Hui Guan

Lijun Zhang, Xiao Liu, Antoni Viros Martin et al.

Watermarking images is critical for tracking image provenance and proving ownership. With the advent of generative models, such as stable diffusion, that can create fake but realistic images, watermarking has become particularly important to make human-created images reliably identifiable. Unfortunately, the very same stable diffusion technology can remove watermarks injected using existing methods. To address this problem, we present ZoDiac, which uses a pre-trained stable diffusion model to inject a watermark into the trainable latent space, resulting in watermarks that can be reliably detected in the latent vector even when attacked. We evaluate ZoDiac on three benchmarks, MS-COCO, DiffusionDB, and WikiArt, and find that ZoDiac is robust against state-of-the-art watermark attacks, with a watermark detection rate above 98% and a false positive rate below 6.4%, outperforming state-of-the-art watermarking methods. We hypothesize that the reciprocating denoising process in diffusion models may inherently enhance the robustness of the watermark when faced with strong attacks and validate the hypothesis. Our research demonstrates that stable diffusion is a promising approach to robust watermarking, able to withstand even stable-diffusion--based attack methods. ZoDiac is open-sourced and available at https://github.com/zhanglijun95/ZoDiac.

Mohammad Mehdi Rastikerdar, Jin Huang, Hui Guan et al.

Resource-constrained IoT devices increasingly rely on deep learning models, however, these models experience significant accuracy drops due to domain shifts when encountering variations in lighting, weather, and seasonal conditions. While cloud-based retraining can address this issue, many IoT deployments operate with limited connectivity and energy constraints, making traditional fine-tuning approaches impractical. We explore this challenge through the lens of wildlife ecology, where camera traps must maintain accurate species classification across changing seasons, weather, and habitats without reliable connectivity. We introduce WildFit, an autonomous in-situ adaptation framework that leverages the key insight that background scenes change more frequently than the visual characteristics of monitored species. WildFit combines background-aware synthesis to generate training samples on-device with drift-aware fine-tuning that triggers model updates only when necessary to conserve resources. Our background-aware synthesis surpasses efficient baselines by 7.3% and diffusion models by 3.0% while being orders of magnitude faster, our drift-aware fine-tuning achieves Pareto optimality with 50% fewer updates and 1.5% higher accuracy, and the end-to-end system outperforms domain adaptation approaches by 20-35% while consuming only 11.2 Wh over 37 days-enabling battery-powered deployment.

Mingcan Xiang, Steven Jiaxun Tang, Qizheng Yang et al.

In the domain of multimedia and multimodal processing, the efficient handling of diverse data streams such as images, video, and sensor data is paramount. Model compression and multitask learning (MTL) are crucial in this field, offering the potential to address the resource-intensive demands of processing and interpreting multiple forms of media simultaneously. However, effectively compressing a multitask model presents significant challenges due to the complexities of balancing sparsity allocation and accuracy performance across multiple tasks. To tackle these challenges, we propose AdapMTL, an adaptive pruning framework for MTL models. AdapMTL leverages multiple learnable soft thresholds independently assigned to the shared backbone and the task-specific heads to capture the nuances in different components' sensitivity to pruning. During training, it co-optimizes the soft thresholds and MTL model weights to automatically determine the suitable sparsity level at each component to achieve both high task accuracy and high overall sparsity. It further incorporates an adaptive weighting mechanism that dynamically adjusts the importance of task-specific losses based on each task's robustness to pruning. We demonstrate the effectiveness of AdapMTL through comprehensive experiments on popular multitask datasets, namely NYU-v2 and Tiny-Taskonomy, with different architectures, showcasing superior performance compared to state-of-the-art pruning methods.

Mohammad Mehdi Rastikerdar, Jin Huang, Shiwei Fang et al.

While existing strategies to execute deep learning-based classification on low-power platforms assume the models are trained on all classes of interest, this paper posits that adopting context-awareness i.e. narrowing down a classification task to the current deployment context consisting of only recent inference queries can substantially enhance performance in resource-constrained environments. We propose a new paradigm, CACTUS, for scalable and efficient context-aware classification where a micro-classifier recognizes a small set of classes relevant to the current context and, when context change happens (e.g., a new class comes into the scene), rapidly switches to another suitable micro-classifier. CACTUS features several innovations, including optimizing the training cost of context-aware classifiers, enabling on-the-fly context-aware switching between classifiers, and balancing context switching costs and performance gains via simple yet effective switching policies. We show that CACTUS achieves significant benefits in accuracy, latency, and compute budget across a range of datasets and IoT platforms.

Xiao Liu, Lijun Zhang, Deepak Ganesan et al.

Transformer models power many AI applications but suffer from high inference latency, limiting their use in real-time settings. Multi-device inference can reduce latency by parallelizing computation. Yet, existing methods require high inter-device bandwidth, making them impractical for bandwidth-constrained environments. We propose ASTRA, a communication-efficient framework that accelerates Transformer inference through a novel integration of sequence parallelism and a Mixed-Precision Attention mechanism designed to minimize inter-device communication. ASTRA compresses non-local token embeddings via vector quantization and preserves task accuracy through two optimizations, Noise-Augmented Quantization and Distributed Class Tokens. Experiments on ViT and GPT2 across vision and NLP tasks show that ASTRA achieves up to 2.64X speedups over single-device inference and up to 15.25X speedups over state-of-the-art multi-device inferences, while operating under bandwidths as low as 10 Mbps. ASTRA is open-sourced at https://github.com/xl1990/Astra.

Zhenning Yang, Hui Guan, Victor Nicolet et al.

Cloud infrastructure is managed through a mix of interfaces -- traditionally, cloud consoles, command-line interfaces (CLI), and SDKs are the tools of choice. Recently, Infrastructure-as-Code/IaC frameworks (e.g., Terraform) have quickly gained popularity. Unlike conventional tools, IaC~frameworks encode the infrastructure in a "source-of-truth" configuration. They are capable of automatically carrying out modifications to the cloud -- deploying, updating, or destroying resources -- to bring the actual infrastructure into alignment with the IaC configuration. However, when IaC is used alongside consoles, CLIs, or SDKs, it loses visibility into external changes, causing infrastructure drift, where the configuration becomes outdated, and later IaC operations may undo valid updates or trigger errors. We present NSync, an automated system for IaC reconciliation that propagates out-of-band changes back into the IaC program. Our key insight is that infrastructure changes eventually all occur via cloud API invocations -- the lowest layer for cloud management operations. NSync gleans insights from API traces to detect drift (i.e., non-IaC changes) and reconcile it (i.e., update the IaC configuration to capture the changes). It employs an agentic architecture that leverages LLMs to infer high-level intents from noisy API sequences, synthesize targeted IaC updates using specialized tools, and continually improve through a self-evolving knowledge base of past reconciliations. We further introduce a novel evaluation pipeline for injecting realistic drifts into cloud infrastructure and assessing reconciliation performance. Experiments across five real-world Terraform projects and 372 drift scenarios show that NSync outperforms the baseline both in terms of accuracy (from 0.71 to 0.97 pass@3) and token efficiency (1.47$\times$ improvement).

Jin Zhou, Hanmei Yang, Steven et al.

Fine-tuning with Reinforcement Learning with Human Feedback (RLHF) is essential for aligning large language models (LLMs). However, RLHF often encounters significant memory challenges. This study is the first to examine memory usage in the RLHF context, exploring various memory management strategies and unveiling the reasons behind excessive memory consumption. Additionally, we introduce a simple yet effective approach that substantially reduces the memory required for RLHF fine-tuning.

Hanmei Yang, Jin Zhou, Yao Fu et al.

It is extremely memory-hungry to train Large Language Models (LLM). To solve this problem, existing work exploits the combination of CPU and GPU for the training process, such as ZeRO-Offload. Such a technique largely democratizes billion-scale model training, making it possible to train with few consumer graphics cards. However, based on our observation, existing frameworks often provide coarse-grained memory management and require experienced experts in configuration tuning, leading to suboptimal hardware utilization and performance. This paper proposes ProTrain, a novel training system that intelligently balances memory usage and performance by coordinating memory, computation, and IO. ProTrain achieves adaptive memory management through Chunk-Based Model State Management and Block-Wise Activation Management, guided by a Memory-Aware Runtime Profiler without user intervention. ProTrain does not change the training algorithm and thus does not compromise accuracy. Experiments show that ProTrain improves training throughput by 1.43$\times$ to 2.71$\times$ compared to the SOTA training systems.

Yuezhou Sun, Wenlong Zhao, Lijun Zhang et al.

This paper investigates deep neural network (DNN) compression from the perspective of compactly representing and storing trained parameters. We explore the previously overlooked opportunity of cross-layer architecture-agnostic representation sharing for DNN parameters. To do this, we decouple feedforward parameters from DNN architectures and leverage additive quantization, an extreme lossy compression method invented for image descriptors, to compactly represent the parameters. The representations are then finetuned on task objectives to improve task accuracy. We conduct extensive experiments on MobileNet-v2, VGG-11, ResNet-50, Feature Pyramid Networks, and pruned DNNs trained for classification, detection, and segmentation tasks. The conceptually simple scheme consistently outperforms iterative unstructured pruning. Applied to ResNet-50 with 76.1% top-1 accuracy on the ILSVRC12 classification challenge, it achieves a $7.2\times$ compression ratio with no accuracy loss and a $15.3\times$ compression ratio at 74.79% accuracy. Further analyses suggest that representation sharing can frequently happen across network layers and that learning shared representations for an entire DNN can achieve better accuracy at the same compression ratio than compressing the model as multiple separate parts. We release PyTorch code to facilitate DNN deployment on resource-constrained devices and spur future research on efficient representations and storage of DNN parameters.

Sian Jin, Chengming Zhang, Xintong Jiang et al.

Training wide and deep neural networks (DNNs) require large amounts of storage resources such as memory because the intermediate activation data must be saved in the memory during forward propagation and then restored for backward propagation. However, state-of-the-art accelerators such as GPUs are only equipped with very limited memory capacities due to hardware design constraints, which significantly limits the maximum batch size and hence performance speedup when training large-scale DNNs. Traditional memory saving techniques either suffer from performance overhead or are constrained by limited interconnect bandwidth or specific interconnect technology. In this paper, we propose a novel memory-efficient CNN training framework (called COMET) that leverages error-bounded lossy compression to significantly reduce the memory requirement for training, to allow training larger models or to accelerate training. Different from the state-of-the-art solutions that adopt image-based lossy compressors (such as JPEG) to compress the activation data, our framework purposely adopts error-bounded lossy compression with a strict error-controlling mechanism. Specifically, we perform a theoretical analysis on the compression error propagation from the altered activation data to the gradients, and empirically investigate the impact of altered gradients over the training process. Based on these analyses, we optimize the error-bounded lossy compression and propose an adaptive error-bound control scheme for activation data compression. We evaluate our design against state-of-the-art solutions with five widely-adopted CNNs and ImageNet dataset. Experiments demonstrate that our proposed framework can significantly reduce the training memory consumption by up to 13.5X over the baseline training and 1.8X over another state-of-the-art compression-based framework, respectively, with little or no accuracy loss.

Marco Serafini, Hui Guan

Graph Neural Networks (GNNs) are a new and increasingly popular family of deep neural network architectures to perform learning on graphs. Training them efficiently is challenging due to the irregular nature of graph data. The problem becomes even more challenging when scaling to large graphs that exceed the capacity of single devices. Standard approaches to distributed DNN training, such as data and model parallelism, do not directly apply to GNNs. Instead, two different approaches have emerged in the literature: whole-graph and sample-based training. In this paper, we review and compare the two approaches. Scalability is challenging with both approaches, but we make a case that research should focus on sample-based training since it is a more promising approach. Finally, we review recent systems supporting sample-based training.

Lijun Zhang, Xiao Liu, Erik Learned-Miller et al.

When capturing images in low-light conditions, the images often suffer from low visibility, which not only degrades the visual aesthetics of images, but also significantly degenerates the performance of many computer vision algorithms. In this paper, we propose a self-supervised low-light image enhancement framework (SID-NISM), which consists of two components, a Self-supervised Image Decomposition Network (SID-Net) and a Nonlinear Illumination Saturation Mapping function (NISM). As a self-supervised network, SID-Net could decompose the given low-light image into its reflectance, illumination and noise directly without any prior training or reference image, which distinguishes it from existing supervised-learning methods greatly. Then, the decomposed illumination map will be enhanced by NISM. Having the restored illumination map, the enhancement can be achieved accordingly. Experiments on several public challenging low-light image datasets reveal that the images enhanced by SID-NISM are more natural and have less unexpected artifacts.

Hui Guan, Andrey Malevich, Jiyan Yang et al.

Continuous representations have been widely adopted in recommender systems where a large number of entities are represented using embedding vectors. As the cardinality of the entities increases, the embedding components can easily contain millions of parameters and become the bottleneck in both storage and inference due to large memory consumption. This work focuses on post-training 4-bit quantization on the continuous embeddings. We propose row-wise uniform quantization with greedy search and codebook-based quantization that consistently outperforms state-of-the-art quantization approaches on reducing accuracy degradation. We deploy our uniform quantization technique on a production model in Facebook and demonstrate that it can reduce the model size to only 13.89% of the single-precision version while the model quality stays neutral.

Hui Guan, Lin Ning, Zhen Lin et al.

Convolutional Neural Networks (CNN) are being actively explored for safety-critical applications such as autonomous vehicles and aerospace, where it is essential to ensure the reliability of inference results in the presence of possible memory faults. Traditional methods such as error correction codes (ECC) and Triple Modular Redundancy (TMR) are CNN-oblivious and incur substantial memory overhead and energy cost. This paper introduces in-place zero-space ECC assisted with a new training scheme weight distribution-oriented training. The new method provides the first known zero space cost memory protection for CNNs without compromising the reliability offered by traditional ECC.