Restricted Boltzmann machines (RBMs) and their extensions, called 'deep-belief networks', are powerful neural networks that have found applications in the fields of machine learning and artificial intelligence. The standard way to training these models resorts to an iterative unsupervised procedure based on Gibbs sampling, called 'contrastive divergence' (CD), and additional supervised tuning via back-propagation. However, this procedure has been shown not to follow any gradient and can lead to suboptimal solutions. In this paper, we show an efficient alternative to CD by means of simulations of digital memcomputing machines (DMMs). We test our approach on pattern recognition using a modified version of the MNIST data set. DMMs sample effectively the vast phase space given by the model distribution of the RBM, and provide a very good approximation close to the optimum. This efficient search significantly reduces the number of pretraining iterations necessary to achieve a given level of accuracy, as well as a total performance gain over CD. In fact, the acceleration of pretraining achieved by simulating DMMs is comparable to, in number of iterations, the recently reported hardware application of the quantum annealing method on the same network and data set. Notably, however, DMMs perform far better than the reported quantum annealing results in terms of quality of the training. We also compare our method to advances in supervised training, like batch-normalization and rectifiers, that work to reduce the advantage of pretraining. We find that the memcomputing method still maintains a quality advantage ($>1\%$ in accuracy, and a $20\%$ reduction in error rate) over these approaches. Furthermore, our method is agnostic about the connectivity of the network. Therefore, it can be extended to train full Boltzmann machines, and even deep networks at once.
Deep Learning has been one of the most disruptive technological advancements in recent times. The high performance of deep learning models comes at the expense of high computational, storage and power requirements. Sensing the immediate need for accelerating and compressing these models to improve on-device performance, we introduce Deeplite Neutrino for production-ready optimization of the models and Deeplite Runtime for deployment of ultra-low bit quantized models on Arm-based platforms. We implement low-level quantization kernels for Armv7 and Armv8 architectures enabling deployment on the …
Technical University of Munich（慕尼黑工业大学）；Audi AG（奥迪股份公司）
While abstraction is a classic tool of verification to scale it up, it is not used very often for verifying neural networks. However, it can help with the still open task of scaling existing algorithms to state-of-the-art network architectures. We introduce an abstraction framework applicable to fully-connected feed-forward neural networks based on clustering of neurons that behave similarly on some inputs. For the particular case of ReLU, we additionally provide error bounds incurred by the abstraction. We show how the abstraction reduces the size of the network, while preserving its accuracy…
University of California, Berkeley（加州大学伯克利分校）；Google AI（谷歌人工智能）
A significant advance in accelerating neural network training has been the development of normalization methods, permitting the training of deep models both faster and with better accuracy. These advances come with practical challenges: for instance, batch normalization ties the prediction of individual examples with other examples within a batch, resulting in a network that is heavily dependent on batch size. Layer normalization and group normalization are data-dependent and thus must be continually used, even at test-time. To address the issues that arise from using explicit normalization te…
While Deep Learning (DL) technologies are a promising tool to solve networking problems that map to classification tasks, their computational complexity is still too high with respect to real-time traffic measurements requirements. To reduce the DL inference cost, we propose a novel caching paradigm, that we named approximate-key caching, which returns approximate results for lookups of selected input based on cached DL inference results. While approximate cache hits alleviate DL inference workload and increase the system throughput, they however introduce an approximation error. As such, we c…
As neural network model sizes have dramatically increased, so has the interest in various techniques to reduce their parameter counts and accelerate their execution. An active area of research in this field is sparsity - encouraging zero values in parameters that can then be discarded from storage or computations. While most research focuses on high levels of sparsity, there are challenges in universally maintaining model accuracy as well as achieving significant speedups over modern matrix-math hardware. To make sparsity adoption practical, the NVIDIA Ampere GPU architecture introduces sparsi…
The University of Arizona（亚利桑那大学）；Lawrence Berkeley National Laboratory（劳伦斯伯克利国家实验室）
GMRES is a powerful numerical solver used to find solutions to extremely large systems of linear equations. These systems of equations appear in many applications in science and engineering. Here we demonstrate a real-time machine learning algorithm that can be used to accelerate the time-to-solution for GMRES. Our framework is novel in that is integrates the deep learning algorithm in an in situ fashion: the AI-accelerator gradually learns how to optimizes the time to solution without requiring user input (such as a pre-trained data set). We describe how our algorithm collects data and optimi…
Industrial Internet of Things (IIoT) applications can benefit from leveraging edge computing. For example, applications underpinned by deep neural networks (DNN) models can be sliced and distributed across the IIoT device and the edge of the network for improving the overall performance of inference and for enhancing privacy of the input data, such as industrial product images. However, low network performance between IIoT devices and the edge is often a bottleneck. In this study, we develop ScissionLite, a holistic framework for accelerating distributed DNN inference using the Transfer Layer …
Chinese University of Hong Kong（香港中文大学）；Xidian University（西安电子科技大学）；Tsinghua University（清华大学）；IBM T. J. Watson（IBM托马斯·J·沃森研究中心）
Deep neural networks (DNNs) have achieved significant success in a variety of real world applications, i.e., image classification. However, tons of parameters in the networks restrict the efficiency of neural networks due to the large model size and the intensive computation. To address this issue, various approximation techniques have been investigated, which seek for a light weighted network with little performance degradation in exchange of smaller model size or faster inference. Both low-rankness and sparsity are appealing properties for the network approximation. In this paper we propose …
William & Mary（威廉与玛丽学院）；Northeastern University（东北大学）；Augusta University（奥古斯塔大学）
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 conn…
A Survey on Design Methodologies for Accelerating Deep Learning on Heterogeneous Architectures
Politecnico di Milano（米兰理工大学）；Università di Bologna（博洛尼亚大学）；Università degli Studi di Ferrara（费拉拉大学）；INFN（意大利国家核物理研究所）；Università degli Studi di Catania（卡塔尼亚大学）；Università degli Studi di Roma Tor Vergata（罗马-托尔维加塔大学）；Università degli Studi della Calabria（卡拉布里亚大学）
Syracuse University（雪城大学）；City University of New York, City College（纽约市立大学城市学院）；University of Southern California（南加州大学）；California State University Fullerton（加州州立大学富勒顿分校）；Northeastern University（东北大学）
Large-scale deep neural networks (DNNs) are both compute and memory intensive. As the size of DNNs continues to grow, it is critical to improve the energy efficiency and performance while maintaining accuracy. For DNNs, the model size is an important factor affecting performance, scalability and energy efficiency. Weight pruning achieves good compression ratios but suffers from three drawbacks: 1) the irregular network structure after pruning; 2) the increased training complexity; and 3) the lack of rigorous guarantee of compression ratio and inference accuracy. To overcome these limitations, …
Recent advances in Neural Architecture Search (NAS) which extract specialized hardware-aware configurations (a.k.a. "sub-networks") from a hardware-agnostic "super-network" have become increasingly popular. While considerable effort has been employed towards improving the first stage, namely, the training of the super-network, the search for derivative high-performing sub-networks is still largely under-explored. For example, some recent network morphism techniques allow a super-network to be trained once and then have hardware-specific networks extracted from it as needed. These methods decou…
Pacific Northwest National Laboratory（太平洋西北国家实验室）；University of Texas at Arlington（德克萨斯大学阿灵顿分校）
Deep Learning is a very powerful machine learning model. Deep Learning trains a large number of parameters for multiple layers and is very slow when data is in large scale and the architecture size is large. Inspired from the shrinking technique used in accelerating computation of Support Vector Machines (SVM) algorithm and screening technique used in LASSO, we propose a shrinking Deep Learning with recall (sDLr) approach to speed up deep learning computation. We experiment shrinking Deep Learning with recall (sDLr) using Deep Neural Network (DNN), Deep Belief Network (DBN) and Convolution Neu…
In this paper, we propose an adaptive pruning method. This method can cut off the channel and layer adaptively. The proportion of the layer and the channel to be cut is learned adaptively. The pruning method proposed in this paper can reduce half of the parameters, and the accuracy will not decrease or even be higher than baseline.
We explore techniques to significantly improve the compute efficiency and performance of Deep Convolution Networks without impacting their accuracy. To improve the compute efficiency, we focus on achieving high accuracy with extremely low-precision (2-bit) weight networks, and to accelerate the execution time, we aggressively skip operations on zero-values. We achieve the highest reported accuracy of 76.6% Top-1/93% Top-5 on the Imagenet object classification challenge with low-precision network\footnote{github release of the source code coming soon} while reducing the compute requirement by ~…
In this paper, we propose a principled deep reinforcement learning (RL) approach that is able to accelerate the convergence rate of general deep neural networks (DNNs). With our approach, a deep RL agent (synonym for optimizer in this work) is used to automatically learn policies about how to schedule learning rates during the optimization of a DNN. The state features of the agent are learned from the weight statistics of the optimizee during training. The reward function of this agent is designed to learn policies that minimize the optimizee's training time given a certain performance goal. T…
University of Chinese Academy of Sciences（中国科学院大学）；Columbia University（哥伦比亚大学）
Deep unsupervised domain adaptation (UDA) has recently received increasing attention from researchers. However, existing methods are computationally intensive due to the computation cost of Convolutional Neural Networks (CNN) adopted by most work. To date, there is no effective network compression method for accelerating these models. In this paper, we propose a unified Transfer Channel Pruning (TCP) approach for accelerating UDA models. TCP is capable of compressing the deep UDA model by pruning less important channels while simultaneously learning transferable features by reducing the cross-…
We introduce a novel method to compute a rank $m$ approximation of the inverse of the Hessian matrix in the distributed regime. By leveraging the differences in gradients and parameters of multiple Workers, we are able to efficiently implement a distributed approximation of the Newton-Raphson method. We also present preliminary results which underline advantages and challenges of second-order methods for large stochastic optimization problems. In particular, our work suggests that novel strategies for combining gradients provide further information on the loss surface.
Current deep reinforcement learning (DRL) algorithms utilize randomness in simulation environments to assume complete coverage in the state space. However, particularly in high dimensions, relying on randomness may lead to gaps in coverage of the trained DRL neural network model, which in turn may lead to drastic and often fatal real-world situations. To the best of the author's knowledge, the assessment of coverage for DRL is lacking in current research literature. Therefore, in this paper, a novel measure, Approximate Pseudo-Coverage (APC), is proposed for assessing the coverage in DRL appli…
This paper investigates the use of prior computation to estimate the value function to improve sample efficiency in on-policy policy gradient methods in reinforcement learning. Our approach is to estimate the value function from prior computations, such as from the Q-network learned in DQN or the value function trained for different but related environments. In particular, we learn a new value function for the target task while combining it with a value estimate from the prior computation. Finally, the resulting value function is used as a baseline in the policy gradient method. This use of a …
Recently, deep neural networks (DNNs) have shown advantages in accelerating optimization algorithms. One approach is to unfold finite number of iterations of conventional optimization algorithms and to learn parameters in the algorithms. However, these are forward methods and are indeed neither iterative nor convergent. Here, we present a novel DNN-based convergent iterative algorithm that accelerates conventional optimization algorithms. We train a DNN to yield parameters in scaled gradient projection method. So far, these parameters have been chosen heuristically, but have shown to be crucia…
Reinforcement learning augmented by the representational power of deep neural networks, has shown promising results on high-dimensional problems, such as game playing and robotic control. However, the sequential nature of these problems poses a fundamental challenge for computational efficiency. Recently, alternative approaches such as evolutionary strategies and deep neuroevolution demonstrated competitive results with faster training time on distributed CPU cores. Here, we report record training times (running at about 1 million frames per second) for Atari 2600 games using deep neuroevoluti…
We propose a reconfigurable hardware architecture for deep neural networks (DNNs) capable of online training and inference, which uses algorithmically pre-determined, structured sparsity to significantly lower memory and computational requirements. This novel architecture introduces the notion of edge-processing to provide flexibility and combines junction pipelining and operational parallelization to speed up training. The overall effect is to reduce network complexity by factors up to 30x and training time by up to 35x relative to GPUs, while maintaining high fidelity of inference results. T…
This paper aims to accelerate the test-time computation of convolutional neural networks (CNNs), especially very deep CNNs that have substantially impacted the computer vision community. Unlike previous methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units into account. We develop an effective solution to the resulting nonlinear optimization problem without the need of stochastic gradient descent (SGD). More importantly, while previous methods mainly focus on optimizing one or two layers, our nonlinear method enables an asymmetric …
UniDT；Research Institute of Intelligent Complex Systems, Fudan University（复旦大学智能复杂系统研究院）；Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University（复旦大学类脑智能科学与技术研究院）
Deep Q-learning Network (DQN) is a successful way which combines reinforcement learning with deep neural networks and leads to a widespread application of reinforcement learning. One challenging problem when applying DQN or other reinforcement learning algorithms to real world problem is data collection. Therefore, how to improve data efficiency is one of the most important problems in the research of reinforcement learning. In this paper, we propose a framework which uses the Max-Mean loss in Deep Q-Network (M$^2$DQN). Instead of sampling one batch of experiences in the training step, we samp…
Software frameworks for neural networks play a key role in the development and application of deep learning methods. In this paper, we introduce the Chainer framework, which intends to provide a flexible, intuitive, and high performance means of implementing the full range of deep learning models needed by researchers and practitioners. Chainer provides acceleration using Graphics Processing Units with a familiar NumPy-like API through CuPy, supports general and dynamic models in Python through Define-by-Run, and also provides add-on packages for state-of-the-art computer vision models as well…
University of Perugia（佩鲁贾大学）；Free University of Bozen-Bolzano（博尔扎诺自由大学）
Emerging applications such as Deep Learning are often data-driven, thus traditional approaches based on auto-tuners are not performance effective across the wide range of inputs used in practice. In the present paper, we start an investigation of predictive models based on machine learning techniques in order to optimize Convolution Neural Networks (CNNs). As a use-case, we focus on the ARM Compute Library which provides three different implementations of the convolution operator at different numeric precision. Starting from a collation of benchmarks, we build and validate models learned by De…
Decentralized training of deep learning models is a key element for enabling data privacy and on-device learning over networks. In realistic learning scenarios, the presence of heterogeneity across different clients' local datasets poses an optimization challenge and may severely deteriorate the generalization performance. In this paper, we investigate and identify the limitation of several decentralized optimization algorithms for different degrees of data heterogeneity. We propose a novel momentum-based method to mitigate this decentralized training difficulty. We show in extensive empirical…
Point source detection techniques are used to identify and localise point sources in radio astronomical surveys. With the development of the Square Kilometre Array (SKA) telescope, survey images will see a massive increase in size from Gigapixels to Terapixels. Point source detection has already proven to be a challenge in recent surveys performed by SKA pathfinder telescopes. This paper proposes the Point Proposal Network (PPN): a point source detector that utilises deep convolutional neural networks for fast source detection. Results measured on simulated MeerKAT images show that, although l…
Accelerating the inference of a trained DNN is a well studied subject. In this paper we switch the focus to the training of DNNs. The training phase is compute intensive, demands complicated data communication, and contains multiple levels of data dependencies and parallelism. This paper presents an algorithm/architecture space exploration of efficient accelerators to achieve better network convergence rates and higher energy efficiency for training DNNs. We further demonstrate that an architecture with hierarchical support for collective communication semantics provides flexibility in trainin…
In recent years, there has been a flurry of research in deep neural network pruning and compression. Early approaches prune weights individually. However, it is difficult to take advantage of the resulting unstructured sparsity patterns on modern hardware like GPUs. As a result, pruning strategies which impose sparsity structures in the weights have become more popular. However,these structured pruning approaches typically lead to higher losses in accuracy than unstructured pruning. In this paper, we present SparseRT, a code generator that leverage unstructured sparsity to accelerate sparse li…
University of Luxembourg（卢森堡大学）；Institute of Fundamental Technological Research, Polish Academy of Sciences（波兰科学院基础技术研究所）
Deep learning surrogate models are being increasingly used in accelerating scientific simulations as a replacement for costly conventional numerical techniques. However, their use remains a significant challenge when dealing with real-world complex examples. In this work, we demonstrate three types of neural network architectures for efficient learning of highly non-linear deformations of solid bodies. The first two architectures are based on the recently proposed CNN U-NET and MAgNET (graph U-NET) frameworks which have shown promising performance for learning on mesh-based data. The third arc…
United Imaging Intelligence（联影智能）；UIH America, Inc.（联影美国公司）；United Imaging Healthcare（上海联影医疗科技股份有限公司）
Multi-contrast MRI images provide complementary contrast information about the characteristics of anatomical structures and are commonly used in clinical practice. Recently, a multi-flip-angle (FA) and multi-echo GRE method (MULTIPLEX MRI) has been developed to simultaneously acquire multiple parametric images with just one single scan. However, it poses two challenges for MULTIPLEX to be used in the 3D high-resolution setting: a relatively long scan time and the huge amount of 3D multi-contrast data for reconstruction. Currently, no DL based method has been proposed for 3D MULTIPLEX data reco…
Universidad Nacional de Colombia（哥伦比亚国立大学）；Romanian Institute of Science and Technology (RIST)（罗马尼亚科学技术研究院（RIST））；Munich Re Markets（慕尼黑再保险市场部）
Markov Chain Monte Carlo (MCMC) algorithms are commonly used for their versatility in sampling from complicated probability distributions. However, as the dimension of the distribution gets larger, the computational costs for a satisfactory exploration of the sampling space become challenging. Adaptive MCMC methods employing a choice of proposal distribution can address this issue speeding up the convergence. In this paper we show an alternative way of performing adaptive MCMC, by using the outcome of Bayesian Neural Networks as the initial proposal for the Markov Chain. This combined approach…
Huawei Noah’s Ark Lab（华为诺亚方舟实验室）；University of Sydney（悉尼大学）
Exploring deep convolutional neural networks of high efficiency and low memory usage is very essential for a wide variety of machine learning tasks. Most of existing approaches used to accelerate deep models by manipulating parameters or filters without data, e.g., pruning and decomposition. In contrast, we study this problem from a different perspective by respecting the difference between data. An instance-wise feature pruning is developed by identifying informative features for different instances. Specifically, by investigating a feature decay regularization, we expect intermediate feature…
IBM T. J. Watson Research Center（IBM托马斯·J·沃森研究中心）；Data Centric Systems（数据中心系统）
Data loading can dominate deep neural network training time on large-scale systems. We present a comprehensive study on accelerating data loading performance in large-scale distributed training. We first identify performance and scalability issues in current data loading implementations. We then propose optimizations that utilize CPU resources to the data loader design. We use an analytical model to characterize the impact of data loading on the overall training time and establish the performance trend as we scale up distributed training. Our model suggests that I/O rate limits the scalability…
Hefei University of Technology（合肥工业大学）；Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；National University of Singapore（新加坡国立大学）
Generative neural network is a new category of neural networks and it has been widely utilized in applications such as content generation, unsupervised learning, segmentation and pose estimation. It typically involves massive computing-intensive deconvolution operations that cannot be fitted to conventional neural network processors directly. However, prior works mainly investigated specialized hardware architectures through intensive hardware modifications to the existing deep learning processors to accelerate deconvolution together with the convolution. In contrast, this work proposes a nove…
NVIDIA（英伟达）；Mila（米拉研究所）；McGill University（麦吉尔大学）；Vector Institute（向量研究所）；University of Toronto（多伦多大学）；Université de Montréal（蒙特利尔大学）
We present Kaolin, a PyTorch library aiming to accelerate 3D deep learning research. Kaolin provides efficient implementations of differentiable 3D modules for use in deep learning systems. With functionality to load and preprocess several popular 3D datasets, and native functions to manipulate meshes, pointclouds, signed distance functions, and voxel grids, Kaolin mitigates the need to write wasteful boilerplate code. Kaolin packages together several differentiable graphics modules including rendering, lighting, shading, and view warping. Kaolin also supports an array of loss functions and ev…
Efi Arazi School of Computer Science（埃菲·阿拉齐计算机学院）；The Interdisciplinary Center（跨学科中心）
We present a selective sampling method designed to accelerate the training of deep neural networks. To this end, we introduce a novel measurement, the minimal margin score (MMS), which measures the minimal amount of displacement an input should take until its predicted classification is switched. For multi-class linear classification, the MMS measure is a natural generalization of the margin-based selection criterion, which was thoroughly studied in the binary classification setting. In addition, the MMS measure provides an interesting insight into the progress of the training process and can …
Tsinghua University（清华大学）；National Supercomputing Center in Wuxi（国家超级计算无锡中心）；Beihang University（北京航空航天大学）
This paper reports our efforts on swCaffe, a highly efficient parallel framework for accelerating deep neural networks (DNNs) training on Sunway TaihuLight, the current fastest supercomputer in the world that adopts a unique many-core heterogeneous architecture, with 40,960 SW26010 processors connected through a customized communication network. First, we point out some insightful principles to fully exploit the performance of the innovative many-core architecture. Second, we propose a set of optimization strategies for redesigning a variety of neural network layers based on Caffe. Third, we p…
To address the limitations of existing magnitude-based pruning algorithms in cases where model weights or activations are of large and similar magnitude, we propose a novel perspective to discover parameter redundancy among channels and accelerate deep CNNs via channel pruning. Precisely, we argue that channels revealing similar feature information have functional overlap and that most channels within each such similarity group can be removed without compromising model's representational power. After deriving an effective metric for evaluating channel similarity through probabilistic modeling,…
Deep learning has significantly improved 2D image recognition. Extending into 3D may advance many new applications including autonomous vehicles, virtual and augmented reality, authoring 3D content, and even improving 2D recognition. However despite growing interest, 3D deep learning remains relatively underexplored. We believe that some of this disparity is due to the engineering challenges involved in 3D deep learning, such as efficiently processing heterogeneous data and reframing graphics operations to be differentiable. We address these challenges by introducing PyTorch3D, a library of mo…
Research Center for Brain-inspired Intelligence, CASIA（中国科学院自动化研究所脑启发智能研究中心）；University of Chinese Academy of Sciences（中国科学院大学）；TuSimple（图森未来）；National Laboratory of Pattern Recognition, CASIA（中国科学院自动化研究所模式识别国家重点实验室）；Center for Excellence in Brain Science and Intelligence Technology, CAS（中国科学院脑科学与智能技术卓越创新中心）
We have witnessed rapid evolution of deep neural network architecture design in the past years. These latest progresses greatly facilitate the developments in various areas such as computer vision and natural language processing. However, along with the extraordinary performance, these state-of-the-art models also bring in expensive computational cost. Directly deploying these models into applications with real-time requirement is still infeasible. Recently, Hinton etal. have shown that the dark knowledge within a powerful teacher model can significantly help the training of a smaller and fast…
This paper introduces Selective-Backprop, a technique that accelerates the training of deep neural networks (DNNs) by prioritizing examples with high loss at each iteration. Selective-Backprop uses the output of a training example's forward pass to decide whether to use that example to compute gradients and update parameters, or to skip immediately to the next example. By reducing the number of computationally-expensive backpropagation steps performed, Selective-Backprop accelerates training. Evaluation on CIFAR10, CIFAR100, and SVHN, across a variety of modern image models, shows that Selecti…
Click-through rate (CTR) prediction is a crucial task in online display advertising. The embedding-based neural networks have been proposed to learn both explicit feature interactions through a shallow component and deep feature interactions using a deep neural network (DNN) component. These sophisticated models, however, slow down the prediction inference by at least hundreds of times. To address the issue of significantly increased serving delay and high memory usage for ad serving in production, this paper presents \emph{DeepLight}: a framework to accelerate the CTR predictions in three asp…
Over the last few years, Deep Neural Networks (DNNs) have become ubiquitous owing to their high accuracy on real-world tasks. However, this increase in accuracy comes at the cost of computationally expensive models leading to higher prediction latencies. Prior efforts to reduce this latency such as quantization, model distillation, and any-time prediction models typically trade-off accuracy for performance. In this work, we observe that caching intermediate layer outputs can help us avoid running all the layers of a DNN for a sizeable fraction of inference requests. We find that this can poten…
The method of choice for parameter aggregation in Deep Neural Network (DNN) training, a network-intensive task, is shifting from the Parameter Server model to decentralized aggregation schemes (AllReduce) inspired by theoretical guarantees of better performance. However, current implementations of AllReduce overlook the interdependence of communication and computation, resulting in significant performance degradation. In this paper, we develop Caramel, a system that accelerates decentralized distributed deep learning through model-aware computation scheduling and communication optimizations fo…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
State-of-the-art deep learning systems rely on iterative distributed training to tackle the increasing complexity of models and input data. The iteration time in these communication-heavy systems depends on the computation time, communication time and the extent of overlap of computation and communication. In this work, we identify a shortcoming in systems with graph representation for computation, such as TensorFlow and PyTorch, that result in high variance in iteration time --- random order of received parameters across workers. We develop a system, TicTac, to improve the iteration time by f…
