Search

Recent advances in fast optical switching technology show promise in meeting the high goodput and low latency requirements of datacenter networks (DCN). We present NegotiaToR, a simple network architecture for optical reconfigurable DCNs that utilizes on-demand scheduling to handle dynamic traffic. In NegotiaToR, racks exchange scheduling messages through an in-band control plane and distributedly calculate non-conflicting paths from binary traffic demand information. Optimized for incasts, it also provides opportunities to bypass scheduling delays. NegotiaToR is compatible with prevalent flat topologies, and is tailored towards a minimalist design for on-demand reconfigurable DCNs, enhancing practicality. Through large-scale simulations, we show that NegotiaToR achieves both small mice flow completion time (FCT) and high goodput on two representative flat topologies, especially under heavy loads. Particularly, the FCT of mice flows is one to two orders of magnitude better than the state-of-the-art traffic-oblivious reconfigurable DCN design., Comment: This paper is accepted by ACM SIGCOMM 2024

Large language models (LLMs) exhibit remarkable capabilities in understanding and generating natural language. However, these models can inadvertently memorize private information, posing significant privacy risks. This study addresses the challenge of enabling LLMs to protect specific individuals' private data without the need for complete retraining. We propose \return, a Real-world pErsonal daTa UnleaRNing dataset, comprising 2,492 individuals from Wikipedia with associated QA pairs, to evaluate machine unlearning (MU) methods for protecting personal data in a realistic scenario. Additionally, we introduce the Name-Aware Unlearning Framework (NAUF) for Privacy Protection, which enables the model to learn which individuals' information should be protected without affecting its ability to answer questions related to other unrelated individuals. Our extensive experiments demonstrate that NAUF achieves a state-of-the-art average unlearning score, surpassing the best baseline method by 5.65 points, effectively protecting target individuals' personal data while maintaining the model's general capabilities.

Large Language Models (LLMs) have shown their impressive capabilities, while also raising concerns about the data contamination problems due to privacy issues and leakage of benchmark datasets in the pre-training phase. Therefore, it is vital to detect the contamination by checking whether an LLM has been pre-trained on the target texts. Recent studies focus on the generated texts and compute perplexities, which are superficial features and not reliable. In this study, we propose to utilize the probing technique for pre-training data detection by examining the model's internal activations. Our method is simple and effective and leads to more trustworthy pre-training data detection. Additionally, we propose ArxivMIA, a new challenging benchmark comprising arxiv abstracts from Computer Science and Mathematics categories. Our experiments demonstrate that our method outperforms all baselines, and achieves state-of-the-art performance on both WikiMIA and ArxivMIA, with additional experiments confirming its efficacy (Our code and dataset are available at https://github.com/zhliu0106/probing-lm-data)., Comment: Accepted by ACL-2024 main conference

The demand for large language model (LLM) inference is gradually dominating the artificial intelligence workloads. Therefore, there is an urgent need for cost-efficient inference serving. Existing work focuses on single-worker optimization and lacks consideration of cluster-level management for both inference queries and computing resources. However, placing requests and managing resources without considering the query features easily causes SLO violations or resource underutilization. Providers are forced to allocate extra computing resources to guarantee user experience, leading to additional serving costs. In this paper we introduce Aladdin, a scheduler that co-adaptively places queries and scales computing resources with SLO awareness. For a stream of inference queries, Aladdin first predicts minimal computing resources and the corresponding serving workers' configuration required to fulfill the SLOs for all queries. Then, it places the queries to each serving worker according to the prefill and decode latency models of batched LLM inference to maximize each worker's utilization. Results show that Aladdin reduces the serving cost of a single model by up to 71% for the same SLO level compared with the baselines, which can be millions of dollars per year.

Speculative decoding has demonstrated its effectiveness in accelerating the inference of large language models while maintaining a consistent sampling distribution. However, the conventional approach of training a separate draft model to achieve a satisfactory token acceptance rate can be costly. Drawing inspiration from early exiting, we propose a novel self-speculative decoding framework \emph{Kangaroo}, which uses a fixed shallow sub-network as a self-draft model, with the remaining layers serving as the larger target model. We train a lightweight and efficient adapter module on top of the sub-network to bridge the gap between the sub-network and the full model's representation ability. It is noteworthy that the inference latency of the self-draft model may no longer be negligible compared to the large model, necessitating strategies to increase the token acceptance rate while minimizing the drafting steps of the small model. To address this challenge, we introduce an additional early exiting mechanism for generating draft tokens. Specifically, we halt the small model's subsequent prediction during the drafting phase once the confidence level for the current token falls below a certain threshold. Extensive experiments on the Spec-Bench demonstrate the effectiveness of Kangaroo. Under single-sequence verification, Kangaroo achieves speedups up to $1.68\times$ on Spec-Bench, outperforming Medusa-1 with 88.7\% fewer additional parameters (67M compared to 591M). The code for Kangaroo is available at https://github.com/Equationliu/Kangaroo.

Compact neural networks are specially designed for applications on edge devices with faster inference speed yet modest performance. However, training strategies of compact models are borrowed from that of conventional models at present, which ignores their difference in model capacity and thus may impede the performance of compact models. In this paper, by systematically investigating the impact of different training ingredients, we introduce a strong training strategy for compact models. We find that the appropriate designs of re-parameterization and knowledge distillation are crucial for training high-performance compact models, while some commonly used data augmentations for training conventional models, such as Mixup and CutMix, lead to worse performance. Our experiments on ImageNet-1K dataset demonstrate that our specialized training strategy for compact models is applicable to various architectures, including GhostNetV2, MobileNetV2 and ShuffleNetV2. Specifically, equipped with our strategy, GhostNetV3 1.3$\times$ achieves a top-1 accuracy of 79.1% with only 269M FLOPs and a latency of 14.46ms on mobile devices, surpassing its ordinarily trained counterpart by a large margin. Moreover, our observation can also be extended to object detection scenarios. PyTorch code and checkpoints can be found at https://github.com/huawei-noah/Efficient-AI-Backbones/tree/master/ghostnetv3_pytorch.

In real-life conversations, the content is diverse, and there exists the one-to-many problem that requires diverse generation. Previous studies attempted to introduce discrete or Gaussian-based continuous latent variables to address the one-to-many problem, but the diversity is limited. Recently, diffusion models have made breakthroughs in computer vision, and some attempts have been made in natural language processing. In this paper, we propose DiffusionDialog, a novel approach to enhance the diversity of dialogue generation with the help of diffusion model. In our approach, we introduce continuous latent variables into the diffusion model. The problem of using latent variables in the dialog task is how to build both an effective prior of the latent space and an inferring process to obtain the proper latent given the context. By combining the encoder and latent-based diffusion model, we encode the response's latent representation in a continuous space as the prior, instead of fixed Gaussian distribution or simply discrete ones. We then infer the latent by denoising step by step with the diffusion model. The experimental results show that our model greatly enhances the diversity of dialog responses while maintaining coherence. Furthermore, in further analysis, we find that our diffusion model achieves high inference efficiency, which is the main challenge of applying diffusion models in natural language processing., Comment: LREC-COLING 2024 camera ready

Data augmentation (DA) is crucial to mitigate model training instability and over-fitting problems in low-resource open-domain dialogue generation. However, traditional DA methods often neglect semantic data diversity, restricting the overall quality. Recently, large language models (LLM) have been used for DA to generate diversified dialogues. However, they have limited controllability and tend to generate dialogues with a distribution shift compared to the seed dialogues. To maximize the augmentation diversity and address the controllability problem, we propose \textbf{S}ummary-based \textbf{D}ialogue \textbf{A}ugmentation with LLM (SDA). Our approach enhances the controllability of LLM by using dialogue summaries as a planning tool. Based on summaries, SDA can generate high-quality and diverse dialogue data even with a small seed dataset. To evaluate the efficacy of data augmentation methods for open-domain dialogue, we designed a clustering-based metric to characterize the semantic diversity of the augmented dialogue data. The experimental results show that SDA can augment high-quality and semantically diverse dialogues given a small seed dataset and an LLM, and the augmented data can boost the performance of open-domain dialogue models., Comment: 13 pages, 5 figures

Adjusting batch sizes and adaptively tuning other hyperparameters can significantly speed up deep neural network (DNN) training. Despite the ubiquity of heterogeneous clusters, existing adaptive DNN training techniques solely consider homogeneous environments. Optimizing distributed DNN training over heterogeneous clusters is technically challenging, and directly adapting existing techniques results in low utilization and poor performance. To solve this problem, we introduce Cannikin -- a novel data-parallel distributed training system. Cannikin achieves efficient and near-optimal performance by accurately modeling the optimal system performance and predicting adaptive batch size training metrics for DNNs in heterogeneous clusters. We implemented Cannikin in PyTorch and conducted experiments over 16 GPUs in Chameleon. Empirical results show that Cannikin reduces DNN training in heterogeneous clusters by up to $52\%$ compared to the state-of-the-art adaptive training system and up to $85\%$ compared to native PyTorch DistributedDataParallel.

De Lellis and coauthors have proved a sharp regularity theorem for area-minimizing currents in finite coefficient homology. They prove that area-minimizing mod $v$ currents are smooth outside of a singular set of codimension at least $1.$ Classical examples like triple junctions demonstrate that their result is sharp. Surprisingly, even though their regularity theorem cannot be improved for any fixed $v$, if one instead fixes the homology class, then $v$ asymptotically one can always achieve more regularity. For any integral homology class $[\Sigma]$ on any Riemannian manifold, we show that for $v$ large, any area-minimizing mod $v$ current in $[\Sigma\mod v]$ must be an integral current, thus having a singular set of codimension at least $2$ in general and of codimension at least $7$ in the hypersurface case. Similar results are obtained for Plateau problems in Euclidean space. Our work is inspired by Morgan's work and based on De Lellis' and coauthors' work.

In 1974, Federer proved that all area-minimizing hypersurfaces on orientable manifolds were calibrated by weakly closed differential forms. However, in this manuscript, we prove the contrary in higher codimensions: calibrated area-minimizers are non-generic. This is surprising given that almost all known examples of area-minimizing surfaces are confirmed to be minimizing via calibration. Let integers $d\ge 1$ and $c\ge 2$ denote dimensions and codimensions, respectively. Let $M^{d+c}$ denote a closed, orientable, smooth manifold of dimension $d+c$. For each $d$-dimensional integral homology class $[\Sigma]$ on $M$, we introduce $\Omega_{[\Sigma]}$ as the set of metrics for which any $d$-dimensional homologically area-minimizing surface in the homology class $[\Sigma]$ in any $g\in \Omega_{[\Sigma]}$ cannot be calibrated by any weakly closed measurable differential form. Our main result is that $\Omega_{[\Sigma]}$ is always a non-empty open set. To exemplify the prevalence of such phenomenon, we show that for any homology class $[\Sigma]$ on $\mathbb{CP}^n,$ the closure of $\Omega_{[\Sigma]}$ contains the Fubini-Study metric. In the hypersurface case, we show that even when a smooth area-minimizer is present, the calibration forms are compelled in some cases to have a non-empty singular set. This provides an answer to a question posed by Michael Freedman. Finally, we show that the ratio of the integral minimal mass to the real minimal mass is unbounded when we consider all Riemannian metrics. Also, it is always possible to fill a multiple of a homology class with an arbitrarily small area compared to the class itself. This settles the Riemannian version of several conjectures by Frank Morgan, Brian White and Robert Young.

Data centers have significant energy needs, both embodied and operational, affecting sustainability adversely. The current techniques and tools for collecting, aggregating, and reporting verifiable sustainability data are vulnerable to cyberattacks and misuse, requiring new security and privacy-preserving solutions. This paper outlines security challenges and research directions for addressing these pressing requirements.

Generating a long-distance quantum entanglement is one of the most essential functions of a quantum network to support quantum communication and computing applications. The successful entanglement rate during a probabilistic entanglement process decreases dramatically with distance, and swapping is a widely-applied quantum technique to address this issue. Most existing entanglement routing protocols use a classic entanglement-swapping method based on Bell State measurements that can only fuse two successful entanglement links. This paper appeals to a more general and efficient swapping method, namely n-fusion based on Greenberger-Horne-Zeilinger measurements that can fuse n successful entanglement links, to maximize the entanglement rate for multiple quantum-user pairs over a quantum network. We propose efficient entanglement routing algorithms that utilize the properties of n-fusion for quantum networks with general topologies. Evaluation results highlight that our proposed algorithm under n-fusion can greatly improve the network performance compared with existing ones.

This paper is a continuation of our work on a conjecture of Almgren on area-minimizing surfaces with fractal singular sets. First, we prove that area-minimizing surfaces with fractal singular sets are prevalent on the homology level on general manifolds. They exist even in metrics arbitrarily close to the flat metric on a torus. Second, we partially determine the moduli space of area-minimizing currents near the ones with fractal singular set that we constructed in this and previous work. Around strata of codimension at least three, the fractal self-intersections deform into transverse immersions under generic perturbations of metric. The result is sharp and we construct concrete examples to show that our fractal singularities do not necessarily completely dissolve under generic perturbations. Finally, we improve our previous results to give examples with different strata limiting to each other and refine the first stratum to fractal subsets of finite combinatorial graphs., Comment: Updated for NSF grant info

With the development of multimedia technology, Video Copy Detection has been a crucial problem for social media platforms. Meta AI hold Video Similarity Challenge on CVPR 2023 to push the technology forward. In this report, we share our winner solutions on Matching Track. We propose a Similarity Alignment Model(SAM) for video copy segment matching. Our SAM exhibits superior performance compared to other competitors, with a 0.108 / 0.144 absolute improvement over the second-place competitor in Phase 1 / Phase 2. Code is available at https://github.com/FeipengMa6/VSC22-Submission/tree/main/VSC22-Matching-Track-1st.

With the development of multimedia technology, Video Copy Detection has been a crucial problem for social media platforms. Meta AI hold Video Similarity Challenge on CVPR 2023 to push the technology forward. In this paper, we share our winner solutions on both tracks to help progress in this area. For Descriptor Track, we propose a dual-level detection method with Video Editing Detection (VED) and Frame Scenes Detection (FSD) to tackle the core challenges on Video Copy Detection. Experimental results demonstrate the effectiveness and efficiency of our proposed method. Code is available at https://github.com/FeipengMa6/VSC22-Submission.

In this paper, a novel Diffusion-based 3D Pose estimation (D3DP) method with Joint-wise reProjection-based Multi-hypothesis Aggregation (JPMA) is proposed for probabilistic 3D human pose estimation. On the one hand, D3DP generates multiple possible 3D pose hypotheses for a single 2D observation. It gradually diffuses the ground truth 3D poses to a random distribution, and learns a denoiser conditioned on 2D keypoints to recover the uncontaminated 3D poses. The proposed D3DP is compatible with existing 3D pose estimators and supports users to balance efficiency and accuracy during inference through two customizable parameters. On the other hand, JPMA is proposed to assemble multiple hypotheses generated by D3DP into a single 3D pose for practical use. It reprojects 3D pose hypotheses to the 2D camera plane, selects the best hypothesis joint-by-joint based on the reprojection errors, and combines the selected joints into the final pose. The proposed JPMA conducts aggregation at the joint level and makes use of the 2D prior information, both of which have been overlooked by previous approaches. Extensive experiments on Human3.6M and MPI-INF-3DHP datasets show that our method outperforms the state-of-the-art deterministic and probabilistic approaches by 1.5% and 8.9%, respectively. Code is available at https://github.com/paTRICK-swk/D3DP., Comment: ICCV 2023

Let dimensions $d \ge 3$ and codimensions $c \ge 3$ be positive integers, and define the exceptional set $E = \{3, 4\}$. We prove that, for $d \notin E$ with arbitrary $c$, or for $d \in E$ with $c \le d$, for any $d$-dimensional integral homology class $[\Sigma]$ on a compact (not necessarily orientable) $(d+c)$-dimensional smooth manifold $M$, there exist open sets $\Omega_{[\Sigma]}$ in the space of smooth Riemannian metrics such that all area-minimizing integral currents in $[\Sigma]$ are singular for metrics in $\Omega_{[\Sigma]}$. This resolves a conjecture by White regarding the generic regularity of area-minimizing surfaces and provides a sharp dimension-wise answer. As a corollary, we determine the moduli space of area-minimizing currents in the vicinity of any area-minimizing transverse immersion with dimension $d \ge 3$ and codimension $c \ge 3$ that satisfies an angle condition of asymptotically sharp order in $d$. Analogous conclusions apply to mod $2$ area-minimizing surfaces., Comment: Updated for NSF grant info

Quantum networks are considered as a promising future platform for quantum information exchange and quantum applications, which have capabilities far beyond the traditional communication networks. Remote quantum entanglement is an essential component of a quantum network. How to efficiently design a multi-routing entanglement protocol is a fundamental yet challenging problem. In this paper, we study a quantum entanglement routing problem to simultaneously maximize the number of quantum-user pairs and their expected throughput. Our approach is to formulate the problem as two sequential integer programming steps. We propose efficient entanglement routing algorithms for the two integer programming steps and analyze their time complexity and performance bounds. Results of evaluation highlight that our approach outperforms existing solutions in both served quantum-user pairs numbers and the network expected throughput.

This paper introduces a novel Pre-trained Spatial Temporal Many-to-One (P-STMO) model for 2D-to-3D human pose estimation task. To reduce the difficulty of capturing spatial and temporal information, we divide this task into two stages: pre-training (Stage I) and fine-tuning (Stage II). In Stage I, a self-supervised pre-training sub-task, termed masked pose modeling, is proposed. The human joints in the input sequence are randomly masked in both spatial and temporal domains. A general form of denoising auto-encoder is exploited to recover the original 2D poses and the encoder is capable of capturing spatial and temporal dependencies in this way. In Stage II, the pre-trained encoder is loaded to STMO model and fine-tuned. The encoder is followed by a many-to-one frame aggregator to predict the 3D pose in the current frame. Especially, an MLP block is utilized as the spatial feature extractor in STMO, which yields better performance than other methods. In addition, a temporal downsampling strategy is proposed to diminish data redundancy. Extensive experiments on two benchmarks show that our method outperforms state-of-the-art methods with fewer parameters and less computational overhead. For example, our P-STMO model achieves 42.1mm MPJPE on Human3.6M dataset when using 2D poses from CPN as inputs. Meanwhile, it brings a 1.5-7.1 times speedup to state-of-the-art methods. Code is available at https://github.com/paTRICK-swk/P-STMO., Comment: ECCV 2022