arXiv subjects

Pengfei Chen

Publications and source records attributed to Pengfei Chen.

At least 55 records · Page 3Linked to original sources

arXiv2024

ENOVA: Autoscaling towards Cost-effective and Stable Serverless LLM Serving

Since the increasing popularity of large language model (LLM) backend systems, it is common and necessary to deploy stable serverless serving of LLM on multi-GPU clusters with autoscaling. However, there exist challenges because the diversity and co-location of applications in multi-GPU clusters will lead to low service quality and GPU utilization. To address them, we build ENOVA, a deployment, monitoring and autoscaling service towards serverless LLM serving. ENOVA deconstructs the execution process of LLM service comprehensively, based on which ENOVA designs a configuration recommendation module for automatic deployment on any GPU clusters and a performance detection module for autoscaling. On top of them, ENOVA implements a deployment execution engine for multi-GPU cluster scheduling. The experiment results show that ENOVA significantly outperforms other state-of-the-art methods and is suitable for wide deployment in large online systems.

cs.DC↗

arXiv2024

CPR++: Object Localization via Single Coarse Point Supervision

Point-based object localization (POL), which pursues high-performance object sensing under low-cost data annotation, has attracted increased attention. However, the point annotation mode inevitably introduces semantic variance due to the inconsistency of annotated points. Existing POL heavily rely on strict annotation rules, which are difficult to define and apply, to handle the problem. In this study, we propose coarse point refinement (CPR), which to our best knowledge is the first attempt to alleviate semantic variance from an algorithmic perspective. CPR reduces the semantic variance by selecting a semantic centre point in a neighbourhood region to replace the initial annotated point. Furthermore, We design a sampling region estimation module to dynamically compute a sampling region for each object and use a cascaded structure to achieve end-to-end optimization. We further integrate a variance regularization into the structure to concentrate the predicted scores, yielding CPR++. We observe that CPR++ can obtain scale information and further reduce the semantic variance in a global region, thus guaranteeing high-performance object localization. Extensive experiments on four challenging datasets validate the effectiveness of both CPR and CPR++. We hope our work can inspire more research on designing algorithms rather than annotation rules to address the semantic variance problem in POL. The dataset and code will be public at github.com/ucas-vg/PointTinyBenchmark.

cs.CV↗

arXiv2024

AesBench: An Expert Benchmark for Multimodal Large Language Models on Image Aesthetics Perception

With collective endeavors, multimodal large language models (MLLMs) are undergoing a flourishing development. However, their performances on image aesthetics perception remain indeterminate, which is highly desired in real-world applications. An obvious obstacle lies in the absence of a specific benchmark to evaluate the effectiveness of MLLMs on aesthetic perception. This blind groping may impede the further development of more advanced MLLMs with aesthetic perception capacity. To address this dilemma, we propose AesBench, an expert benchmark aiming to comprehensively evaluate the aesthetic perception capacities of MLLMs through elaborate design across dual facets. (1) We construct an Expert-labeled Aesthetics Perception Database (EAPD), which features diversified image contents and high-quality annotations provided by professional aesthetic experts. (2) We propose a set of integrative criteria to measure the aesthetic perception abilities of MLLMs from four perspectives, including Perception (AesP), Empathy (AesE), Assessment (AesA) and Interpretation (AesI). Extensive experimental results underscore that the current MLLMs only possess rudimentary aesthetic perception ability, and there is still a significant gap between MLLMs and humans. We hope this work can inspire the community to engage in deeper explorations on the aesthetic potentials of MLLMs. Source data will be available at https://github.com/yipoh/AesBench.

cs.CV↗

arXiv2023

Nahida: In-Band Distributed Tracing with eBPF

Microservices are commonly used in modern cloud-native applications to achieve agility. However, the complexity of service dependencies in large-scale microservices systems can lead to anomaly propagation, making fault troubleshooting a challenge. To address this issue, distributed tracing systems have been proposed to trace complete request execution paths, enabling developers to troubleshoot anomalous services. However, existing distributed tracing systems have limitations such as invasive instrumentation, trace loss, or inaccurate trace correlation. To overcome these limitations, we propose a new tracing system based on eBPF (extended Berkeley Packet Filter), named Nahida, that can track complete requests in the kernel without intrusion, regardless of programming language or implementation. Our evaluation results show that Nahida can track over 92% of requests with stable accuracy, even under the high concurrency of user requests, while the state-of-the-art non-invasive approaches can not track any of the requests. Importantly, Nahida can track requests served by a multi-threaded application that none of the existing invasive tracing systems can handle by instrumenting tracing codes into libraries. Moreover, the overhead introduced by Nahida is negligible, increasing service latency by only 1.55%-2.1%. Overall, Nahida provides an effective and non-invasive solution for distributed tracing.

cs.OS↗

arXiv2023

LogShrink: Effective Log Compression by Leveraging Commonality and Variability of Log Data

Log data is a crucial resource for recording system events and states during system execution. However, as systems grow in scale, log data generation has become increasingly explosive, leading to an expensive overhead on log storage, such as several petabytes per day in production. To address this issue, log compression has become a crucial task in reducing disk storage while allowing for further log analysis. Unfortunately, existing general-purpose and log-specific compression methods have been limited in their ability to utilize log data characteristics. To overcome these limitations, we conduct an empirical study and obtain three major observations on the characteristics of log data that can facilitate the log compression task. Based on these observations, we propose LogShrink, a novel and effective log compression method by leveraging commonality and variability of log data. An analyzer based on longest common subsequence and entropy techniques is proposed to identify the latent commonality and variability in log messages. The key idea behind this is that the commonality and variability can be exploited to shrink log data with a shorter representation. Besides, a clustering-based sequence sampler is introduced to accelerate the commonality and variability analyzer. The extensive experimental results demonstrate that LogShrink can exceed baselines in compression ratio by 16% to 356% on average while preserving a reasonable compression speed.

cs.SE↗

arXiv2023

Can the Parker Solar Probe Detect a CME-flare Current Sheet?

A current sheet (CS) is the central structure in the disrupting magnetic configuration during solar eruptions. More than 90\% of the free magnetic energy (the difference between the energy in the non-potential magnetic field and that in the potential one) stored in the coronal magnetic field beforehand is converted into heating and kinetic energy of the plasma, as well as accelerating charged particles, by magnetic reconnection occurring in the CS. However, the detailed physical properties and fine structures of the CS are still unknown since there is no relevant information obtained via in situ detections. The Parker Solar Probe (PSP) may provide us such information should it traverse a CS in the eruption. The perihelion of PSP's final orbit is located at about 10 solar radii from the center of the Sun, so it can observe the CS at a very close distance, or even traverses the CS, which provides us a unique opportunity to look into fine properties and structures of the CS, helping reveal the detailed physics of large-scale reconnection that was impossible before. We evaluate the probability that PSP can traverse a CS, and examine the orbit of a PSP-like spacecraft that has the highest probability to traverse a CS.

astro-ph.SR↗

arXiv2023

Understanding the Lateral Drifting of an Erupting Filament with a Data-constrained Magnetohydrodynamic Simulation

Solar filaments often exhibit rotation and deflection during eruptions, which would significantly affect the geoeffectiveness of the corresponding coronal mass ejections (CMEs). Therefore, understanding the mechanisms that lead to such rotation and lateral displacement of filaments is a great concern to space weather forecasting. In this paper, we examine an intriguing filament eruption event observed by the Chinese Hα Solar Explorer (CHASE) and the Solar Dynamics Observatory (SDO). The filament, which eventually evolves into a CME, exhibits significant lateral drifting during its rising. Moreover, the orientation of the CME flux rope axis deviates from that of the pre-eruptive filament observed in the source region. To investigate the physical processes behind these observations, we perform a data-constrained magnetohydrodynamic (MHD) simulation. Many prominent observational features in the eruption are reproduced by our numerical model, including the morphology of the eruptive filament, eruption path, and flare ribbons. The simulation results reveal that the magnetic reconnection between the flux-rope leg and neighboring low-lying sheared arcades may be the primary mechanism responsible for the lateral drifting of the filament material. Such a reconnection geometry leads to flux-rope footpoint migration and a reconfiguration of its morphology. As a consequence, the filament material hosted in the flux rope drifts laterally, and the CME flux rope deviates from the pre-eruptive filament. This finding underscores the importance of external magnetic reconnection in influencing the orientation of a flux rope axis during eruption.

astro-ph.SR↗

arXiv2023

Spatial Self-Distillation for Object Detection with Inaccurate Bounding Boxes

Object detection via inaccurate bounding boxes supervision has boosted a broad interest due to the expensive high-quality annotation data or the occasional inevitability of low annotation quality (\eg tiny objects). The previous works usually utilize multiple instance learning (MIL), which highly depends on category information, to select and refine a low-quality box. Those methods suffer from object drift, group prediction and part domination problems without exploring spatial information. In this paper, we heuristically propose a \textbf{Spatial Self-Distillation based Object Detector (SSD-Det)} to mine spatial information to refine the inaccurate box in a self-distillation fashion. SSD-Det utilizes a Spatial Position Self-Distillation \textbf{(SPSD)} module to exploit spatial information and an interactive structure to combine spatial information and category information, thus constructing a high-quality proposal bag. To further improve the selection procedure, a Spatial Identity Self-Distillation \textbf{(SISD)} module is introduced in SSD-Det to obtain spatial confidence to help select the best proposals. Experiments on MS-COCO and VOC datasets with noisy box annotation verify our method's effectiveness and achieve state-of-the-art performance. The code is available at https://github.com/ucas-vg/PointTinyBenchmark/tree/SSD-Det.

cs.CV↗

arXiv2023

Why "solar tsunamis" rarely leave their imprints in the chromosphere

Solar coronal waves frequently appear as bright disturbances that propagate globally from the eruption center in the solar atmosphere, just like the tsunamis in the ocean on Earth. Theoretically, coronal waves can sweep over the underlying chromosphere and leave an imprint in the form of Moreton wave, due to the enhanced pressure beneath their coronal wavefront. Despite the frequent observations of coronal waves, their counterparts in the chromosphere are rarely detected. Why the chromosphere rarely bears the imprints of solar tsunamis remained a mystery since their discovery three decades ago. To resolve this question, all coronal waves and associated Moreton waves in the last decade have been initially surveyed, though the detection of Moreton waves could be hampered by utilising the low-quality H$α$ data from Global Oscillations Network Group. Here, we present 8 cases (including 5 in Appendix) of the coexistence of coronal and Moreton waves in inclined eruptions where it is argued that the extreme inclination is key to providing an answer to address the question. For all these events, the lowest part of the coronal wavefront near the solar surface appears very bright, and the simultaneous disturbances in the solar transition region and the chromosphere predominantly occur beneath the bright segment. Therefore, evidenced by observations, we propose a scenario for the excitation mechanism of the coronal-Moreton waves in highly inclined eruptions, in which the lowest part of a coronal wave can effectively disturb the chromosphere even for a weak (e.g., B-class) solar flare.

astro-ph.SR↗

arXiv2023

HyperTuner: A Cross-Layer Multi-Objective Hyperparameter Auto-Tuning Framework for Data Analytic Services

Hyper-parameters optimization (HPO) is vital for machine learning models. Besides model accuracy, other tuning intentions such as model training time and energy consumption are also worthy of attention from data analytic service providers. Hence, it is essential to take both model hyperparameters and system parameters into consideration to execute cross-layer multi-objective hyperparameter auto-tuning. Towards this challenging target, we propose HyperTuner in this paper. To address the formulated high-dimensional black-box multi-objective optimization problem, HyperTuner first conducts multi-objective parameter importance ranking with its MOPIR algorithm and then leverages the proposed ADUMBO algorithm to find the Pareto-optimal configuration set. During each iteration, ADUMBO selects the most promising configuration from the generated Pareto candidate set via maximizing a new well-designed metric, which can adaptively leverage the uncertainty as well as the predicted mean across all the surrogate models along with the iteration times. We evaluate HyperTuner on our local distributed TensorFlow cluster and experimental results show that it is always able to find a better Pareto configuration front superior in both convergence and diversity compared with the other four baseline algorithms. Besides, experiments with different training datasets, different optimization objectives and different machine learning platforms verify that HyperTuner can well adapt to various data analytic service scenarios.

cs.LG↗

arXiv2022

Point-to-Box Network for Accurate Object Detection via Single Point Supervision

Object detection using single point supervision has received increasing attention over the years. However, the performance gap between point supervised object detection (PSOD) and bounding box supervised detection remains large. In this paper, we attribute such a large performance gap to the failure of generating high-quality proposal bags which are crucial for multiple instance learning (MIL). To address this problem, we introduce a lightweight alternative to the off-the-shelf proposal (OTSP) method and thereby create the Point-to-Box Network (P2BNet), which can construct an inter-objects balanced proposal bag by generating proposals in an anchor-like way. By fully investigating the accurate position information, P2BNet further constructs an instance-level bag, avoiding the mixture of multiple objects. Finally, a coarse-to-fine policy in a cascade fashion is utilized to improve the IoU between proposals and ground-truth (GT). Benefiting from these strategies, P2BNet is able to produce high-quality instance-level bags for object detection. P2BNet improves the mean average precision (AP) by more than 50% relative to the previous best PSOD method on the MS COCO dataset. It also demonstrates the great potential to bridge the performance gap between point supervised and bounding-box supervised detectors. The code will be released at github.com/ucas-vg/P2BNet.

cs.CV↗

arXiv2022

Constructing Large-Scale Real-World Benchmark Datasets for AIOps

Recently, AIOps (Artificial Intelligence for IT Operations) has been well studied in academia and industry to enable automated and effective software service management. Plenty of efforts have been dedicated to AIOps, including anomaly detection, root cause localization, incident management, etc. However, most existing works are evaluated on private datasets, so their generality and real performance cannot be guaranteed. The lack of public large-scale real-world datasets has prevented researchers and engineers from enhancing the development of AIOps. To tackle this dilemma, in this work, we introduce three public real-world, large-scale datasets about AIOps, mainly aiming at KPI anomaly detection, root cause localization on multi-dimensional data, and failure discovery and diagnosis. More importantly, we held three competitions in 2018/2019/2020 based on these datasets, attracting thousands of teams to participate. In the future, we will continue to publish more datasets and hold competitions to promote the development of AIOps further.

cs.SE↗

arXiv2022

Acknowledging the Unknown for Multi-label Learning with Single Positive Labels

Due to the difficulty of collecting exhaustive multi-label annotations, multi-label datasets often contain partial labels. We consider an extreme of this weakly supervised learning problem, called single positive multi-label learning (SPML), where each multi-label training image has only one positive label. Traditionally, all unannotated labels are assumed as negative labels in SPML, which introduces false negative labels and causes model training to be dominated by assumed negative labels. In this work, we choose to treat all unannotated labels from an alternative perspective, i.e. acknowledging they are unknown. Hence, we propose entropy-maximization (EM) loss to attain a special gradient regime for providing proper supervision signals. Moreover, we propose asymmetric pseudo-labeling (APL), which adopts asymmetric-tolerance strategies and a self-paced procedure, to cooperate with EM loss and then provide more precise supervision. Experiments show that our method significantly improves performance and achieves state-of-the-art results on all four benchmarks. Code is available at https://github.com/Correr-Zhou/SPML-AckTheUnknown.

cs.CV↗

arXiv2022

Robust Medical Image Classification from Noisy Labeled Data with Global and Local Representation Guided Co-training

Deep neural networks have achieved remarkable success in a wide variety of natural image and medical image computing tasks. However, these achievements indispensably rely on accurately annotated training data. If encountering some noisy-labeled images, the network training procedure would suffer from difficulties, leading to a sub-optimal classifier. This problem is even more severe in the medical image analysis field, as the annotation quality of medical images heavily relies on the expertise and experience of annotators. In this paper, we propose a novel collaborative training paradigm with global and local representation learning for robust medical image classification from noisy-labeled data to combat the lack of high quality annotated medical data. Specifically, we employ the self-ensemble model with a noisy label filter to efficiently select the clean and noisy samples. Then, the clean samples are trained by a collaborative training strategy to eliminate the disturbance from imperfect labeled samples. Notably, we further design a novel global and local representation learning scheme to implicitly regularize the networks to utilize noisy samples in a self-supervised manner. We evaluated our proposed robust learning strategy on four public medical image classification datasets with three types of label noise,ie,random noise, computer-generated label noise, and inter-observer variability noise. Our method outperforms other learning from noisy label methods and we also conducted extensive experiments to analyze each component of our method.

eess.IV↗

arXiv2022

Object Localization under Single Coarse Point Supervision

Point-based object localization (POL), which pursues high-performance object sensing under low-cost data annotation, has attracted increased attention. However, the point annotation mode inevitably introduces semantic variance for the inconsistency of annotated points. Existing POL methods heavily reply on accurate key-point annotations which are difficult to define. In this study, we propose a POL method using coarse point annotations, relaxing the supervision signals from accurate key points to freely spotted points. To this end, we propose a coarse point refinement (CPR) approach, which to our best knowledge is the first attempt to alleviate semantic variance from the perspective of algorithm. CPR constructs point bags, selects semantic-correlated points, and produces semantic center points through multiple instance learning (MIL). In this way, CPR defines a weakly supervised evolution procedure, which ensures training high-performance object localizer under coarse point supervision. Experimental results on COCO, DOTA and our proposed SeaPerson dataset validate the effectiveness of the CPR approach. The dataset and code will be available at https://github.com/ucas-vg/PointTinyBenchmark/.

cs.CV↗

arXiv2021

Formation and Characteristics of Filament Threads in Double-Dipped Magnetic Flux Tubes

As one of the main formation mechanisms of solar filament formation, the chromospheric evaporation-coronal condensation model has been confirmed by numerical simulations to explain the formation of filament threads very well in flux tubes with single dips. However, coronal magnetic extrapolations indicated that some magnetic field lines might possess more than one dip. It is expected that the formation process would be significantly different in this case compared to a single-dipped magnetic flux tube. In this paper, based on the evaporation-condensation model, we study filament thread formation in double-dipped magnetic flux tubes by numerical simulations. We find that only with particular combinations of magnetic configuration and heating, e.g., concentrated localized heating and a long magnetic flux tube with deep dips, can two threads form and persist in a double-dipped magnetic flux tube. Comparing our parametric survey with observations, we conclude that such magnetically connected threads due to multiple dips are more likely to exist in quiescent filaments than in active-region filaments. Moreover, we find that these threads are usually shorter than independently trapped threads, which might be one of the reasons why quiescent filaments have short threads. These characteristics of magnetically connected threads could also explain barbs and vertical threads in quiescent filaments.

astro-ph.SR↗

arXiv2021

Magnetic Twists of Solar Filaments

Solar filaments are cold and dense materials situated in magnetic dips, which show distinct radiation characteristics compared to the surrounding coronal plasma. They are associated with coronal sheared and twisted magnetic field lines. However, the exact magnetic configuration supporting a filament material is not easy to be ascertained because of the absence of routine observations of the magnetic field inside filaments. Since many filaments lie above weak-field regions, it is nearly impossible to extrapolate their coronal magnetic structures by applying the traditional methods to noisy photospheric magnetograms, in particular the horizontal components. In this paper, we construct magnetic structures for some filaments with the regularized Biot--Savart laws and calculate their magnetic twists. Moreover, we make a parameter survey for the flux ropes of the Titov-Demoulin-modified model to explore the factors affecting the twist of a force-free magnetic flux rope. It is found that the twist of a force-free flux rope is proportional to its axial length to minor radius ratio, and is basically independent of the overlying background magnetic field strength. Thus, we infer that long quiescent filaments are likely to be supported by more twisted flux ropes than short active-region filaments, which is consistent with observations.

astro-ph.SR↗

arXiv2021

Foresee then Evaluate: Decomposing Value Estimation with Latent Future Prediction

Value function is the central notion of Reinforcement Learning (RL). Value estimation, especially with function approximation, can be challenging since it involves the stochasticity of environmental dynamics and reward signals that can be sparse and delayed in some cases. A typical model-free RL algorithm usually estimates the values of a policy by Temporal Difference (TD) or Monte Carlo (MC) algorithms directly from rewards, without explicitly taking dynamics into consideration. In this paper, we propose Value Decomposition with Future Prediction (VDFP), providing an explicit two-step understanding of the value estimation process: 1) first foresee the latent future, 2) and then evaluate it. We analytically decompose the value function into a latent future dynamics part and a policy-independent trajectory return part, inducing a way to model latent dynamics and returns separately in value estimation. Further, we derive a practical deep RL algorithm, consisting of a convolutional model to learn compact trajectory representation from past experiences, a conditional variational auto-encoder to predict the latent future dynamics and a convex return model that evaluates trajectory representation. In experiments, we empirically demonstrate the effectiveness of our approach for both off-policy and on-policy RL in several OpenAI Gym continuous control tasks as well as a few challenging variants with delayed reward.

cs.LG↗