlights the importance of non-autoregressive design as a scalable paradigm for foundation models in time series.

Chunjie Luo, Jianfeng Zhan, Tianshu Hao et al.

Attention is an effective mechanism to improve the deep model capability. Squeeze-and-Excite (SE) introduces a light-weight attention branch to enhance the network's representational power. The attention branch is gated using the Sigmoid function and multiplied by the feature map's trunk branch. It is too sensitive to coordinate and balance the trunk and attention branches' contributions. To control the attention branch's influence, we propose a new attention method, called Shift-and-Balance (SB). Different from Squeeze-and-Excite, the attention branch is regulated by the learned control factor to control the balance, then added into the feature map's trunk branch. Experiments show that Shift-and-Balance attention significantly improves the accuracy compared to Squeeze-and-Excite when applied in more layers, increasing more size and capacity of a network. Moreover, Shift-and-Balance attention achieves better or close accuracy compared to the state-of-art Dynamic Convolution.

Chunjie Luo, Jianfeng Zhan, Lei Wang et al.

To build light-weight network, we propose a new normalization, Fine-grained Batch Normalization (FBN). Different from Batch Normalization (BN), which normalizes the final summation of the weighted inputs, FBN normalizes the intermediate state of the summation. We propose a novel light-weight network based on FBN, called Finet. At training time, the convolutional layer with FBN can be seen as an inverted bottleneck mechanism. FBN can be fused into convolution at inference time. After fusion, Finet uses the standard convolution with equal channel width, thus makes the inference more efficient. On ImageNet classification dataset, Finet achieves the state-of-art performance (65.706% accuracy with 43M FLOPs, and 73.786% accuracy with 303M FLOPs), Moreover, experiments show that Finet is more efficient than other state-of-art light-weight networks.

Chunjie Luo, Xiwen He, Jianfeng Zhan et al.

Due to increasing amounts of data and compute resources, deep learning achieves many successes in various domains. The application of deep learning on the mobile and embedded devices is taken more and more attentions, benchmarking and ranking the AI abilities of mobile and embedded devices becomes an urgent problem to be solved. Considering the model diversity and framework diversity, we propose a benchmark suite, AIoTBench, which focuses on the evaluation of the inference abilities of mobile and embedded devices. AIoTBench covers three typical heavy-weight networks: ResNet50, InceptionV3, DenseNet121, as well as three light-weight networks: SqueezeNet, MobileNetV2, MnasNet. Each network is implemented by three frameworks which are designed for mobile and embedded devices: Tensorflow Lite, Caffe2, Pytorch Mobile. To compare and rank the AI capabilities of the devices, we propose two unified metrics as the AI scores: Valid Images Per Second (VIPS) and Valid FLOPs Per Second (VOPS). Currently, we have compared and ranked 5 mobile devices using our benchmark. This list will be extended and updated soon after.

Wanling Gao, Fei Tang, Jianfeng Zhan et al.

Modern real-world application scenarios like Internet services consist of a diversity of AI and non-AI modules with huge code sizes and long and complicated execution paths, which raises serious benchmarking or evaluating challenges. Using AI components or micro benchmarks alone can lead to error-prone conclusions. This paper presents a methodology to attack the above challenge. We formalize a real-world application scenario as a Directed Acyclic Graph-based model and propose the rules to distill it into a permutation of essential AI and non-AI tasks, which we call a scenario benchmark. Together with seventeen industry partners, we extract nine typical scenario benchmarks. We design and implement an extensible, configurable, and flexible benchmark framework. We implement two Internet service AI scenario benchmarks based on the framework as proxies to two real-world application scenarios. We consider scenario, component, and micro benchmarks as three indispensable parts for evaluating. Our evaluation shows the advantage of our methodology against using component or micro AI benchmarks alone. The specifications, source code, testbed, and results are publicly available from \url{https://www.benchcouncil.org/aibench/scenario/}.

Fei Tang, Wanling Gao, Jianfeng Zhan et al.

Earlier-stage evaluations of a new AI architecture/system need affordable benchmarks. Only using a few AI component benchmarks like MLPerfalone in the other stages may lead to misleading conclusions. Moreover, the learning dynamics are not well understood, and the benchmarks' shelf-life is short. This paper proposes a balanced benchmarking methodology. We use real-world benchmarks to cover the factors space that impacts the learning dynamics to the most considerable extent. After performing an exhaustive survey on Internet service AI domains, we identify and implement nineteen representative AI tasks with state-of-the-art models. For repeatable performance ranking (RPR subset) and workload characterization (WC subset), we keep two subsets to a minimum for affordability. We contribute by far the most comprehensive AI training benchmark suite. The evaluations show: (1) AIBench Training (v1.1) outperforms MLPerfTraining (v0.7) in terms of diversity and representativeness of model complexity, computational cost, convergent rate, computation, and memory access patterns, and hotspot functions; (2) Against the AIBench full benchmarks, its RPR subset shortens the benchmarking cost by 64%, while maintaining the primary workload characteristics; (3) The performance ranking shows the single-purpose AI accelerator like TPU with the optimized TensorFlowframework performs better than that of GPUs while losing the latter's general support for various AI models. The specification, source code, and performance numbers are available from the AIBench homepage https://www.benchcouncil.org/aibench-training/index.html.

Chunjie Luo, Jianfeng Zhan, Lei Wang et al.

Batch normalization (BN) has become a standard technique for training the modern deep networks. However, its effectiveness diminishes when the batch size becomes smaller, since the batch statistics estimation becomes inaccurate. That hinders batch normalization's usage for 1) training larger model which requires small batches constrained by memory consumption, 2) training on mobile or embedded devices of which the memory resource is limited. In this paper, we propose a simple but effective method, called extended batch normalization (EBN). For NCHW format feature maps, extended batch normalization computes the mean along the (N, H, W) dimensions, as the same as batch normalization, to maintain the advantage of batch normalization. To alleviate the problem caused by small batch size, extended batch normalization computes the standard deviation along the (N, C, H, W) dimensions, thus enlarges the number of samples from which the standard deviation is computed. We compare extended batch normalization with batch normalization and group normalization on the datasets of MNIST, CIFAR-10/100, STL-10, and ImageNet, respectively. The experiments show that extended batch normalization alleviates the problem of batch normalization with small batch size while achieving close performances to batch normalization with large batch size.

Wanling Gao, Fei Tang, Jianfeng Zhan et al.

Domain-specific software and hardware co-design is encouraging as it is much easier to achieve efficiency for fewer tasks. Agile domain-specific benchmarking speeds up the process as it provides not only relevant design inputs but also relevant metrics, and tools. Unfortunately, modern workloads like Big data, AI, and Internet services dwarf the traditional one in terms of code size, deployment scale, and execution path, and hence raise serious benchmarking challenges. This paper proposes an agile domain-specific benchmarking methodology. Together with seventeen industry partners, we identify ten important end-to-end application scenarios, among which sixteen representative AI tasks are distilled as the AI component benchmarks. We propose the permutations of essential AI and non-AI component benchmarks as end-to-end benchmarks. An end-to-end benchmark is a distillation of the essential attributes of an industry-scale application. We design and implement a highly extensible, configurable, and flexible benchmark framework, on the basis of which, we propose the guideline for building end-to-end benchmarks, and present the first end-to-end Internet service AI benchmark. The preliminary evaluation shows the value of our benchmark suite---AIBench against MLPerf and TailBench for hardware and software designers, micro-architectural researchers, and code developers. The specifications, source code, testbed, and results are publicly available from the web site \url{http://www.benchcouncil.org/AIBench/index.html}.

Wanling Gao, Fei Tang, Lei Wang et al.

Today's Internet Services are undergoing fundamental changes and shifting to an intelligent computing era where AI is widely employed to augment services. In this context, many innovative AI algorithms, systems, and architectures are proposed, and thus the importance of benchmarking and evaluating them rises. However, modern Internet services adopt a microservice-based architecture and consist of various modules. The diversity of these modules and complexity of execution paths, the massive scale and complex hierarchy of datacenter infrastructure, the confidential issues of data sets and workloads pose great challenges to benchmarking. In this paper, we present the first industry-standard Internet service AI benchmark suite---AIBench with seventeen industry partners, including several top Internet service providers. AIBench provides a highly extensible, configurable, and flexible benchmark framework that contains loosely coupled modules. We identify sixteen prominent AI problem domains like learning to rank, each of which forms an AI component benchmark, from three most important Internet service domains: search engine, social network, and e-commerce, which is by far the most comprehensive AI benchmarking effort. On the basis of the AIBench framework, abstracting the real-world data sets and workloads from one of the top e-commerce providers, we design and implement the first end-to-end Internet service AI benchmark, which contains the primary modules in the critical paths of an industry scale application and is scalable to deploy on different cluster scales. The specifications, source code, and performance numbers are publicly available from the benchmark council web site http://www.benchcouncil.org/AIBench/index.html.

Wanling Gao, Jianfeng Zhan, Lei Wang et al.

Several fundamental changes in technology indicate domain-specific hardware and software co-design is the only path left. In this context, architecture, system, data management, and machine learning communities pay greater attention to innovative big data and AI algorithms, architecture, and systems. Unfortunately, complexity, diversity, frequently-changed workloads, and rapid evolution of big data and AI systems raise great challenges. First, the traditional benchmarking methodology that creates a new benchmark or proxy for every possible workload is not scalable, or even impossible for Big Data and AI benchmarking. Second, it is prohibitively expensive to tailor the architecture to characteristics of one or more application or even a domain of applications. We consider each big data and AI workload as a pipeline of one or more classes of units of computation performed on different initial or intermediate data inputs, each class of which we call a data motif. On the basis of our previous work that identifies eight data motifs taking up most of the run time of a wide variety of big data and AI workloads, we propose a scalable benchmarking methodology that uses the combination of one or more data motifs---to represent diversity of big data and AI workloads. Following this methodology, we present a unified big data and AI benchmark suite---BigDataBench 4.0, publicly available from~\url{http://prof.ict.ac.cn/BigDataBench}. This unified benchmark suite sheds new light on domain-specific hardware and software co-design: tailoring the system and architecture to characteristics of the unified eight data motifs other than one or more application case by case. Also, for the first time, we comprehensively characterize the CPU pipeline efficiency using the benchmarks of seven workload types in BigDataBench 4.0.

Chunjie Luo, Jianfeng Zhan, Lei Wang et al.

Traditionally, multi-layer neural networks use dot product between the output vector of previous layer and the incoming weight vector as the input to activation function. The result of dot product is unbounded, thus increases the risk of large variance. Large variance of neuron makes the model sensitive to the change of input distribution, thus results in poor generalization, and aggravates the internal covariate shift which slows down the training. To bound dot product and decrease the variance, we propose to use cosine similarity or centered cosine similarity (Pearson Correlation Coefficient) instead of dot product in neural networks, which we call cosine normalization. We compare cosine normalization with batch, weight and layer normalization in fully-connected neural networks as well as convolutional networks on the data sets of MNIST, 20NEWS GROUP, CIFAR-10/100 and SVHN. Experiments show that cosine normalization achieves better performance than other normalization techniques.