Zheng Zhan

Xuan Shen, Pu Zhao, Yifan Gong et al. · harvard

Large Language Models (LLMs) have long held sway in the realms of artificial intelligence research. Numerous efficient techniques, including weight pruning, quantization, and distillation, have been embraced to compress LLMs, targeting memory reduction and inference acceleration, which underscore the redundancy in LLMs. However, most model compression techniques concentrate on weight optimization, overlooking the exploration of optimal architectures. Besides, traditional architecture search methods, limited by the elevated complexity with extensive parameters, struggle to demonstrate their effectiveness on LLMs. In this paper, we propose a training-free architecture search framework to identify optimal subnets that preserve the fundamental strengths of the original LLMs while achieving inference acceleration. Furthermore, after generating subnets that inherit specific weights from the original LLMs, we introduce a reformation algorithm that utilizes the omitted weights to rectify the inherited weights with a small amount of calibration data. Compared with SOTA training-free structured pruning works that can generate smaller networks, our method demonstrates superior performance across standard benchmarks. Furthermore, our generated subnets can directly reduce the usage of GPU memory and achieve inference acceleration. Code: https://github.com/shawnricecake/search-llm

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.

Zifeng Wang, Zheng Zhan, Yifan Gong et al.

Rehearsal-based approaches are a mainstay of continual learning (CL). They mitigate the catastrophic forgetting problem by maintaining a small fixed-size buffer with a subset of data from past tasks. While most rehearsal-based approaches study how to effectively exploit the knowledge from the buffered past data, little attention is paid to the inter-task relationships with the critical task-specific and task-invariant knowledge. By appropriately leveraging inter-task relationships, we propose a novel CL method named DualHSIC to boost the performance of existing rehearsal-based methods in a simple yet effective way. DualHSIC consists of two complementary components that stem from the so-called Hilbert Schmidt independence criterion (HSIC): HSIC-Bottleneck for Rehearsal (HBR) lessens the inter-task interference and HSIC Alignment (HA) promotes task-invariant knowledge sharing. Extensive experiments show that DualHSIC can be seamlessly plugged into existing rehearsal-based methods for consistent performance improvements, and also outperforms recent state-of-the-art regularization-enhanced rehearsal methods. Source code will be released.

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).

Yifan Gong, Zheng Zhan, Pu Zhao et al.

During the deployment of deep neural networks (DNNs) on edge devices, many research efforts are devoted to the limited hardware resource. However, little attention is paid to the influence of dynamic power management. As edge devices typically only have a budget of energy with batteries (rather than almost unlimited energy support on servers or workstations), their dynamic power management often changes the execution frequency as in the widely-used dynamic voltage and frequency scaling (DVFS) technique. This leads to highly unstable inference speed performance, especially for computation-intensive DNN models, which can harm user experience and waste hardware resources. We firstly identify this problem and then propose All-in-One, a highly representative pruning framework to work with dynamic power management using DVFS. The framework can use only one set of model weights and soft masks (together with other auxiliary parameters of negligible storage) to represent multiple models of various pruning ratios. By re-configuring the model to the corresponding pruning ratio for a specific execution frequency (and voltage), we are able to achieve stable inference speed, i.e., keeping the difference in speed performance under various execution frequencies as small as possible. Our experiments demonstrate that our method not only achieves high accuracy for multiple models of different pruning ratios, but also reduces their variance of inference latency for various frequencies, with minimal memory consumption of only one model and one soft mask.

Yifan Gong, Zheng Zhan, Yanyu Li et al.

Good weight initialization serves as an effective measure to reduce the training cost of a deep neural network (DNN) model. The choice of how to initialize parameters is challenging and may require manual tuning, which can be time-consuming and prone to human error. To overcome such limitations, this work takes a novel step towards building a weight generator to synthesize the neural weights for initialization. We use the image-to-image translation task with generative adversarial networks (GANs) as an example due to the ease of collecting model weights spanning a wide range. Specifically, we first collect a dataset with various image editing concepts and their corresponding trained weights, which are later used for the training of the weight generator. To address the different characteristics among layers and the substantial number of weights to be predicted, we divide the weights into equal-sized blocks and assign each block an index. Subsequently, a diffusion model is trained with such a dataset using both text conditions of the concept and the block indexes. By initializing the image translation model with the denoised weights predicted by our diffusion model, the training requires only 43.3 seconds. Compared to training from scratch (i.e., Pix2pix), we achieve a 15x training time acceleration for a new concept while obtaining even better image generation quality.

Kwangjun Ahn, Byron Xu, Natalie Abreu et al.

Orthonormalized updates accelerate training, improve stability, and enable robust hyperparameter transfer, but existing methods like Muon rely on dense matrix operations that clash with sharded weights in large-scale LLM training, causing high compute and communication cost. We introduce Dion (Distributed Orthonormalization), a scalable and efficient update rule that replaces Newton-Schulz iteration with amortized power iteration on a momentum buffer, avoiding full-matrix reconstruction and integrating cleanly with weight sharding. The rank-fraction parameter with error feedback enables low-rank updates that balance quality with significant cost savings. On language models from 160M to 3B parameters, Dion retains the benefits of orthonormalized updates, while markedly reducing wall-clock time at scale, making it a practical optimizer for next-generation foundation models. Code is available at: https://github.com/microsoft/dion/

Qitao Tan, Jun Liu, Zheng Zhan et al.

Large language models (LLMs) excel across various tasks, but standard first-order (FO) fine-tuning demands considerable memory, significantly limiting real-world deployment. Recently, zeroth-order (ZO) optimization stood out as a promising memory-efficient training paradigm, avoiding backward passes and relying solely on forward passes for gradient estimation, making it attractive for resource-constrained scenarios. However, ZO method lags far behind FO method in both convergence speed and accuracy. To bridge the gap, we introduce a novel layer-wise divergence analysis that uncovers the distinct update pattern of FO and ZO optimization. Aiming to resemble the learning capacity of FO method from the findings, we propose Divergence-driven Zeroth-Order (DiZO) optimization. DiZO conducts divergence-driven layer adaptation by incorporating projections to ZO updates, generating diverse-magnitude updates precisely scaled to layer-wise individual optimization needs. Our results demonstrate that DiZO significantly reduces the needed iterations for convergence without sacrificing throughput, cutting training GPU hours by up to 48\% on various datasets. Moreover, DiZO consistently outperforms the representative ZO baselines in fine-tuning RoBERTa-large, OPT-series, and Llama-series on downstream tasks and, in some cases, even surpasses memory-intensive FO fine-tuning. Our code is released at https://github.com/Skilteee/DiZO.

Liliang Ren, Congcong Chen, Haoran Xu et al.

Recent advances in language modeling have demonstrated the effectiveness of State Space Models (SSMs) for efficient sequence modeling. While hybrid architectures such as Samba and the decoder-decoder architecture, YOCO, have shown promising performance gains over Transformers, prior works have not investigated the efficiency potential of representation sharing between SSM layers. In this paper, we introduce the Gated Memory Unit (GMU), a simple yet effective mechanism for efficient memory sharing across layers. We apply it to create SambaY, a decoder-hybrid-decoder architecture that incorporates GMUs in the cross-decoder to share memory readout states from a Samba-based self-decoder. SambaY significantly enhances decoding efficiency, preserves linear pre-filling time complexity, and boosts long-context performance, all while eliminating the need for explicit positional encoding. Through extensive scaling experiments, we demonstrate that our model exhibits a significantly lower irreducible loss compared to a strong YOCO baseline, indicating superior performance scalability under large-scale compute regimes. Our largest model enhanced with Differential Attention, Phi4-mini-Flash-Reasoning, achieves significantly better performance than Phi4-mini-Reasoning on reasoning tasks such as Math500, AIME24/25, and GPQA Diamond without any reinforcement learning, while delivering up to 10x higher decoding throughput on 2K-length prompts with 32K generation length under the vLLM inference framework. We release our training codebase on open-source data at https://github.com/microsoft/ArchScale.

Zhenglun Kong, Zheng Zhan, Shiyue Hou et al. · harvard

Large language models (LLMs) have shown remarkable promise but remain challenging to continually improve through traditional finetuning, particularly when integrating capabilities from other specialized LLMs. Popular methods like ensemble and weight merging require substantial memory and struggle to adapt to changing data environments. Recent efforts have transferred knowledge from multiple LLMs into a single target model; however, they suffer from interference and degraded performance among tasks, largely due to limited flexibility in candidate selection and training pipelines. To address these issues, we propose a framework that adaptively selects and aggregates knowledge from diverse LLMs to build a single, stronger model, avoiding the high memory overhead of ensemble and inflexible weight merging. Specifically, we design an adaptive selection network that identifies the most relevant source LLMs based on their scores, thereby reducing knowledge interference. We further propose a dynamic weighted fusion strategy that accounts for the inherent strengths of candidate LLMs, along with a feedback-driven loss function that prevents the selector from converging on a single subset of sources. Experimental results demonstrate that our method can enable a more stable and scalable knowledge aggregation process while reducing knowledge interference by up to 50% compared to existing approaches. Code is avaliable at https://github.com/ZLKong/LLM_Integration

Yang Xiao, Gen Li, Kaiyuan Deng et al.

Training-free acceleration has emerged as an advanced research area in video generation based on diffusion models. The redundancy of latents in diffusion model inference provides a natural entry point for acceleration. In this paper, we decompose the inference process into the encoding, denoising, and decoding stages, and observe that cache-based acceleration methods often lead to substantial memory surges in the latter two stages. To address this problem, we analyze the characteristics of inference across different stages and propose stage-specific strategies for reducing memory consumption: 1) Asynchronous Cache Swapping. 2) Feature chunk. 3) Slicing latents to decode. At the same time, we ensure that the time overhead introduced by these three strategies remains lower than the acceleration gains themselves. Compared with the baseline, our approach achieves faster inference speed and lower memory usage, while maintaining quality degradation within an acceptable range. The Code is available at https://github.com/NKUShaw/LightCache .

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.

Zichong Meng, Jie Zhang, Changdi Yang et al.

Class Incremental Learning (CIL) is challenging due to catastrophic forgetting. On top of that, Exemplar-free Class Incremental Learning is even more challenging due to forbidden access to previous task data. Recent exemplar-free CIL methods attempt to mitigate catastrophic forgetting by synthesizing previous task data. However, they fail to overcome the catastrophic forgetting due to the inability to deal with the significant domain gap between real and synthetic data. To overcome these issues, we propose a novel exemplar-free CIL method. Our method adopts multi-distribution matching (MDM) diffusion models to unify quality and bridge domain gaps among all domains of training data. Moreover, our approach integrates selective synthetic image augmentation (SSIA) to expand the distribution of the training data, thereby improving the model's plasticity and reinforcing the performance of our method's ultimate component, multi-domain adaptation (MDA). With the proposed integrations, our method then reformulates exemplar-free CIL into a multi-domain adaptation problem to implicitly address the domain gap problem to enhance model stability during incremental training. Extensive experiments on benchmark class incremental datasets and settings demonstrate that our method excels previous exemplar-free CIL methods and achieves state-of-the-art performance.

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).

Zheng Zhan, Yushu Wu, Zhenglun Kong et al. · harvard

Recent advancements in State Space Models (SSMs) have attracted significant interest, particularly in models optimized for parallel training and handling long-range dependencies. Architectures like Mamba have scaled to billions of parameters with selective SSM. To facilitate broader applications using Mamba, exploring its efficiency is crucial. While token reduction techniques offer a straightforward post-training strategy, we find that applying existing methods directly to SSMs leads to substantial performance drops. Through insightful analysis, we identify the reasons for this failure and the limitations of current techniques. In response, we propose a tailored, unified post-training token reduction method for SSMs. Our approach integrates token importance and similarity, thus taking advantage of both pruning and merging, to devise a fine-grained intra-layer token reduction strategy. Extensive experiments show that our method improves the average accuracy by 5.7% to 13.1% on six benchmarks with Mamba-2 compared to existing methods, while significantly reducing computational demands and memory requirements.

Qitao Tan, Sung-En Chang, Rui Xia et al.

Zeroth-order (ZO) optimization is an emerging deep neural network (DNN) training paradigm that offers computational simplicity and memory savings. However, this seemingly promising approach faces a significant and long-ignored challenge. ZO requires generating a substantial number of Gaussian random numbers, which poses significant difficulties and even makes it infeasible for hardware platforms, such as FPGAs and ASICs. In this paper, we identify this critical issue, which arises from the mismatch between algorithm and hardware designers. To address this issue, we proposed PeZO, a perturbation-efficient ZO framework. Specifically, we design random number reuse strategies to significantly reduce the demand for random number generation and introduce a hardware-friendly adaptive scaling method to replace the costly Gaussian distribution with a uniform distribution. Our experiments show that PeZO reduces the required LUTs and FFs for random number generation by 48.6\% and 12.7\%, and saves at maximum 86\% power consumption, all without compromising training performance, making ZO optimization feasible for on-device training. To the best of our knowledge, we are the first to explore the potential of on-device ZO optimization, providing valuable insights for future research.

Xuehai He, Shuohang Wang, Jianwei Yang et al.

Recent advancements in diffusion models have shown great promise in producing high-quality video content. However, efficiently training video diffusion models capable of integrating directional guidance and controllable motion intensity remains a challenging and under-explored area. To tackle these challenges, this paper introduces Mojito, a diffusion model that incorporates both motion trajectory and intensity control for text-to-video generation. Specifically, Mojito features a Directional Motion Control (DMC) module that leverages cross-attention to efficiently direct the generated object's motion without training, alongside a Motion Intensity Modulator (MIM) that uses optical flow maps generated from videos to guide varying levels of motion intensity. Extensive experiments demonstrate Mojito's effectiveness in achieving precise trajectory and intensity control with high computational efficiency, generating motion patterns that closely match specified directions and intensities, providing realistic dynamics that align well with natural motion in real-world scenarios.

Zheng Zhan, Liliang Ren, Shuohang Wang et al.

Linear State Space Models (SSMs) offer remarkable performance gains in efficient sequence modeling, with constant inference-time computation and memory complexity. Recent advances, such as Mamba, further enhance SSMs with input-dependent gating and hardware-aware implementations, positioning them as strong alternatives to Transformers for long sequence modeling. However, efficiently scaling the expressive power of SSMs, particularly with Mixture of Experts (MoE), remains challenging, as naive integration attempts often falter or degrade performance. In this work, we introduce Routing Mamba (RoM), a novel approach that scales SSM parameters using sparse mixtures of linear projection experts. By sharing routing decisions between projection layers and lightweight sub-modules within Mamba across experts, RoM leverages synergies among linear projection experts for effective and efficient sparse scaling of Mamba layers. At a scale of 1.3B active parameters (10B total) and 16K training sequence length, RoM achieves language modeling performance equivalent to a dense Mamba model requiring over 2.3x more active parameters, and demonstrates consistent perplexity across context lengths. Experimental results further show RoM effectively scales hybrid language models, yielding a 23% FLOPS saving compared to dense Mamba scaling for similar performance.

Yifan Gong, Zheng Zhan, Qing Jin et al.

One highly promising direction for enabling flexible real-time on-device image editing is utilizing data distillation by leveraging large-scale text-to-image diffusion models to generate paired datasets used for training generative adversarial networks (GANs). This approach notably alleviates the stringent requirements typically imposed by high-end commercial GPUs for performing image editing with diffusion models. However, unlike text-to-image diffusion models, each distilled GAN is specialized for a specific image editing task, necessitating costly training efforts to obtain models for various concepts. In this work, we introduce and address a novel research direction: can the process of distilling GANs from diffusion models be made significantly more efficient? To achieve this goal, we propose a series of innovative techniques. First, we construct a base GAN model with generalized features, adaptable to different concepts through fine-tuning, eliminating the need for training from scratch. Second, we identify crucial layers within the base GAN model and employ Low-Rank Adaptation (LoRA) with a simple yet effective rank search process, rather than fine-tuning the entire base model. Third, we investigate the minimal amount of data necessary for fine-tuning, further reducing the overall training time. Extensive experiments show that we can efficiently empower GANs with the ability to perform real-time high-quality image editing on mobile devices with remarkably reduced training and storage costs for each concept.

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.

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).

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.

Wei Niu, Pu Zhao, Zheng Zhan et al.

High-end mobile platforms rapidly serve as primary computing devices for a wide range of Deep Neural Network (DNN) applications. However, the constrained computation and storage resources on these devices still pose significant challenges for real-time DNN inference executions. To address this problem, we propose a set of hardware-friendly structured model pruning and compiler optimization techniques to accelerate DNN executions on mobile devices. This demo shows that these optimizations can enable real-time mobile execution of multiple DNN applications, including style transfer, DNN coloring and super resolution.

Tianyun Zhang, Xiaolong Ma, Zheng Zhan et al.

To address the large model size and intensive computation requirement of deep neural networks (DNNs), weight pruning techniques have been proposed and generally fall into two categories, i.e., static regularization-based pruning and dynamic regularization-based pruning. However, the former method currently suffers either complex workloads or accuracy degradation, while the latter one takes a long time to tune the parameters to achieve the desired pruning rate without accuracy loss. In this paper, we propose a unified DNN weight pruning framework with dynamically updated regularization terms bounded by the designated constraint, which can generate both non-structured sparsity and different kinds of structured sparsity. We also extend our method to an integrated framework for the combination of different DNN compression tasks.

Yifan Gong, Zheng Zhan, Zhengang Li et al.

Weight pruning of deep neural networks (DNNs) has been proposed to satisfy the limited storage and computing capability of mobile edge devices. However, previous pruning methods mainly focus on reducing the model size and/or improving performance without considering the privacy of user data. To mitigate this concern, we propose a privacy-preserving-oriented pruning and mobile acceleration framework that does not require the private training dataset. At the algorithm level of the proposed framework, a systematic weight pruning technique based on the alternating direction method of multipliers (ADMM) is designed to iteratively solve the pattern-based pruning problem for each layer with randomly generated synthetic data. In addition, corresponding optimizations at the compiler level are leveraged for inference accelerations on devices. With the proposed framework, users could avoid the time-consuming pruning process for non-experts and directly benefit from compressed models. Experimental results show that the proposed framework outperforms three state-of-art end-to-end DNN frameworks, i.e., TensorFlow-Lite, TVM, and MNN, with speedup up to 4.2X, 2.5X, and 2.0X, respectively, with almost no accuracy loss, while preserving data privacy.

Zhengang Li, Yifan Gong, Xiaolong Ma et al.

Structured weight pruning is a representative model compression technique of DNNs for hardware efficiency and inference accelerations. Previous works in this area leave great space for improvement since sparse structures with combinations of different structured pruning schemes are not exploited fully and efficiently. To mitigate the limitations, we propose SS-Auto, a single-shot, automatic structured pruning framework that can achieve row pruning and column pruning simultaneously. We adopt soft constraint-based formulation to alleviate the strong non-convexity of l0-norm constraints used in state-of-the-art ADMM-based methods for faster convergence and fewer hyperparameters. Instead of solving the problem directly, a Primal-Proximal solution is proposed to avoid the pitfall of penalizing all weights equally, thereby enhancing the accuracy. Extensive experiments on CIFAR-10 and CIFAR-100 datasets demonstrate that the proposed framework can achieve ultra-high pruning rates while maintaining accuracy. Furthermore, significant inference speedup has been observed from the proposed framework through actual measurements on the smartphone.

Xiaolong Ma, Zhengang Li, Yifan Gong et al.

Accelerating DNN execution on various resource-limited computing platforms has been a long-standing problem. Prior works utilize l1-based group lasso or dynamic regularization such as ADMM to perform structured pruning on DNN models to leverage the parallel computing architectures. However, both of the pruning dimensions and pruning methods lack universality, which leads to degraded performance and limited applicability. To solve the problem, we propose a new block-based pruning framework that comprises a general and flexible structured pruning dimension as well as a powerful and efficient reweighted regularization method. Our framework is universal, which can be applied to both CNNs and RNNs, implying complete support for the two major kinds of computation-intensive layers (i.e., CONV and FC layers). To complete all aspects of the pruning-for-acceleration task, we also integrate compiler-based code optimization into our framework that can perform DNN inference in a real-time manner. To the best of our knowledge, it is the first time that the weight pruning framework achieves universal coverage for both CNNs and RNNs with real-time mobile acceleration and no accuracy compromise.

Yanzhi Wang, Zheng Zhan, Jiayu Li et al.

Large-scale deep neural networks are both memory intensive and computation-intensive, thereby posing stringent requirements on the computing platforms. Hardware accelerations of deep neural networks have been extensively investigated in both industry and academia. Specific forms of binary neural networks (BNNs) and stochastic computing based neural networks (SCNNs) are particularly appealing to hardware implementations since they can be implemented almost entirely with binary operations. Despite the obvious advantages in hardware implementation, these approximate computing techniques are questioned by researchers in terms of accuracy and universal applicability. Also it is important to understand the relative pros and cons of SCNNs and BNNs in theory and in actual hardware implementations. In order to address these concerns, in this paper we prove that the "ideal" SCNNs and BNNs satisfy the universal approximation property with probability 1 (due to the stochastic behavior). The proof is conducted by first proving the property for SCNNs from the strong law of large numbers, and then using SCNNs as a "bridge" to prove for BNNs. Based on the universal approximation property, we further prove that SCNNs and BNNs exhibit the same energy complexity. In other words, they have the same asymptotic energy consumption with the growing of network size. We also provide a detailed analysis of the pros and cons of SCNNs and BNNs for hardware implementations and conclude that SCNNs are more suitable for hardware.