Communication overhead poses an important obstacle to distributed DNN training and draws increasing attention in recent years. Despite continuous efforts, prior solutions such as gradient compression/reduction, compute/communication overlapping and layer-wise flow scheduling, etc., are still coarse-grained and insufficient for an efficient distributed training especially when the network is under pressure. We present DLCP, a novel solution exploiting the domain-specific properties of deep learning to optimize communication overhead of DNN training in a fine-grained manner. At its heart, DLCP comprises of several key innovations beyond prior work: e.g., it exploits {\em bounded loss tolerance} of SGD-based training to improve tail communication latency which cannot be avoided purely through gradient compression. It then performs fine-grained packet-level prioritization and dropping, as opposed to flow-level scheduling, based on layers and magnitudes of gradients to further speedup model convergence without affecting accuracy. In addition, it leverages inter-packet order-independency to perform per-packet load balancing without causing classical re-ordering issues. DLCP works with both Parameter Server and collective communication routines. We have implemented DLCP with commodity switches, integrated it with various training frameworks including TensorFlow, MXNet and PyTorch, and deployed it in our small-scale testbed with 10 Nvidia V100 GPUs. Our testbed experiments and large-scale simulations show that DLCP delivers up to $84.3\%$ additional training acceleration over the best existing solutions.
Nonlinear conjugate gradient (NLCG) based optimizers have shown superior loss convergence properties compared to gradient descent based optimizers for traditional optimization problems. However, in Deep Neural Network (DNN) training, the dominant optimization algorithm of choice is still Stochastic Gradient Descent (SGD) and its variants. In this work, we propose and evaluate the stochastic preconditioned nonlinear conjugate gradient algorithm for large scale DNN training tasks. We show that a nonlinear conjugate gradient algorithm improves the convergence speed of DNN training, especially in …
Synchronous strategies with data parallelism, such as the Synchronous StochasticGradient Descent (S-SGD) and the model averaging methods, are widely utilizedin distributed training of Deep Neural Networks (DNNs), largely owing to itseasy implementation yet promising performance. Particularly, each worker ofthe cluster hosts a copy of the DNN and an evenly divided share of the datasetwith the fixed mini-batch size, to keep the training of DNNs convergence. In thestrategies, the workers with different computational capability, need to wait foreach other because of the synchronization and delays …
The increasing complexity of deep neural networks (DNNs) has made it challenging to exploit existing large-scale data processing pipelines for handling massive data and parameters involved in DNN training. Distributed computing platforms and GPGPU-based acceleration provide a mainstream solution to this computational challenge. In this paper, we propose DeepSpark, a distributed and parallel deep learning framework that exploits Apache Spark on commodity clusters. To support parallel operations, DeepSpark automatically distributes workloads and parameters to Caffe/Tensorflow-running nodes using…
Recent developments on large-scale distributed machine learning applications, e.g., deep neural networks, benefit enormously from the advances in distributed non-convex optimization techniques, e.g., distributed Stochastic Gradient Descent (SGD). A series of recent works study the linear speedup property of distributed SGD variants with reduced communication. The linear speedup property enable us to scale out the computing capability by adding more computing nodes into our system. The reduced communication complexity is desirable since communication overhead is often the performance bottleneck…
This paper presents TAG, an automatic system to derive optimized DNN training graph and its deployment onto any device topology, for expedited training in device- and topology- heterogeneous ML clusters. We novelly combine both the DNN computation graph and the device topology graph as input to a graph neural network (GNN), and join the GNN with a search-based method to quickly identify optimized distributed training strategies. To reduce communication in a heterogeneous cluster, we further explore a lossless gradient compression technique and solve a combinatorial optimization problem to auto…
Pukyong National University（釜庆国立大学）；Korea University（高丽大学）；Singapore University of Technology and Design（新加坡科技设计大学）；Kyung Hee University（庆熙大学）
This paper studies a deep learning (DL) framework to solve distributed non-convex constrained optimizations in wireless networks where multiple computing nodes, interconnected via backhaul links, desire to determine an efficient assignment of their states based on local observations. Two different configurations are considered: First, an infinite-capacity backhaul enables nodes to communicate in a lossless way, thereby obtaining the solution by centralized computations. Second, a practical finite-capacity backhaul leads to the deployment of distributed solvers equipped along with quantizers fo…
University of British Columbia（不列颠哥伦比亚大学）；Vector Institute（向量学院）；Carnegie Mellon University（卡内基梅隆大学）；University of Toronto（多伦多大学）
Data parallel training is widely used for scaling distributed deep neural network (DNN) training. However, the performance benefits are often limited by the communication-heavy parameter synchronization step. In this paper, we take advantage of the domain specific knowledge of DNN training and overlap parameter synchronization with computation in order to improve the training performance. We make two key observations: (1) the optimal data representation granularity for the communication may differ from that used by the underlying DNN model implementation and (2) different parameters can afford…
UC Irvine（加州大学欧文分校）；NGD Systems Inc.（NGD Systems公司）
Distributed training is a novel approach to accelerate Deep Neural Networks (DNN) training, but common training libraries fall short of addressing the distributed cases with heterogeneous processors or the cases where the processing nodes get interrupted by other workloads. This paper describes distributed training of DNN on computational storage devices (CSD), which are NAND flash-based, high capacity data storage with internal processing engines. A CSD-based distributed architecture incorporates the advantages of federated learning in terms of performance scalability, resiliency, and data pr…
University of Hong Kong（香港大学）；ByteDance Inc.（字节跳动公司）
Distributed training using multiple devices (e.g., GPUs) has been widely adopted for learning DNN models over large datasets. However, the performance of large-scale distributed training tends to be far from linear speed-up in practice. Given the complexity of distributed systems, it is challenging to identify the root cause(s) of inefficiency and exercise effective performance optimizations when unexpected low training speed occurs. To date, there exists no software tool which diagnoses performance issues and helps expedite distributed DNN training, while the training can be run using differe…
College of Computer and Control Engineering, Nankai University（南开大学计算机与控制工程学院）；Microsoft Research（微软研究院）
Data parallelism has emerged as a necessary technique to accelerate the training of deep neural networks (DNN). In a typical data parallelism approach, the local workers push the latest updates of all the parameters to the parameter server and pull all merged parameters back periodically. However, with the increasing size of DNN models and the large number of workers in practice, this typical data parallelism cannot achieve satisfactory training acceleration, since it usually suffers from the heavy communication cost due to transferring huge amount of information between workers and the parame…
Queen’s University Belfast（贝尔法斯特女王大学）；University of St Andrews（圣安德鲁斯大学）
Partitioning and deploying Deep Neural Networks (DNNs) across edge nodes may be used to meet performance objectives of applications. However, the failure of a single node may result in cascading failures that will adversely impact the delivery of the service and will result in failure to meet specific objectives. The impact of these failures needs to be minimised at runtime. Three techniques are explored in this paper, namely repartitioning, early-exit and skip-connection. When an edge node fails, the repartitioning technique will repartition and redeploy the DNN thus avoiding the failed nodes…
It is important to scale out deep neural network (DNN) training for reducing model training time. The high communication overhead is one of the major performance bottlenecks for distributed DNN training across multiple GPUs. Our investigations have shown that popular open-source DNN systems could only achieve 2.5 speedup ratio on 64 GPUs connected by 56 Gbps network. To address this problem, we propose a communication backend named GradientFlow for distributed DNN training, and employ a set of network optimization techniques. First, we integrate ring-based allreduce, mixed-precision training, …
The Chinese University of Hong Kong（香港中文大学）；Huawei Technologies Co. Ltd（华为技术有限公司）
A good parallelization strategy can significantly improve the efficiency or reduce the cost for the distributed training of deep neural networks (DNNs). Recently, several methods have been proposed to find efficient parallelization strategies but they all optimize a single objective (e.g., execution time, memory consumption) and produce only one strategy. We propose FT, an efficient algorithm that searches for an optimal set of parallelization strategies to allow the trade-off among different objectives. FT can adapt to different scenarios by minimizing the memory consumption when the number o…
Hong Kong University of Science and Technology（香港科技大学）；Huawei（华为）
Bulk synchronous parallel (BSP) is the de-facto paradigm for distributed DNN training in today's production clusters. However, due to the global synchronization nature, its performance can be significantly influenced by network bottlenecks caused by either static topology heterogeneity or dynamic bandwidth contentions. Existing solutions, either system-level optimizations strengthening BSP (e.g., Ring or Hierarchical All-reduce) or algorithmic optimizations replacing BSP (e.g., ASP or SSP, which relax the global barriers), do not completely solve the problem, as they may still suffer from comm…
This paper proposes DisCo, an automatic deep learning compilation module for data-parallel distributed training. Unlike most deep learning compilers that focus on training or inference on a single device, DisCo optimizes a DNN model for distributed training over multiple GPU machines. Existing single-device compilation strategies do not work well in distributed training, due mainly to communication inefficiency that they incur. DisCo generates optimized, joint computation operator and communication tensor fusion strategies to enable highly efficient distributed training. A GNN-based simulator …
PipeDream is a Deep Neural Network(DNN) training system for GPUs that parallelizes computation by pipelining execution across multiple machines. Its pipeline parallel computing model avoids the slowdowns faced by data-parallel training when large models and/or limited network bandwidth induce high communication-to-computation ratios. PipeDream reduces communication by up to 95% for large DNNs relative to data-parallel training, and allows perfect overlap of communication and computation. PipeDream keeps all available GPUs productive by systematically partitioning DNN layers among them to balan…
Training deep neural networks (DNNs) in large-cluster computing environments is increasingly necessary, as networks grow in size and complexity. Local memory and processing limitations require robust data and model parallelism for crossing compute node boundaries. We propose a linear-algebraic approach to model parallelism in deep learning, which allows parallel distribution of any tensor in the DNN. Rather than rely on automatic differentiation tools, which do not universally support distributed memory parallelism models, we show that parallel data movement operations, e.g., broadcast, sum-re…
The training process of Deep Neural Network (DNN) is compute-intensive, often taking days to weeks to train a DNN model. Therefore, parallel execution of DNN training on GPUs is a widely adopted approach to speed up the process nowadays. Due to the implementation simplicity, data parallelism is currently the most commonly used parallelization method. Nonetheless, data parallelism suffers from excessive inter-GPU communication overhead due to frequent weight synchronization among GPUs. Another approach is pipelined model parallelism, which partitions a DNN model among GPUs, and processes multip…
Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this survey, we describe the problem from a theoretical perspective, followed by approaches for its parallelization. We present trends in DNN architectures and the resulting implications on parallelization strategies. We then review and model the different types of concurrency in DNNs: from the single operator, through parallelism in network inference and training, to distri…
Federated Learning (FL) empowers Industrial Internet of Things (IIoT) with distributed intelligence of industrial automation thanks to its capability of distributed machine learning without any raw data exchange. However, it is rather challenging for lightweight IIoT devices to perform computation-intensive local model training over large-scale deep neural networks (DNNs). Driven by this issue, we develop a communication-computation efficient FL framework for resource-limited IIoT networks that integrates DNN partition technique into the standard FL mechanism, wherein IIoT devices perform loca…
Following the recent success of deep neural networks (DNN) on video computer vision tasks, performing DNN inferences on videos that originate from mobile devices has gained practical significance. As such, previous approaches developed methods to offload DNN inference computations for images to cloud servers to manage the resource constraints of mobile devices. However, when it comes to video data, communicating information of every frame consumes excessive network bandwidth and renders the entire system susceptible to adverse network conditions such as congestion. Thus, in this work, we seek …
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 …
Distributed Deep Learning using Stochastic Gradient Staleness
Information Technology Department, Posts and Telecommunications Institute of Technology (PTIT)（信息科技部，邮电技术研究所）；Department of Mechanical and Robotics Engineering, Gwangju Institute of Science and Technology（机械与机器人工程系，全州科学技术院）
Hong Kong University of Science and Technology（香港科技大学）；University of Michigan（密歇根大学）
Training deep neural networks (DNNs) is time-consuming. While most existing solutions try to overlap/schedule computation and communication for efficient training, this paper goes one step further by skipping computing and communication through DNN layer freezing. Our key insight is that the training progress of internal DNN layers differs significantly, and front layers often become well-trained much earlier than deep layers. To explore this, we first introduce the notion of training plasticity to quantify the training progress of internal DNN layers. Then we design Egeria, a knowledge-guided…
University of Otago（奥塔哥大学）；Qilu University of Technology（齐鲁工业大学）
Communication efficiency plays an important role in accelerating the distributed training of Deep Neural Networks (DNN). All-reduce is the crucial communication primitive to reduce model parameters in distributed DNN training. Most existing all-reduce algorithms are designed for traditional electrical interconnect systems, which cannot meet the communication requirements for distributed training of large DNNs due to the low data bandwidth of the electrical interconnect systems. One of the promising alternatives for electrical interconnect is optical interconnect, which can provide high bandwid…
Many recent pattern recognition applications rely on complex distributed architectures in which sensing and computational nodes interact together through a communication network. Deep neural networks (DNNs) play an important role in this scenario, furnishing powerful decision mechanisms, at the price of a high computational effort. Consequently, powerful state-of-the-art DNNs are frequently split over various computational nodes, e.g., a first part stays on an embedded device and the rest on a server. Deciding where to split a DNN is a challenge in itself, making the design of deep learning ap…
The last decade has witnessed growth in the computational requirements for training deep neural networks. Current approaches (e.g., data/model parallelism, pipeline parallelism) parallelize training tasks onto multiple devices. However, these approaches always rely on specific deep learning frameworks and requires elaborate manual design, which make it difficult to maintain and share between different type of models. In this paper, we propose Auto-MAP, a framework for exploring distributed execution plans for DNN workloads, which can automatically discovering fast parallelization strategies th…
To train modern large DNN models, pipeline parallelism has recently emerged, which distributes the model across GPUs and enables different devices to process different microbatches in pipeline. Earlier pipeline designs allow multiple versions of model parameters to co-exist (similar to asynchronous training), and cannot ensure the same model convergence and accuracy performance as without pipelining. Synchronous pipelining has recently been proposed which ensures model performance by enforcing a synchronization barrier between training iterations. Nonetheless, the synchronization barrier requi…
Duke University（杜克大学）；University of Southern California（南加州大学）
With the rise of artificial intelligence in recent years, Deep Neural Networks (DNNs) have been widely used in many domains. To achieve high performance and energy efficiency, hardware acceleration (especially inference) of DNNs is intensively studied both in academia and industry. However, we still face two challenges: large DNN models and datasets, which incur frequent off-chip memory accesses; and the training of DNNs, which is not well-explored in recent accelerator designs. To truly provide high throughput and energy efficient acceleration for the training of deep and large models, we ine…
Hong Kong University of Science and Technology（香港科技大学）；National University of Defense Technology（国防科技大学）
ByteScheduler partitions and rearranges tensor transmissions to improve the communication efficiency of distributed Deep Neural Network (DNN) training. The configuration of hyper-parameters (i.e., the partition size and the credit size) is critical to the effectiveness of partitioning and rearrangement. Currently, ByteScheduler adopts Bayesian Optimization (BO) to find the optimal configuration for the hyper-parameters beforehand. In practice, however, various runtime factors (e.g., worker node status and network conditions) change over time, making the statically-determined one-shot configura…
Recently, Deep Neural Networks (DNNs) have recorded great success in handling medical and other complex classification tasks. However, as the sizes of a DNN model and the available dataset increase, the training process becomes more complex and computationally intensive, which usually takes a longer time to complete. In this work, we have proposed a generic full end-to-end hybrid parallelization approach combining both model and data parallelism for efficiently distributed and scalable training of DNN models. We have also proposed a Genetic Algorithm based heuristic resources allocation mechan…
SlimEdge: Performance and Device Aware Distributed DNN Deployment on Resource-Constrained Edge Hardware
Amrita Vishwa Vidyapeetham（阿姆里塔印度教学院）；Intel Corporation（英特尔公司）；Indian Institute of Science（印度科学研究院）；Analog Devices, Inc.（模拟器件公司）；Birla Institute of Technology and Science-Pilani, Dubai Campus（比拉技术科学学院-迪拜校区）
TensorFlow has been the most widely adopted Machine/Deep Learning framework. However, little exists in the literature that provides a thorough understanding of the capabilities which TensorFlow offers for the distributed training of large ML/DL models that need computation and communication at scale. Most commonly used distributed training approaches for TF can be categorized as follows: 1) Google Remote Procedure Call (gRPC), 2) gRPC+X: X=(InfiniBand Verbs, Message Passing Interface, and GPUDirect RDMA), and 3) No-gRPC: Baidu Allreduce with MPI, Horovod with MPI, and Horovod with NVIDIA NCCL.…
We propose TopoOpt, a novel direct-connect fabric for deep neural network (DNN) training workloads. TopoOpt co-optimizes the distributed training process across three dimensions: computation, communication, and network topology. We demonstrate the mutability of AllReduce traffic, and leverage this property to construct efficient network topologies for DNN training jobs. TopoOpt then uses an alternating optimization technique and a group theory-inspired algorithm called TotientPerms to find the best network topology and routing plan, together with a parallelization strategy. We build a fully fu…
Most work in the deep learning systems community has focused on faster inference, but arriving at a trained model requires lengthy experiments. Accelerating training lets developers iterate faster and come up with better models. DNN training is often seen as a compute-bound problem, best done in a single large compute node with many GPUs. As DNNs get bigger, training requires going distributed. Distributed deep neural network (DDNN) training constitutes an important workload on the cloud. Larger DNN models and faster compute engines shift the training performance bottleneck from computation to…
With the increased penetration and proliferation of Internet of Things (IoT) devices, there is a growing trend towards distributing the power of deep learning (DL) across edge devices rather than centralizing it in the cloud. This development enables better privacy preservation, real-time responses, and user-specific models. To deploy deep and complex models to edge devices with limited resources, model partitioning of deep neural networks (DNN) model is necessary, and has been widely studied. However, most of the existing literature only considers distributing the inference model while still …
Parameter updating is an important stage in parallelism-based distributed deep learning. Synchronous methods are widely used in distributed training the Deep Neural Networks (DNNs). To reduce the communication and synchronization overhead of synchronous methods, decreasing the synchronization frequency (e.g., every $n$ mini-batches) is a straightforward approach. However, it often suffers from poor convergence. In this paper, we propose a new algorithm of integrating Particle Swarm Optimization (PSO) into the distributed training process of DNNs to automatically compute new parameters. In the …
The Hong Kong University of Science and Technology（香港科技大学）
The second-order optimization methods, notably the D-KFAC (Distributed Kronecker Factored Approximate Curvature) algorithms, have gained traction on accelerating deep neural network (DNN) training on GPU clusters. However, existing D-KFAC algorithms require to compute and communicate a large volume of second-order information, i.e., Kronecker factors (KFs), before preconditioning gradients, resulting in large computation and communication overheads as well as a high memory footprint. In this paper, we propose DP-KFAC, a novel distributed preconditioning scheme that distributes the KF construct…
Tri-Institutional Research Center for Neuroimaging and Data Science (TReNDS)（三机构神经影像与数据科学研究中心（TReNDS））；Georgia State University（佐治亚州立大学）；Georgia Institue of Technology（佐治亚理工学院）；Maynooth University（梅努斯大学）
Although distributed machine learning has opened up many new and exciting research frontiers, fragmentation of models and data across different machines, nodes, and sites still results in considerable communication overhead, impeding reliable training in real-world contexts. The focus on gradients as the primary shared statistic during training has spawned a number of intuitive algorithms for distributed deep learning; however, gradient-centric training of large deep neural networks (DNNs) tends to be communication-heavy, often requiring additional adaptations such as sparsity constraints, com…
Recent trends in high-performance computing and deep learning have led to the proliferation of studies on large-scale deep neural network training. However, the frequent communication requirements among computation nodes drastically slows the overall training speeds, which causes bottlenecks in distributed training, particularly in clusters with limited network bandwidths. To mitigate the drawbacks of distributed communications, researchers have proposed various optimization strategies. In this paper, we provide a comprehensive survey of communication strategies from both an algorithm viewpoin…
University of Science and Technology of China（中国科学技术大学）；Huawei Technologies（华为技术有限公司）
The size of deep neural networks (DNNs) grows rapidly as the complexity of the machine learning algorithm increases. To satisfy the requirement of computation and memory of DNN training, distributed deep learning based on model parallelism has been widely recognized. We propose a new pipeline parallelism training framework, BaPipe, which can automatically explore pipeline parallelism training methods and balanced partition strategies for DNN distributed training. In BaPipe, each accelerator calculates the forward propagation and backward propagation of different parts of networks to implement …
Hamad Bin Khalifa University（哈马德·本·哈利法大学）；Qatar University（卡塔尔大学）；Mohamed Bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）
Although Deep Neural Networks (DNN) have become the backbone technology of several ubiquitous applications, their deployment in resource-constrained machines, e.g., Internet of Things (IoT) devices, is still challenging. To satisfy the resource requirements of such a paradigm, collaborative deep inference with IoT synergy was introduced. However, the distribution of DNN networks suffers from severe data leakage. Various threats have been presented, including black-box attacks, where malicious participants can recover arbitrary inputs fed into their devices. Although many countermeasures were d…
Carnegie Mellon University（卡内基梅隆大学）；Seoul National University（首尔国立大学）；Institute for Infocomm Research（资讯通信研究院）
Deep learning (DL) has achieved notable successes in many machine learning tasks. A number of frameworks have been developed to expedite the process of designing and training deep neural networks (DNNs), such as Caffe, Torch and Theano. Currently they can harness multiple GPUs on a single machine, but are unable to use GPUs that are distributed across multiple machines; as even average-sized DNNs can take days to train on a single GPU with 100s of GBs to TBs of data, distributed GPUs present a prime opportunity for scaling up DL. However, the limited bandwidth available on commodity Ethernet n…
