Wei Niu

Gen Li, Jie Ji, Minghai Qin et al.

As deep convolutional neural networks (DNNs) are widely used in various fields of computer vision, leveraging the overfitting ability of the DNN to achieve video resolution upscaling has become a new trend in the modern video delivery system. By dividing videos into chunks and overfitting each chunk with a super-resolution model, the server encodes videos before transmitting them to the clients, thus achieving better video quality and transmission efficiency. However, a large number of chunks are expected to ensure good overfitting quality, which substantially increases the storage and consumes more bandwidth resources for data transmission. On the other hand, decreasing the number of chunks through training optimization techniques usually requires high model capacity, which significantly slows down execution speed. To reconcile such, we propose a novel method for high-quality and efficient video resolution upscaling tasks, which leverages the spatial-temporal information to accurately divide video into chunks, thus keeping the number of chunks as well as the model size to minimum. Additionally, we advance our method into a single overfitting model by a data-aware joint training technique, which further reduces the storage requirement with negligible quality drop. We deploy our models on an off-the-shelf mobile phone, and experimental results show that our method achieves real-time video super-resolution with high video quality. Compared with the state-of-the-art, our method achieves 28 fps streaming speed with 41.6 PSNR, which is 14$\times$ faster and 2.29 dB better in the live video resolution upscaling tasks. Code available in https://github.com/coulsonlee/STDO-CVPR2023.git

Fei Dou, Jin Ye, Geng Yuan et al.

Artificial General Intelligence (AGI), possessing the capacity to comprehend, learn, and execute tasks with human cognitive abilities, engenders significant anticipation and intrigue across scientific, commercial, and societal arenas. This fascination extends particularly to the Internet of Things (IoT), a landscape characterized by the interconnection of countless devices, sensors, and systems, collectively gathering and sharing data to enable intelligent decision-making and automation. This research embarks on an exploration of the opportunities and challenges towards achieving AGI in the context of the IoT. Specifically, it starts by outlining the fundamental principles of IoT and the critical role of Artificial Intelligence (AI) in IoT systems. Subsequently, it delves into AGI fundamentals, culminating in the formulation of a conceptual framework for AGI's seamless integration within IoT. The application spectrum for AGI-infused IoT is broad, encompassing domains ranging from smart grids, residential environments, manufacturing, and transportation to environmental monitoring, agriculture, healthcare, and education. However, adapting AGI to resource-constrained IoT settings necessitates dedicated research efforts. Furthermore, the paper addresses constraints imposed by limited computing resources, intricacies associated with large-scale IoT communication, as well as the critical concerns pertaining to security and privacy.

Zifeng Wang, Zheng Zhan, Yifan Gong et al.

Existing work in continual learning (CL) focuses on mitigating catastrophic forgetting, i.e., model performance deterioration on past tasks when learning a new task. However, the training efficiency of a CL system is under-investigated, which limits the real-world application of CL systems under resource-limited scenarios. In this work, we propose a novel framework called Sparse Continual Learning(SparCL), which is the first study that leverages sparsity to enable cost-effective continual learning on edge devices. SparCL achieves both training acceleration and accuracy preservation through the synergy of three aspects: weight sparsity, data efficiency, and gradient sparsity. Specifically, we propose task-aware dynamic masking (TDM) to learn a sparse network throughout the entire CL process, dynamic data removal (DDR) to remove less informative training data, and dynamic gradient masking (DGM) to sparsify the gradient updates. Each of them not only improves efficiency, but also further mitigates catastrophic forgetting. SparCL consistently improves the training efficiency of existing state-of-the-art (SOTA) CL methods by at most 23X less training FLOPs, and, surprisingly, further improves the SOTA accuracy by at most 1.7%. SparCL also outperforms competitive baselines obtained from adapting SOTA sparse training methods to the CL setting in both efficiency and accuracy. We also evaluate the effectiveness of SparCL on a real mobile phone, further indicating the practical potential of our method.

Zheng Zhan, Zhenglun Kong, Yifan Gong et al. · harvard

State Space Models (SSMs) have the advantage of keeping linear computational complexity compared to attention modules in transformers, and have been applied to vision tasks as a new type of powerful vision foundation model. Inspired by the observations that the final prediction in vision transformers (ViTs) is only based on a subset of most informative tokens, we take the novel step of enhancing the efficiency of SSM-based vision models through token-based pruning. However, direct applications of existing token pruning techniques designed for ViTs fail to deliver good performance, even with extensive fine-tuning. To address this issue, we revisit the unique computational characteristics of SSMs and discover that naive application disrupts the sequential token positions. This insight motivates us to design a novel and general token pruning method specifically for SSM-based vision models. We first introduce a pruning-aware hidden state alignment method to stabilize the neighborhood of remaining tokens for performance enhancement. Besides, based on our detailed analysis, we propose a token importance evaluation method adapted for SSM models, to guide the token pruning. With efficient implementation and practical acceleration methods, our method brings actual speedup. Extensive experiments demonstrate that our approach can achieve significant computation reduction with minimal impact on performance across different tasks. Notably, we achieve 81.7\% accuracy on ImageNet with a 41.6\% reduction in the FLOPs for pruned PlainMamba-L3. Furthermore, our work provides deeper insights into understanding the behavior of SSM-based vision models for future research.

Yushu Wu, Yifan Gong, Pu Zhao et al.

Deep learning-based super-resolution (SR) has gained tremendous popularity in recent years because of its high image quality performance and wide application scenarios. However, prior methods typically suffer from large amounts of computations and huge power consumption, causing difficulties for real-time inference, especially on resource-limited platforms such as mobile devices. To mitigate this, we propose a compiler-aware SR neural architecture search (NAS) framework that conducts depth search and per-layer width search with adaptive SR blocks. The inference speed is directly taken into the optimization along with the SR loss to derive SR models with high image quality while satisfying the real-time inference requirement. Instead of measuring the speed on mobile devices at each iteration during the search process, a speed model incorporated with compiler optimizations is leveraged to predict the inference latency of the SR block with various width configurations for faster convergence. With the proposed framework, we achieve real-time SR inference for implementing 720p resolution with competitive SR performance (in terms of PSNR and SSIM) on GPU/DSP of mobile platforms (Samsung Galaxy S21).

Jou-An Chen, Wei Niu, Bin Ren et al.

Data redundancy is ubiquitous in the inputs and intermediate results of Deep Neural Networks (DNN). It offers many significant opportunities for improving DNN performance and efficiency and has been explored in a large body of work. These studies have scattered in many venues across several years. The targets they focus on range from images to videos and texts, and the techniques they use to detect and exploit data redundancy also vary in many aspects. There is not yet a systematic examination and summary of the many efforts, making it difficult for researchers to get a comprehensive view of the prior work, the state of the art, differences and shared principles, and the areas and directions yet to explore. This article tries to fill the void. It surveys hundreds of recent papers on the topic, introduces a novel taxonomy to put the various techniques into a single categorization framework, offers a comprehensive description of the main methods used for exploiting data redundancy in improving multiple kinds of DNNs on data, and points out a set of research opportunities for future to explore.

Jun Liu, Pu Zhao, Zhenglun Kong et al. · harvard

Embodied robotic systems increasingly rely on large language model (LLM)-based agents to support high-level reasoning, planning, and decision-making during interactions with the environment. However, invoking LLM reasoning introduces substantial computational latency and resource overhead, which can interrupt action execution and reduce system reliability. Excessive reasoning may delay actions, while insufficient reasoning often leads to incorrect decisions and task failures. This raises a fundamental question for embodied agents: when should the agent reason, and when should it act? In this work, we propose RARRL (Resource-Aware Reasoning via Reinforcement Learning), a hierarchical framework for resource-aware orchestration of embodied agents. Rather than learning low-level control policies, RARRL learns a high-level orchestration policy that operates at the agent's decision-making layer. This policy enables the agent to adaptively determine whether to invoke reasoning, which reasoning role to employ, and how much computational budget to allocate based on current observations, execution history, and remaining resources. Extensive experiments, including evaluations with empirical latency profiles derived from the ALFRED benchmark, show that RARRL consistently improves task success rates while reducing execution latency and enhancing robustness compared with fixed or heuristic reasoning strategies. These results demonstrate that adaptive reasoning control is essential for building reliable and efficient embodied robotic agents.

Gen Li, Zhihao Shu, Jie Ji et al.

Deep neural networks (DNNs) are frequently employed in a variety of computer vision applications. Nowadays, an emerging trend in the current video distribution system is to take advantage of DNN's overfitting properties to perform video resolution upscaling. By splitting videos into chunks and applying a super-resolution (SR) model to overfit each chunk, this scheme of SR models plus video chunks is able to replace traditional video transmission to enhance video quality and transmission efficiency. However, many models and chunks are needed to guarantee high performance, which leads to tremendous overhead on model switching and memory footprints at the user end. To resolve such problems, we propose a Dynamic Deep neural network assisted by a Content-Aware data processing pipeline to reduce the model number down to one (Dy-DCA), which helps promote performance while conserving computational resources. Additionally, to achieve real acceleration on the user end, we designed a framework that optimizes dynamic features (e.g., dynamic shapes, sizes, and control flow) in Dy-DCA to enable a series of compilation optimizations, including fused code generation, static execution planning, etc. By employing such techniques, our method achieves better PSNR and real-time performance (33 FPS) on an off-the-shelf mobile phone. Meanwhile, assisted by our compilation optimization, we achieve a 1.7$\times$ speedup while saving up to 1.61$\times$ memory consumption. Code available in https://github.com/coulsonlee/Dy-DCA-ECCV2024.

Wenqi Jia, Zhewen Hu, Youyuan Liu et al.

Large-scale scientific simulations generate massive datasets, posing challenges for storage and I/O. Traditional lossy compression struggles to advance more in balancing compression ratio, data quality, and adaptability to diverse scientific data features. While deep learning-based solutions have been explored, their common practice of relying on large models and offline training limits adaptability to dynamic data characteristics and computational efficiency. To address these challenges, we propose NeurLZ, a neural method designed to enhance lossy compression by integrating online learning, cross-field learning, and robust error regulation. Key innovations of NeurLZ include: (1) compression-time online neural learning with lightweight skipping DNN models, adapting to residual errors without costly offline pertaining, (2) the error-mitigating capability, recovering fine details from compression errors overlooked by conventional compressors, (3) $1\times$ and $2\times$ error-regulation modes, ensuring strict adherence to $1\times$ user-input error bounds strictly or relaxed 2$\times$ bounds for better overall quality, and (4) cross-field learning leveraging inter-field correlations in scientific data to improve conventional methods. Comprehensive evaluations on representative HPC datasets, e.g., Nyx, Miranda, Hurricane, against state-of-the-art compressors show NeurLZ's effectiveness. During the first five learning epochs, NeurLZ achieves an 89% bit rate reduction, with further optimization yielding up to around 94% reduction at equivalent distortion, significantly outperforming existing methods, demonstrating NeurLZ's superior performance in enhancing scientific lossy compression as a scalable and efficient solution.

Xuan Shen, Zhao Song, Yufa Zhou et al.

Diffusion Transformers have emerged as the preeminent models for a wide array of generative tasks, demonstrating superior performance and efficacy across various applications. The promising results come at the cost of slow inference, as each denoising step requires running the whole transformer model with a large amount of parameters. In this paper, we show that performing the full computation of the model at each diffusion step is unnecessary, as some computations can be skipped by lazily reusing the results of previous steps. Furthermore, we show that the lower bound of similarity between outputs at consecutive steps is notably high, and this similarity can be linearly approximated using the inputs. To verify our demonstrations, we propose the \textbf{LazyDiT}, a lazy learning framework that efficiently leverages cached results from earlier steps to skip redundant computations. Specifically, we incorporate lazy learning layers into the model, effectively trained to maximize laziness, enabling dynamic skipping of redundant computations. Experimental results show that LazyDiT outperforms the DDIM sampler across multiple diffusion transformer models at various resolutions. Furthermore, we implement our method on mobile devices, achieving better performance than DDIM with similar latency. Code: https://github.com/shawnricecake/lazydit

Yanyu Li, Xuan Shen, Geng Yuan et al.

In this work we demonstrate real-time portrait stylization, specifically, translating self-portrait into cartoon or anime style on mobile devices. We propose a latency-driven differentiable architecture search method, maintaining realistic generative quality. With our framework, we obtain $10\times$ computation reduction on the generative model and achieve real-time video stylization on off-the-shelf smartphone using mobile GPUs.

Yanjun Lin, Zimo Xiao, Kartik Natarajan et al.

Task-oriented dialogue systems -- handling transactions, reservations, and service requests -- require predictable behavior, yet the moderately-sized LLMs needed for practical latency are prone to hallucination and format errors that cascade into incorrect actions (e.g., a hotel booked for the wrong date). We propose ReacTOD, a bounded neuro-symbolic architecture that reformulates NLU as discrete tool calls within a self-correcting ReAct loop governed by deterministic validation. A bounded ReAct loop enables iterative self-correction, improving accuracy by up to 9.3 percentage points over single-pass inference on MultiWOZ. A symbolic validator enforces action compliance, schema conformance, and coreference consistency on every dialogue state update, achieving a 93.1% self-correction rate on intercepted errors and producing structured execution traces. Incremental state prediction and on-demand history retrieval keep prompts compact, empirically improving instruction adherence in parameter-constrained models. On MultiWOZ 2.1, ReacTOD achieves a new zero-shot state-of-the-art: gpt-oss-20B reaches 52.71% joint goal accuracy, surpassing the previous best by 14 percentage points, while Qwen3-8B achieves 47.34% with only 8B parameters. On the Schema-Guided Dialogue (SGD) benchmark, ReacTOD with Claude-Opus-4.6 achieves 80.68% JGA under fully end-to-end evaluation with predicted domains, and Qwen3-32B reaches 64.09% -- demonstrating cross-benchmark generalization without task-specific training data.

Xuan Shen, Weize Ma, Jing Liu et al.

Monocular Depth Estimation (MDE) has emerged as a pivotal task in computer vision, supporting numerous real-world applications. However, deploying accurate depth estimation models on resource-limited edge devices, especially Application-Specific Integrated Circuits (ASICs), is challenging due to the high computational and memory demands. Recent advancements in foundational depth estimation deliver impressive results but further amplify the difficulty of deployment on ASICs. To address this, we propose QuartDepth which adopts post-training quantization to quantize MDE models with hardware accelerations for ASICs. Our approach involves quantizing both weights and activations to 4-bit precision, reducing the model size and computation cost. To mitigate the performance degradation, we introduce activation polishing and compensation algorithm applied before and after activation quantization, as well as a weight reconstruction method for minimizing errors in weight quantization. Furthermore, we design a flexible and programmable hardware accelerator by supporting kernel fusion and customized instruction programmability, enhancing throughput and efficiency. Experimental results demonstrate that our framework achieves competitive accuracy while enabling fast inference and higher energy efficiency on ASICs, bridging the gap between high-performance depth estimation and practical edge-device applicability. Code: https://github.com/shawnricecake/quart-depth

Yuqing Zhou, Zhuoer Wang, Jie Yuan et al.

Large language model (LLM)-based agents are widely deployed in user-facing services but remain error-prone in new tasks, tend to repeat the same failure patterns, and show substantial run-to-run variability. Fixing failures via environment-specific training or manual patching is costly and hard to scale. To enable self-evolving agents in user-facing service environments, we propose WISE-Flow, a workflow-centric framework that converts historical service interactions into reusable procedural experience by inducing workflows with prerequisite-augmented action blocks. At deployment, WISE-Flow aligns the agent's execution trajectory to retrieved workflows and performs prerequisite-aware feasibility reasoning to achieve state-grounded next actions. Experiments on ToolSandbox and $τ^2$-bench show consistent improvement across base models.

Yangming Zhang, Jian Xu, Kunxiong Zhu et al.

3D Gaussian Splatting (3DGS) has revolutionized novel view synthesis with high-quality rendering through continuous aggregations of millions of 3D Gaussian primitives. However, it suffers from a substantial memory footprint, particularly during training due to uncontrolled densification, posing a critical bottleneck for deployment on memory-constrained edge devices. While existing methods prune redundant Gaussians post-training, they fail to address the peak memory spikes caused by the abrupt growth of Gaussians early in the training process. To solve the training memory consumption problem, we propose a systematic memory-bounded training framework that dynamically optimizes Gaussians through iterative growth and pruning. In other words, the proposed framework alternates between incremental pruning of low-impact Gaussians and strategic growing of new primitives with an adaptive Gaussian compensation, maintaining a near-constant low memory usage while progressively refining rendering fidelity. We comprehensively evaluate the proposed training framework on various real-world datasets under strict memory constraints, showing significant improvements over existing state-of-the-art methods. Particularly, our proposed method practically enables memory-efficient 3DGS training on NVIDIA Jetson AGX Xavier, achieving similar visual quality with up to 80% lower peak training memory consumption than the original 3DGS.

Jingxu Ng, Cheng Lv, Pu Zhao et al.

Stable diffusion plays a crucial role in generating high-quality images. However, image generation is time-consuming and memory-intensive. To address this, stable-diffusion.cpp (Sdcpp) emerges as an efficient inference framework to accelerate the diffusion models. Although it is lightweight, the current implementation of ggml_conv_2d operator in Sdcpp is suboptimal, exhibiting both high inference latency and massive memory usage. To address this, in this work, we present an optimized version of Sdcpp leveraging the Winograd algorithm to accelerate 2D convolution operations, which is the primary bottleneck in the pipeline. By analyzing both dependent and independent computation graphs, we exploit the device's locality and parallelism to achieve substantial performance improvements. Our framework delivers correct end-to-end results across various stable diffusion models, including SDv1.4, v1.5, v2.1, SDXL, and SDXL-Turbo. Our evaluation results demonstrate a speedup up to 2.76x for individual convolutional layers and an inference speedup up to 4.79x for the overall image generation process, compared with the original Sdcpp on M1 pro. Homepage: https://github.com/SealAILab/stable-diffusion-cpp

Geng Yuan, Xiaolong Ma, Wei Niu et al.

Recently, a new trend of exploring sparsity for accelerating neural network training has emerged, embracing the paradigm of training on the edge. This paper proposes a novel Memory-Economic Sparse Training (MEST) framework targeting for accurate and fast execution on edge devices. The proposed MEST framework consists of enhancements by Elastic Mutation (EM) and Soft Memory Bound (&S) that ensure superior accuracy at high sparsity ratios. Different from the existing works for sparse training, this current work reveals the importance of sparsity schemes on the performance of sparse training in terms of accuracy as well as training speed on real edge devices. On top of that, the paper proposes to employ data efficiency for further acceleration of sparse training. Our results suggest that unforgettable examples can be identified in-situ even during the dynamic exploration of sparsity masks in the sparse training process, and therefore can be removed for further training speedup on edge devices. Comparing with state-of-the-art (SOTA) works on accuracy, our MEST increases Top-1 accuracy significantly on ImageNet when using the same unstructured sparsity scheme. Systematical evaluation on accuracy, training speed, and memory footprint are conducted, where the proposed MEST framework consistently outperforms representative SOTA works. A reviewer strongly against our work based on his false assumptions and misunderstandings. On top of the previous submission, we employ data efficiency for further acceleration of sparse training. And we explore the impact of model sparsity, sparsity schemes, and sparse training algorithms on the number of removable training examples. Our codes are publicly available at: https://github.com/boone891214/MEST.

Yuxuan Cai, Hongjia Li, Geng Yuan et al.

The rapid development and wide utilization of object detection techniques have aroused attention on both accuracy and speed of object detectors. However, the current state-of-the-art object detection works are either accuracy-oriented using a large model but leading to high latency or speed-oriented using a lightweight model but sacrificing accuracy. In this work, we propose YOLObile framework, a real-time object detection on mobile devices via compression-compilation co-design. A novel block-punched pruning scheme is proposed for any kernel size. To improve computational efficiency on mobile devices, a GPU-CPU collaborative scheme is adopted along with advanced compiler-assisted optimizations. Experimental results indicate that our pruning scheme achieves 14$\times$ compression rate of YOLOv4 with 49.0 mAP. Under our YOLObile framework, we achieve 17 FPS inference speed using GPU on Samsung Galaxy S20. By incorporating our proposed GPU-CPU collaborative scheme, the inference speed is increased to 19.1 FPS, and outperforms the original YOLOv4 by 5$\times$ speedup. Source code is at: \url{https://github.com/nightsnack/YOLObile}.

Yun He, Jianling Wang, Wei Niu et al.

User-generated item lists are a popular feature of many different platforms. Examples include lists of books on Goodreads, playlists on Spotify and YouTube, collections of images on Pinterest, and lists of answers on question-answer sites like Zhihu. Recommending item lists is critical for increasing user engagement and connecting users to new items, but many approaches are designed for the item-based recommendation, without careful consideration of the complex relationships between items and lists. Hence, in this paper, we propose a novel user-generated list recommendation model called AttList. Two unique features of AttList are careful modeling of (i) hierarchical user preference, which aggregates items to characterize the list that they belong to, and then aggregates these lists to estimate the user preference, naturally fitting into the hierarchical structure of item lists; and (ii) item and list consistency, through a novel self-attentive aggregation layer designed for capturing the consistency of neighboring items and lists to better model user preference. Through experiments over three real-world datasets reflecting different kinds of user-generated item lists, we find that AttList results in significant improvements in NDCG, Precision@k, and Recall@k versus a suite of state-of-the-art baselines. Furthermore, all code and data are available at https://github.com/heyunh2015/AttList.

Zhihao Shu, Md Musfiqur Rahman Sanim, Hangyu Zheng et al.

The increasing size and complexity of modern deep neural networks (DNNs) pose significant challenges for on-device inference on mobile GPUs, with limited memory and computational resources. Existing DNN acceleration frameworks primarily deploy a weight preloading strategy, where all model parameters are loaded into memory before execution on mobile GPUs. We posit that this approach is not adequate for modern DNN workloads that comprise very large model(s) and possibly execution of several distinct models in succession. In this work, we introduce FlashMem, a memory streaming framework designed to efficiently execute large-scale modern DNNs and multi-DNN workloads while minimizing memory consumption and reducing inference latency. Instead of fully preloading weights, FlashMem statically determines model loading schedules and dynamically streams them on demand, leveraging 2.5D texture memory to minimize data transformations and improve execution efficiency. Experimental results on 11 models demonstrate that FlashMem achieves 2.0x to 8.4x memory reduction and 1.7x to 75.0x speedup compared to existing frameworks, enabling efficient execution of large-scale models and multi-DNN support on resource-constrained mobile GPUs.

Kaiyuan Deng, Hangyu Zheng, Minghai Qing et al.

Deploying models, especially large language models (LLMs), is becoming increasingly attractive to a broader user base, including those without specialized expertise. However, due to the resource constraints of certain hardware, maintaining high accuracy with larger model while meeting the hardware requirements remains a significant challenge. Model quantization technique helps mitigate memory and compute bottlenecks, yet the added complexities of tuning and deploying quantized models further exacerbates these challenges, making the process unfriendly to most of the users. We introduce the Hardware-Aware Quantization Agent (HAQA), an automated framework that leverages LLMs to streamline the entire quantization and deployment process by enabling efficient hyperparameter tuning and hardware configuration, thereby simultaneously improving deployment quality and ease of use for a broad range of users. Our results demonstrate up to a 2.3x speedup in inference, along with increased throughput and improved accuracy compared to unoptimized models on Llama. Additionally, HAQA is designed to implement adaptive quantization strategies across diverse hardware platforms, as it automatically finds optimal settings even when they appear counterintuitive, thereby reducing extensive manual effort and demonstrating superior adaptability. Code will be released.

Ruitong Sun, Tianze Yang, Wei Niu et al.

Diffusion models have emerged as the dominant paradigm for high-quality image generation, yet their computational expense remains substantial due to iterative denoising. Classifier-Free Guidance (CFG) significantly enhances generation quality and controllability but doubles the computation by requiring both conditional and unconditional forward passes at every timestep. We present OUSAC (Optimized gUidance Scheduling with Adaptive Caching), a framework that accelerates diffusion transformers (DiT) through systematic optimization. Our key insight is that variable guidance scales enable sparse computation: adjusting scales at certain timesteps can compensate for skipping CFG at others, enabling both fewer total sampling steps and fewer CFG steps while maintaining quality. However, variable guidance patterns introduce denoising deviations that undermine standard caching methods, which assume constant CFG scales across steps. Moreover, different transformer blocks are affected at different levels under dynamic conditions. This paper develops a two-stage approach leveraging these insights. Stage-1 employs evolutionary algorithms to jointly optimize which timesteps to skip and what guidance scale to use, eliminating up to 82% of unconditional passes. Stage-2 introduces adaptive rank allocation that tailors calibration efforts per transformer block, maintaining caching effectiveness under variable guidance. Experiments demonstrate that OUSAC significantly outperforms state-of-the-art acceleration methods, achieving 53% computational savings with 15% quality improvement on DiT-XL/2 (ImageNet 512x512), 60% savings with 16.1% improvement on PixArt-alpha (MSCOCO), and 5x speedup on FLUX while improving CLIP Score over the 50-step baseline.

Zheng Zhan, Yushu Wu, Yifan Gong et al. · harvard

The rapid progress in artificial intelligence-generated content (AIGC), especially with diffusion models, has significantly advanced development of high-quality video generation. However, current video diffusion models exhibit demanding computational requirements and high peak memory usage, especially for generating longer and higher-resolution videos. These limitations greatly hinder the practical application of video diffusion models on standard hardware platforms. To tackle this issue, we present a novel, training-free framework named Streamlined Inference, which leverages the temporal and spatial properties of video diffusion models. Our approach integrates three core components: Feature Slicer, Operator Grouping, and Step Rehash. Specifically, Feature Slicer effectively partitions input features into sub-features and Operator Grouping processes each sub-feature with a group of consecutive operators, resulting in significant memory reduction without sacrificing the quality or speed. Step Rehash further exploits the similarity between adjacent steps in diffusion, and accelerates inference through skipping unnecessary steps. Extensive experiments demonstrate that our approach significantly reduces peak memory and computational overhead, making it feasible to generate high-quality videos on a single consumer GPU (e.g., reducing peak memory of AnimateDiff from 42GB to 11GB, featuring faster inference on 2080Ti).

Wei Niu, Md Musfiqur Rahman Sanim, Zhihao Shu et al.

This work is motivated by recent developments in Deep Neural Networks, particularly the Transformer architectures underlying applications such as ChatGPT, and the need for performing inference on mobile devices. Focusing on emerging transformers (specifically the ones with computationally efficient Swin-like architectures) and large models (e.g., Stable Diffusion and LLMs) based on transformers, we observe that layout transformations between the computational operators cause a significant slowdown in these applications. This paper presents SmartMem, a comprehensive framework for eliminating most layout transformations, with the idea that multiple operators can use the same tensor layout through careful choice of layout and implementation of operations. Our approach is based on classifying the operators into four groups, and considering combinations of producer-consumer edges between the operators. We develop a set of methods for searching such layouts. Another component of our work is developing efficient memory layouts for 2.5 dimensional memory commonly seen in mobile devices. Our experimental results show that SmartMem outperforms 5 state-of-the-art DNN execution frameworks on mobile devices across 18 varied neural networks, including CNNs, Transformers with both local and global attention, as well as LLMs. In particular, compared to DNNFusion, SmartMem achieves an average speedup of 2.8$\times$, and outperforms TVM and MNN with speedups of 6.9$\times$ and 7.9$\times$, respectively, on average.

Jun Liu, Zhenglun Kong, Pu Zhao et al. · harvard

Structured pruning for large language models (LLMs) has garnered significant academic interest due to its ability to efficiently compress and accelerate LLMs by eliminating redundant weight groups at a coarse-grained granularity. Current structured pruning methods for LLMs typically depend on a singular granularity for assessing weight importance, resulting in notable performance degradation in downstream tasks. Intriguingly, our empirical investigations reveal that utilizing unstructured pruning, which achieves better performance retention by pruning weights at a finer granularity, \emph{i.e.}, individual weights, yields significantly varied sparse LLM structures when juxtaposed to structured pruning. This suggests that evaluating both holistic and individual assessment for weight importance is essential for LLM pruning. Building on this insight, we introduce the Hybrid-grained Weight Importance Assessment (HyWIA), a novel method that merges fine-grained and coarse-grained evaluations of weight importance for the pruning of LLMs. Leveraging an attention mechanism, HyWIA adaptively determines the optimal blend of granularity in weight importance assessments in an end-to-end pruning manner. Extensive experiments on LLaMA-V1/V2, Vicuna, Baichuan, and Bloom across various benchmarks demonstrate the effectiveness of HyWIA in pruning LLMs. For example, HyWIA surpasses the cutting-edge LLM-Pruner by an average margin of 2.82% in accuracy across seven downstream tasks when pruning LLaMA-7B by 50%.

Yangming Zhang, Wenqi Jia, Wei Niu et al.

3D Gaussian Splatting (3DGS) has emerged as a mainstream for novel view synthesis, leveraging continuous aggregations of Gaussian functions to model scene geometry. However, 3DGS suffers from substantial memory requirements to store the multitude of Gaussians, hindering its practicality. To address this challenge, we introduce GaussianSpa, an optimization-based simplification framework for compact and high-quality 3DGS. Specifically, we formulate the simplification as an optimization problem associated with the 3DGS training. Correspondingly, we propose an efficient "optimizing-sparsifying" solution that alternately solves two independent sub-problems, gradually imposing strong sparsity onto the Gaussians in the training process. Our comprehensive evaluations on various datasets show the superiority of GaussianSpa over existing state-of-the-art approaches. Notably, GaussianSpa achieves an average PSNR improvement of 0.9 dB on the real-world Deep Blending dataset with 10$\times$ fewer Gaussians compared to the vanilla 3DGS. Our project page is available at https://noodle-lab.github.io/gaussianspa/.

Jun Liu, Zhenglun Kong, Peiyan Dong et al. · harvard

Fine-tuning helps large language models (LLM) recover degraded information and enhance task performance. Although Low-Rank Adaptation (LoRA) is widely used and effective for fine-tuning, we have observed that its scaling factor can limit or even reduce performance as the rank size increases. To address this issue, we propose RoRA (Rank-adaptive Reliability Optimization), a simple yet effective method for optimizing LoRA's scaling factor. By replacing $α/r$ with $α/\sqrt{r}$, RoRA ensures improved performance as rank size increases. Moreover, RoRA enhances low-rank adaptation in fine-tuning uncompressed models and excels in the more challenging task of accuracy recovery when fine-tuning pruned models. Extensive experiments demonstrate the effectiveness of RoRA in fine-tuning both uncompressed and pruned models. RoRA surpasses the state-of-the-art (SOTA) in average accuracy and robustness on LLaMA-7B/13B, LLaMA2-7B, and LLaMA3-8B, specifically outperforming LoRA and DoRA by 6.5% and 2.9% on LLaMA-7B, respectively. In pruned model fine-tuning, RoRA shows significant advantages; for SHEARED-LLAMA-1.3, a LLaMA-7B with 81.4% pruning, RoRA achieves 5.7% higher average accuracy than LoRA and 3.9% higher than DoRA.

Jun Liu, Zhenglun Kong, Pu Zhao et al. · harvard

Autonomous driving platforms encounter diverse driving scenarios, each with varying hardware resources and precision requirements. Given the computational limitations of embedded devices, it is crucial to consider computing costs when deploying on target platforms like the NVIDIA\textsuperscript{\textregistered} DRIVE PX 2. Our objective is to customize the semantic segmentation network according to the computing power and specific scenarios of autonomous driving hardware. We implement dynamic adaptability through a three-tier control mechanism -- width multiplier, classifier depth, and classifier kernel -- allowing fine-grained control over model components based on hardware constraints and task requirements. This adaptability facilitates broad model scaling, targeted refinement of the final layers, and scenario-specific optimization of kernel sizes, leading to improved resource allocation and performance. Additionally, we leverage Bayesian Optimization with surrogate modeling to efficiently explore hyperparameter spaces under tight computational budgets. Our approach addresses scenario-specific and task-specific requirements through automatic parameter search, accommodating the unique computational complexity and accuracy needs of autonomous driving. It scales its Multiply-Accumulate Operations (MACs) for Task-Specific Learning Adaptation (TSLA), resulting in alternative configurations tailored to diverse self-driving tasks. These TSLA customizations maximize computational capacity and model accuracy, optimizing hardware utilization.

Wei Niu, Gagan Agrawal, Bin Ren

Though many compilation and runtime systems have been developed for DNNs in recent years, the focus has largely been on static DNNs. Dynamic DNNs, where tensor shapes and sizes and even the set of operators used are dependent upon the input and/or execution, are becoming common. This paper presents SoD$^2$, a comprehensive framework for optimizing Dynamic DNNs. The basis of our approach is a classification of common operators that form DNNs, and the use of this classification towards a Rank and Dimension Propagation (RDP) method. This framework statically determines the shapes of operators as known constants, symbolic constants, or operations on these. Next, using RDP we enable a series of optimizations, like fused code generation, execution (order) planning, and even runtime memory allocation plan generation. By evaluating the framework on 10 emerging Dynamic DNNs and comparing it against several existing systems, we demonstrate both reductions in execution latency and memory requirements, with RDP-enabled key optimizations responsible for much of the gains. Our evaluation results show that SoD$^2$ runs up to $3.9\times$ faster than these systems while saving up to $88\%$ peak memory consumption.

Wenqi Jia, Sian Jin, Jinzhen Wang et al.

The rapid expansion of computational capabilities and the ever-growing scale of modern HPC systems present formidable challenges in managing exascale scientific data. Faced with such vast datasets, traditional lossless compression techniques prove insufficient in reducing data size to a manageable level while preserving all information intact. In response, researchers have turned to error-bounded lossy compression methods, which offer a balance between data size reduction and information retention. However, despite their utility, these compressors employing conventional techniques struggle with limited reconstruction quality. To address this issue, we draw inspiration from recent advancements in deep learning and propose GWLZ, a novel group-wise learning-based lossy compression framework with multiple lightweight learnable enhancer models. Leveraging a group of neural networks, GWLZ significantly enhances the decompressed data reconstruction quality with negligible impact on the compression efficiency. Experimental results on different fields from the Nyx dataset demonstrate remarkable improvements by GWLZ, achieving up to 20% quality enhancements with negligible overhead as low as 0.0003x.

Jun Liu, Zhenglun Kong, Changdi Yang et al. · harvard

Multi-agent large language model (LLM) systems have shown strong potential in complex reasoning and collaborative decision-making tasks. However, most existing coordination schemes rely on static or full-context routing strategies, which lead to excessive token consumption, redundant memory exposure, and limited adaptability across interaction rounds. We introduce RCR-Router, a modular and role-aware context routing framework designed to enable efficient, adaptive collaboration in multi-agent LLMs. To our knowledge, this is the first routing approach that dynamically selects semantically relevant memory subsets for each agent based on its role and task stage, while adhering to a strict token budget. A lightweight scoring policy guides memory selection, and agent outputs are iteratively integrated into a shared memory store to facilitate progressive context refinement. To better evaluate model behavior, we further propose an Answer Quality Score metric that captures LLM-generated explanations beyond standard QA accuracy. Experiments on three multi-hop QA benchmarks -- HotPotQA, MuSiQue, and 2WikiMultihop -- demonstrate that RCR-Router reduces token usage (up to 30%) while improving or maintaining answer quality. These results highlight the importance of structured memory routing and output-aware evaluation in advancing scalable multi-agent LLM systems.

Jun Liu, Zhenglun Kong, Peiyan Dong et al. · harvard

Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into {[REASON]} and {[ACT]} spans, applying segment-specific losses to align each component with the teacher's behavior. This structure-aware supervision enables compact agents to better replicate the teacher's decision process. Experiments on ALFWorld, HotPotQA-ReAct, and WebShop show that our approach consistently outperforms token-level and imitation learning baselines, achieving significant compression with minimal performance drop. Scaling and ablation results further highlight the importance of span-level alignment for efficient and deployable agents.

Yuanbin Man, Ying Huang, Chengming Zhang et al.

The advancements in large language models (LLMs) have propelled the improvement of video understanding tasks by incorporating LLMs with visual models. However, most existing LLM-based models (e.g., VideoLLaMA, VideoChat) are constrained to processing short-duration videos. Recent attempts to understand long-term videos by extracting and compressing visual features into a fixed memory size. Nevertheless, those methods leverage only visual modality to merge video tokens and overlook the correlation between visual and textual queries, leading to difficulties in effectively handling complex question-answering tasks. To address the challenges of long videos and complex prompts, we propose AdaCM$^2$, which, for the first time, introduces an adaptive cross-modality memory reduction approach to video-text alignment in an auto-regressive manner on video streams. Our extensive experiments on various video understanding tasks, such as video captioning, video question answering, and video classification, demonstrate that AdaCM$^2$ achieves state-of-the-art performance across multiple datasets while significantly reducing memory usage. Notably, it achieves a 4.5% improvement across multiple tasks in the LVU dataset with a GPU memory consumption reduction of up to 65%.

Xuanqing Liu, Luyang Kong, Wei Niu et al.

Large language models (LLMs) have demonstrated remarkable capabilities in handling complex dialogue tasks without requiring use case-specific fine-tuning. However, analyzing live dialogues in real-time necessitates low-latency processing systems, making it impractical to deploy models with billions of parameters due to latency constraints. As a result, practitioners often prefer smaller models with millions of parameters, trained on high-quality, human-annotated datasets. Yet, curating such datasets is both time-consuming and costly. Consequently, there is a growing need to combine the scalability of LLM-generated labels with the precision of human annotations, enabling fine-tuned smaller models to achieve both higher speed and accuracy comparable to larger models. In this paper, we introduce a simple yet effective framework to address this challenge. Our approach is specifically designed for per-utterance classification problems, which encompass tasks such as intent detection, dialogue state tracking, and more. To mitigate the impact of labeling errors from LLMs -- the primary source of inaccuracies in student models -- we propose a noise-reduced preference learning loss. Experimental results demonstrate that our method significantly improves accuracy across utterance-level dialogue tasks, including sentiment detection (over $2\%$), dialogue act classification (over $1.5\%$), etc.

Md Musfiqur Rahman Sanim, Zhihao Shu, Bahram Afsharmanesh et al.

Image-based 3D scene reconstruction, which transforms multi-view images into a structured 3D representation of the surrounding environment, is a common task across many modern applications. 3D Gaussian Splatting (3DGS) is a new paradigm to address this problem and offers considerable efficiency as compared to the previous methods. Motivated by this, and considering various benefits of mobile device deployment (data privacy, operating without internet connectivity, and potentially faster responses), this paper develops Texture3dgs, an optimized mapping of 3DGS for a mobile GPU. A critical challenge in this area turns out to be optimizing for the two-dimensional (2D) texture cache, which needs to be exploited for faster executions on mobile GPUs. As a sorting method dominates the computations in 3DGS on mobile platforms, the core of Texture3dgs is a novel sorting algorithm where the processing, data movement, and placement are highly optimized for 2D memory. The properties of this algorithm are analyzed in view of a cost model for the texture cache. In addition, we accelerate other steps of the 3DGS algorithm through improved variable layout design and other optimizations. End-to-end evaluation shows that Texture3dgs delivers up to 4.1$\times$ and 1.7$\times$ speedup for the sorting and overall 3D scene reconstruction, respectively -- while also reducing memory usage by up to 1.6$\times$ -- demonstrating the effectiveness of our design for efficient mobile 3D scene reconstruction.

Wentian Zhu, Zhen Xiang, Wei Niu et al.

Unlike regular tokens derived from existing text corpora, special tokens are artificially created to annotate structured conversations during the fine-tuning process of Large Language Models (LLMs). Serving as metadata of training data, these tokens play a crucial role in instructing LLMs to generate coherent and context-aware responses. We demonstrate that special tokens can be exploited to construct four attack primitives, with which malicious users can reliably bypass the internal safety alignment of online LLM services and circumvent state-of-the-art (SOTA) external content moderation systems simultaneously. Moreover, we found that addressing this threat is challenging, as aggressive defense mechanisms-such as input sanitization by removing special tokens entirely, as suggested in academia-are less effective than anticipated. This is because such defense can be evaded when the special tokens are replaced by regular ones with high semantic similarity within the tokenizer's embedding space. We systemically evaluated our method, named MetaBreak, on both lab environment and commercial LLM platforms. Our approach achieves jailbreak rates comparable to SOTA prompt-engineering-based solutions when no content moderation is deployed. However, when there is content moderation, MetaBreak outperforms SOTA solutions PAP and GPTFuzzer by 11.6% and 34.8%, respectively. Finally, since MetaBreak employs a fundamentally different strategy from prompt engineering, the two approaches can work synergistically. Notably, empowering MetaBreak on PAP and GPTFuzzer boosts jailbreak rates by 24.3% and 20.2%, respectively.

Qitao Tan, Xiaoying Song, Jin Lu et al.

Quantization is an effective technique to reduce the deployment cost of large language models (LLMs), and post-training quantization (PTQ) has been widely studied due to its efficiency. However, existing PTQ methods are limited by their inability to fine-tune model parameters and often suffer significant accuracy loss in low-bit scenarios. Quantization-aware training (QAT) provides a more principled solution, but its reliance on backpropagation incurs prohibitive memory costs, limiting its practicality for LLM deployment. To address these challenges, we propose ZeroQAT, a zeroth-order optimization-based QAT framework that supports both weight and activation quantization. ZeroQAT leverages forward-only gradient estimation to eliminate backpropagation, substantially reducing computational and memory overhead while retaining the benefits of end-to-end optimization. We further introduce a lightweight variant of ZeroQAT for quantized fine-tuning, which freezes and pre-quantizes most parameters to further cut memory usage. Experiments show that ZeroQAT consistently outperforms representative PTQ and QAT baselines while requiring significantly less memory. For example, ZeroQAT enables fine-tuning of a 13B model at extremely low bit-widths (e.g., 2-4 bits) on a single 8GB GPU, and even allows fine-tuning a 6.7B model on a OnePlus 12 smartphone, demonstrating its practicality for end-to-end QAT on resource-limited edge devices.

Zhenglun Kong, Peiyan Dong, Xiaolong Ma et al.

Recently, Vision Transformer (ViT) has continuously established new milestones in the computer vision field, while the high computation and memory cost makes its propagation in industrial production difficult. Pruning, a traditional model compression paradigm for hardware efficiency, has been widely applied in various DNN structures. Nevertheless, it stays ambiguous on how to perform exclusive pruning on the ViT structure. Considering three key points: the structural characteristics, the internal data pattern of ViTs, and the related edge device deployment, we leverage the input token sparsity and propose a computation-aware soft pruning framework, which can be set up on vanilla Transformers of both flatten and CNN-type structures, such as Pooling-based ViT (PiT). More concretely, we design a dynamic attention-based multi-head token selector, which is a lightweight module for adaptive instance-wise token selection. We further introduce a soft pruning technique, which integrates the less informative tokens generated by the selector module into a package token that will participate in subsequent calculations rather than being completely discarded. Our framework is bound to the trade-off between accuracy and computation constraints of specific edge devices through our proposed computation-aware training strategy. Experimental results show that our framework significantly reduces the computation cost of ViTs while maintaining comparable performance on image classification. Moreover, our framework can guarantee the identified model to meet resource specifications of mobile devices and FPGA, and even achieve the real-time execution of DeiT-T on mobile platforms. For example, our method reduces the latency of DeiT-T to 26 ms (26%$\sim $41% superior to existing works) on the mobile device with 0.25%$\sim $4% higher top-1 accuracy on ImageNet.

Yifan Gong, Geng Yuan, Zheng Zhan et al.

Weight pruning is an effective model compression technique to tackle the challenges of achieving real-time deep neural network (DNN) inference on mobile devices. However, prior pruning schemes have limited application scenarios due to accuracy degradation, difficulty in leveraging hardware acceleration, and/or restriction on certain types of DNN layers. In this paper, we propose a general, fine-grained structured pruning scheme and corresponding compiler optimizations that are applicable to any type of DNN layer while achieving high accuracy and hardware inference performance. With the flexibility of applying different pruning schemes to different layers enabled by our compiler optimizations, we further probe into the new problem of determining the best-suited pruning scheme considering the different acceleration and accuracy performance of various pruning schemes. Two pruning scheme mapping methods, one is search-based and the other is rule-based, are proposed to automatically derive the best-suited pruning regularity and block size for each layer of any given DNN. Experimental results demonstrate that our pruning scheme mapping methods, together with the general fine-grained structured pruning scheme, outperform the state-of-the-art DNN optimization framework with up to 2.48$\times$ and 1.73$\times$ DNN inference acceleration on CIFAR-10 and ImageNet dataset without accuracy loss.

Hsin-Hsuan Sung, Yuanchao Xu, Jiexiong Guan et al.

Autonomous driving is of great interest in both research and industry. The high cost has been one of the major roadblocks that slow down the development and adoption of autonomous driving in practice. This paper, for the first-time, shows that it is possible to run level-4 (i.e., fully autonomous driving) software on a single off-the-shelf card (Jetson AGX Xavier) for less than $1k, an order of magnitude less than the state-of-the-art systems, while meeting all the requirements of latency. The success comes from the resolution of some important issues shared by existing practices through a series of measures and innovations. The study overturns the common perceptions of the computing resources required by level-4 autonomous driving, points out a promising path for the industry to lower the cost, and suggests a number of research opportunities for rethinking the architecture, software design, and optimizations of autonomous driving.

Wei Niu, Jiexiong Guan, Yanzhi Wang et al.

Deep Neural Networks (DNNs) have emerged as the core enabler of many major applications on mobile devices. To achieve high accuracy, DNN models have become increasingly deep with hundreds or even thousands of operator layers, leading to high memory and computational requirements for inference. Operator fusion (or kernel/layer fusion) is key optimization in many state-of-the-art DNN execution frameworks, such as TensorFlow, TVM, and MNN. However, these frameworks usually adopt fusion approaches based on certain patterns that are too restrictive to cover the diversity of operators and layer connections. Polyhedral-based loop fusion techniques, on the other hand, work on a low-level view of the computation without operator-level information, and can also miss potential fusion opportunities. To address this challenge, this paper proposes a novel and extensive loop fusion framework called DNNFusion. The basic idea of this work is to work at an operator view of DNNs, but expand fusion opportunities by developing a classification of both individual operators and their combinations. In addition, DNNFusion includes 1) a novel mathematical-property-based graph rewriting framework to reduce evaluation costs and facilitate subsequent operator fusion, 2) an integrated fusion plan generation that leverages the high-level analysis and accurate light-weight profiling, and 3) additional optimizations during fusion code generation. DNNFusion is extensively evaluated on 15 DNN models with varied types of tasks, model sizes, and layer counts. The evaluation results demonstrate that DNNFusion finds up to 8.8x higher fusion opportunities, outperforms four state-of-the-art DNN execution frameworks with 9.3x speedup. The memory requirement reduction and speedups can enable the execution of many of the target models on mobile devices and even make them part of a real-time application.

Wei Niu, Zhengang Li, Xiaolong Ma et al.

It is appealing but challenging to achieve real-time deep neural network (DNN) inference on mobile devices because even the powerful modern mobile devices are considered as ``resource-constrained'' when executing large-scale DNNs. It necessitates the sparse model inference via weight pruning, i.e., DNN weight sparsity, and it is desirable to design a new DNN weight sparsity scheme that can facilitate real-time inference on mobile devices while preserving a high sparse model accuracy. This paper designs a novel mobile inference acceleration framework GRIM that is General to both convolutional neural networks (CNNs) and recurrent neural networks (RNNs) and that achieves Real-time execution and high accuracy, leveraging fine-grained structured sparse model Inference and compiler optimizations for Mobiles. We start by proposing a new fine-grained structured sparsity scheme through the Block-based Column-Row (BCR) pruning. Based on this new fine-grained structured sparsity, our GRIM framework consists of two parts: (a) the compiler optimization and code generation for real-time mobile inference; and (b) the BCR pruning optimizations for determining pruning hyperparameters and performing weight pruning. We compare GRIM with Alibaba MNN, TVM, TensorFlow-Lite, a sparse implementation based on CSR, PatDNN, and ESE (a representative FPGA inference acceleration framework for RNNs), and achieve up to 14.08x speedup.

Zheng Zhan, Yifan Gong, Pu Zhao et al.

Though recent years have witnessed remarkable progress in single image super-resolution (SISR) tasks with the prosperous development of deep neural networks (DNNs), the deep learning methods are confronted with the computation and memory consumption issues in practice, especially for resource-limited platforms such as mobile devices. To overcome the challenge and facilitate the real-time deployment of SISR tasks on mobile, we combine neural architecture search with pruning search and propose an automatic search framework that derives sparse super-resolution (SR) models with high image quality while satisfying the real-time inference requirement. To decrease the search cost, we leverage the weight sharing strategy by introducing a supernet and decouple the search problem into three stages, including supernet construction, compiler-aware architecture and pruning search, and compiler-aware pruning ratio search. With the proposed framework, we are the first to achieve real-time SR inference (with only tens of milliseconds per frame) for implementing 720p resolution with competitive image quality (in terms of PSNR and SSIM) on mobile platforms (Samsung Galaxy S20).

Pu Zhao, Wei Niu, Geng Yuan et al.

Object detection plays an important role in self-driving cars for security development. However, mobile systems on self-driving cars with limited computation resources lead to difficulties for object detection. To facilitate this, we propose a compiler-aware neural pruning search framework to achieve high-speed inference on autonomous vehicles for 2D and 3D object detection. The framework automatically searches the pruning scheme and rate for each layer to find a best-suited pruning for optimizing detection accuracy and speed performance under compiler optimization. Our experiments demonstrate that for the first time, the proposed method achieves (close-to) real-time, 55ms and 99ms inference times for YOLOv4 based 2D object detection and PointPillars based 3D detection, respectively, on an off-the-shelf mobile phone with minor (or no) accuracy loss.

Xuan Shen, Geng Yuan, Wei Niu et al.

The rapid development of autonomous driving, abnormal behavior detection, and behavior recognition makes an increasing demand for multi-person pose estimation-based applications, especially on mobile platforms. However, to achieve high accuracy, state-of-the-art methods tend to have a large model size and complex post-processing algorithm, which costs intense computation and long end-to-end latency. To solve this problem, we propose an architecture optimization and weight pruning framework to accelerate inference of multi-person pose estimation on mobile devices. With our optimization framework, we achieve up to 2.51x faster model inference speed with higher accuracy compared to representative lightweight multi-person pose estimator.

Wei Niu, Zhenglun Kong, Geng Yuan et al.

Transformer-based deep learning models have increasingly demonstrated high accuracy on many natural language processing (NLP) tasks. In this paper, we propose a compression-compilation co-design framework that can guarantee the identified model to meet both resource and real-time specifications of mobile devices. Our framework applies a compiler-aware neural architecture optimization method (CANAO), which can generate the optimal compressed model that balances both accuracy and latency. We are able to achieve up to 7.8x speedup compared with TensorFlow-Lite with only minor accuracy loss. We present two types of BERT applications on mobile devices: Question Answering (QA) and Text Generation. Both can be executed in real-time with latency as low as 45ms. Videos for demonstrating the framework can be found on https://www.youtube.com/watch?v=_WIRvK_2PZI

Pu Zhao, Wei Niu, Geng Yuan et al.

3D object detection is an important task, especially in the autonomous driving application domain. However, it is challenging to support the real-time performance with the limited computation and memory resources on edge-computing devices in self-driving cars. To achieve this, we propose a compiler-aware unified framework incorporating network enhancement and pruning search with the reinforcement learning techniques, to enable real-time inference of 3D object detection on the resource-limited edge-computing devices. Specifically, a generator Recurrent Neural Network (RNN) is employed to provide the unified scheme for both network enhancement and pruning search automatically, without human expertise and assistance. And the evaluated performance of the unified schemes can be fed back to train the generator RNN. The experimental results demonstrate that the proposed framework firstly achieves real-time 3D object detection on mobile devices (Samsung Galaxy S20 phone) with competitive detection performance.

Zhengang Li, Geng Yuan, Wei Niu et al.

With the increasing demand to efficiently deploy DNNs on mobile edge devices, it becomes much more important to reduce unnecessary computation and increase the execution speed. Prior methods towards this goal, including model compression and network architecture search (NAS), are largely performed independently and do not fully consider compiler-level optimizations which is a must-do for mobile acceleration. In this work, we first propose (i) a general category of fine-grained structured pruning applicable to various DNN layers, and (ii) a comprehensive, compiler automatic code generation framework supporting different DNNs and different pruning schemes, which bridge the gap of model compression and NAS. We further propose NPAS, a compiler-aware unified network pruning, and architecture search. To deal with large search space, we propose a meta-modeling procedure based on reinforcement learning with fast evaluation and Bayesian optimization, ensuring the total number of training epochs comparable with representative NAS frameworks. Our framework achieves 6.7ms, 5.9ms, 3.9ms ImageNet inference times with 78.2%, 75% (MobileNet-V3 level), and 71% (MobileNet-V2 level) Top-1 accuracy respectively on an off-the-shelf mobile phone, consistently outperforming prior work.

Chengming Zhang, Geng Yuan, Wei Niu et al.

Convolutional neural networks (CNNs) are becoming increasingly deeper, wider, and non-linear because of the growing demand on prediction accuracy and analysis quality. The wide and deep CNNs, however, require a large amount of computing resources and processing time. Many previous works have studied model pruning to improve inference performance, but little work has been done for effectively reducing training cost. In this paper, we propose ClickTrain: an efficient and accurate end-to-end training and pruning framework for CNNs. Different from the existing pruning-during-training work, ClickTrain provides higher model accuracy and compression ratio via fine-grained architecture-preserving pruning. By leveraging pattern-based pruning with our proposed novel accurate weight importance estimation, dynamic pattern generation and selection, and compiler-assisted computation optimizations, ClickTrain generates highly accurate and fast pruned CNN models for direct deployment without any extra time overhead, compared with the baseline training. ClickTrain also reduces the end-to-end time cost of the pruning-after-training method by up to 2.3X with comparable accuracy and compression ratio. Moreover, compared with the state-of-the-art pruning-during-training approach, ClickTrain provides significant improvements both accuracy and compression ratio on the tested CNN models and datasets, under similar limited training time.

Wei Niu, Zhenglun Kong, Geng Yuan et al.

Pre-trained large-scale language models have increasingly demonstrated high accuracy on many natural language processing (NLP) tasks. However, the limited weight storage and computational speed on hardware platforms have impeded the popularity of pre-trained models, especially in the era of edge computing. In this paper, we seek to find the best model structure of BERT for a given computation size to match specific devices. We propose the first compiler-aware neural architecture optimization framework. Our framework can guarantee the identified model to meet both resource and real-time specifications of mobile devices, thus achieving real-time execution of large transformer-based models like BERT variants. We evaluate our model on several NLP tasks, achieving competitive results on well-known benchmarks with lower latency on mobile devices. Specifically, our model is 5.2x faster on CPU and 4.1x faster on GPU with 0.5-2% accuracy loss compared with BERT-base. Our overall framework achieves up to 7.8x speedup compared with TensorFlow-Lite with only minor accuracy loss.