Neural network pruning is a popular technique used to reduce the inference costs of modern, potentially overparameterized, networks. Starting from a pre-trained network, the process is as follows: remove redundant parameters, retrain, and repeat while maintaining the same test accuracy. The result is a model that is a fraction of the size of the original with comparable predictive performance (test accuracy). Here, we reassess and evaluate whether the use of test accuracy alone in the terminating condition is sufficient to ensure that the resulting model performs well across a wide spectrum of "harder" metrics such as generalization to out-of-distribution data and resilience to noise. Across evaluations on varying architectures and data sets, we find that pruned networks effectively approximate the unpruned model, however, the prune ratio at which pruned networks achieve commensurate performance varies significantly across tasks. These results call into question the extent of \emph{genuine} overparameterization in deep learning and raise concerns about the practicability of deploying pruned networks, specifically in the context of safety-critical systems, unless they are widely evaluated beyond test accuracy to reliably predict their performance. Our code is available at https://github.com/lucaslie/torchprune.

New hardware can substantially increase the speed and efficiency of deep neural network training. To guide the development of future hardware architectures, it is pertinent to explore the hardware and machine learning properties of alternative training algorithms. In this work we evaluate the use of small batch, fine-grained Pipelined Backpropagation, an asynchronous pipeline parallel training algorithm that has significant hardware advantages. We introduce two methods, Spike Compensation and Linear Weight Prediction, that effectively mitigate the downsides caused by the asynchronicity of Pipelined Backpropagation and outperform existing techniques in our setting. We show that appropriate normalization and small batch sizes can also aid training. With our methods, fine-grained Pipelined Backpropagation using a batch size of one can match the accuracy of SGD for multiple networks trained on CIFAR-10 and ImageNet. Simple scaling rules allow the use of existing hyperparameters for traditional training without additional tuning.

The paper proposes and optimizes a partial recovery training system, CPR, for recommendation models. CPR relaxes the consistency requirement by enabling non-failed nodes to proceed without loading checkpoints when a node fails during training, improving failure-related overheads. The paper is the first to the extent of our knowledge to perform a data-driven, in-depth analysis of applying partial recovery to recommendation models and identified a trade-off between accuracy and performance. Motivated by the analysis, we present CPR, a partial recovery training system that can reduce the training time and maintain the desired level of model accuracy by (1) estimating the benefit of partial recovery, (2) selecting an appropriate checkpoint saving interval, and (3) prioritizing to save updates of more frequently accessed parameters. Two variants of CPR, CPR-MFU and CPR-SSU, reduce the checkpoint-related overhead from 8.2--8.5% to 0.53--0.68% compared to full recovery, on a setup emulating the failure pattern and overhead of a production-scale cluster. While reducing overhead significantly, CPR achieves model quality on par with the more expensive full recovery scheme, training the state-of-the-art recommendation model using Criteo’s Terabyte CTR dataset. Our results also suggest that CPR can speed up training on a real production-scale cluster, without notably degrading the accuracy.

Certifying the robustness of neural networks against adversarial attacks is critical to their reliable adoption in real-world systems including autonomous driving and medical diagnosis. Unfortunately, state-of-the-art verifiers either do not scale to larger networks or are too imprecise to prove robustness, which limits their practical adoption. In this work, we introduce GPUPoly, a scalable verifier that can prove the robustness of significantly larger deep neural networks than possible with prior work. The key insight behind GPUPoly is the design of custom, sound polyhedra algorithms for neural network verification on a GPU. Our algorithms leverage the available GPU parallelism and the inherent sparsity of the underlying verification task. GPUPoly scales to very large networks: for example, it can prove the robustness of a 1M neuron, 34-layer deep residual network in $\approx$ 22 seconds. We believe GPUPoly is a promising step towards the practical verification of large real-world networks.

Data access between on- and off-chip memories account for a large fraction of overall energy consumption during inference with deep learning networks. On-chip memory compression can greatly reduce this energy cost as long as it balances the simplicity and low cost of the compression/decompression implementation and its effectiveness in data size reduction. We present Boveda, a simple and effective on-chip lossless memory compression technique for fixed-point precision networks. It reduces data widths by exploiting the value distribution deep learning applications naturally exhibit. Boveda can increase the effective on-chip capacity, reduce off-chip traffic, and/or achieve a desired performance/energy target while using smaller on-chip memories. Boveda can be placed after any memory block in the on-chip memory hierarchy and can work with \textul{any} data-parallel processing units such as the vector-like or the tensorcore units of modern graphics processors, systolic arrays such as that used in the Tensor Processing Unit, and units that process sparse tensors such as those used in the SCNN accelerator. To demonstrate the potential of Boveda, we implement it over (i) SCNN, a state-of-the-art accelerator for sparse networks, (ii) a Tensorcore-like architecture, and (iii) TPU. Boveda reduces memory footprint by 34\% for SCNN and sparse models on top of zero compression. For dense models, Boveda improves compression by 47\%. We also present a prototype FPGA implementation.

The boom of edge AI applications has spawned a great many neural network (NN) algorithms and inference platforms. Unfortunately, the fast pace of development in their fields have magnified the gaps between them. A well-designed NN algorithm with reduced number of computation operations and memory accesses can easily result in increased inference latency in real-world deployment, due to a mismatch between the algorithm and the features of target platforms.

Therefore, it is critical to understand the behaviour characteristics of NN design space on target platforms. However, none of existing NN benchmarking or characterization studies can serve this purpose. They only evaluate some sparse configurations in the design space for the purpose of platform optimization rather than the scaling in every design dimension for NN algorithm efficiency. This paper presents the first empirical study on the NN design space to learn NN behaviour characteristics on different inference platforms. The revealed characteristics can be used as guidelines to design efficient NN algorithms. We profile ten-thousand configurations from a cutting-edge NN design space on seven industrial edge AI platforms. Seven key findings as well as their causes and implications for efficient NN design are highlighted.

Deep learning implementations on CPUs (Central Processing Units) are gaining more traction. Enhanced AI capabilities on commodity x86 architectures are commercially appealing due to the reuse of existing hardware and virtualization ease. A notable work in this direction is the SLIDE system. SLIDE is a C++ implementation of a sparse hash table based back-propagation, which was shown to be significantly faster than GPUs in training hundreds of million parameter neural models. In this paper, we argue that SLIDE's current implementation is sub-optimal and does not exploit several opportunities available in modern CPUs. In particular, we show how SLIDE's computations allow for a unique possibility of vectorization via AVX (Advanced Vector Extensions)-512. Furthermore, we highlight opportunities for different kinds of memory optimization and quantizations. Combining all of them, we obtain up to 7x speedup in the computations on the same hardware. Our experiments are focused on large (hundreds of millions of parameters) recommendation and NLP models. Our work highlights several novel perspectives and opportunities for implementing randomized algorithms for deep learning on modern CPUs.

Swift for TensorFlow is a deep learning platform that scales from mobile devices to clusters of hardware accelerators in data centers. It combines a language-integrated automatic differentiation system and multiple Tensor implementations within a modern ahead-of-time compiled language oriented around mutable value semantics. The resulting platform has been validated through use in over 30 deep learning models and and has been employed across data center and mobile applications.

One of the major optimizations employed in deep learning frameworks is graph rewriting. Production frameworks rely on heuristics to decide if rewrite rules should be applied and in which order. Prior research has shown that one can discover more optimal tensor computation graphs if we search for a better sequence of substitutions instead of relying on heuristics. However, we observe that existing approaches for tensor graph superoptimization both in production and research frameworks apply substitutions in a sequential manner. Such sequential search methods are sensitive to the order in which the substitutions are applied and often only explore a small fragment of the exponential space of equivalent graphs. This paper presents a novel technique for tensor graph superoptimization that employs equality saturation to apply all possible substitutions at once. We show that our approach can find optimized graphs with up to 16% speedup over state-of-the-art, while spending on average 48x less time optimizing.

Distributed training techniques have been widely deployed in large-scale deep models training on dense-GPU clusters. However, on public cloud clusters, due to the moderate inter-connection bandwidth between instances, traditional state-of-the-art distributed training systems cannot scale well in training large-scale models. In this paper, we propose a new computing and communication efficient top-k sparsification communication library for distributed training. To further improve the system scalability, we optimize I/O by proposing a simple yet efficient multi-level data caching mechanism and optimize the update operation by introducing a novel parallel tensor operator. Experimental results on a 16-node Tencent Cloud cluster (each node with 8 Nvidia Tesla V100 GPUs) show that our system achieves 25%-40% faster than existing state-of-the-art systems on CNNs and Transformer. We finally break the record on DAWNBench on training ResNet-50 to 93% top-5 accuracy on ImageNet.

We study the problem of using low computational cost to automate the choices of learners and hyperparameters for an ad-hoc training dataset and error metric, by conducting trials of different configurations on the given training data. We investigate the joint impact of multiple factors on both trial cost and model error, and propose several design guidelines. Following them, we build a fast and lightweight library FLAML which optimizes for low computational resource in finding accurate models. FLAML integrates several simple but effective search strategies into an adaptive system. It significantly outperforms top-ranked AutoML libraries on a large open source AutoML benchmark under equal, or sometimes orders of magnitude smaller budget constraints.

Outlier detection (OD) is a key machine learning (ML) task for identifying abnormal objects from general samples with numerous high-stake applications including fraud detection and intrusion detection. Due to the lack of ground truth labels, practitioners often have to build a large number of unsupervised, heterogeneous models (i.e., different algorithms with varying hyperparameters) for further combination and analysis, rather than relying on a single model. How to accelerate the training and scoring on new-coming samples by outlyingness (referred as prediction throughout the paper) with a large number of unsupervised, heterogeneous OD models? In this study, we propose a modular acceleration system, called SUOD, to address it. The proposed system focuses on three complementary acceleration aspects (data reduction for high-dimensional data, approximation for costly models, and taskload imbalance optimization for distributed environment), while maintaining performance accuracy. Extensive experiments on more than 20 benchmark datasets demonstrate SUOD's effectiveness in heterogeneous OD acceleration, along with a real-world deployment case on fraudulent claim analysis at IQVIA, a leading healthcare firm. We open-source SUOD for reproducibility and accessibility.

Structured matrices, such as those derived from Kronecker products (KP), are effective at compressing neural networks, but can lead to unacceptable accuracy loss when applied to large models. In this paper, we propose the notion of doping - addition of an extremely sparse matrix to a structured matrix. Doping facilitates additional degrees of freedom for a small number of parameters, allowing them to independently diverge from the fixed structure. To train LSTMs with doped structured matrices, we introduce the additional parameter matrix while slowly annealing its sparsity level. However, we find that performance degrades as we slowly sparsify the doping matrix, due to co-matrix adaptation (CMA) between the structured and the sparse matrices. We address this over dependence on the sparse matrix using a co-matrix dropout regularization (CMR) scheme. We provide empirical evidence to show that doping, CMA and CMR are concepts generally applicable to multiple structured matrices (Kronecker Product, LMF, Hybrid Matrix Decomposition). Additionally, results with doped kronecker product matrices demonstrate state-of-the-art accuracy at large compression factors (10 − 25x) across 4 natural language processing applications with minor loss in accuracy. Doped KP compression technique outperforms previous state-of-the art compression results by achieving 1.3−2.4x higher compression factor at a similar accuracy, while also beating strong alternatives like pruning and low-rank methods by a large margin (8% or more). Additionally, we show that doped KP can be deployed on commodity hardware using the current software stack and achieve 2.5 − 5.5x inference run-time speed-up over baseline.

The rapid demand for memory and computational resources by the emerging complex applications requires multi-core parallel systems capable to scale the execution of these applications. In this paper, we propose a distributed graph-theoretic framework for automatic parallelization in multi-core systems, where the goal is to minimize the data communication while accounting for intrinsic functional interdependence and balancing the workloads among cores to improve the overall performance. Specifically, we design a general and flexible greedy-based vertex cut framework for partitioning LLVM IR graphs into clusters while taking into consideration the data communication and workload balance among clusters. Then, we map the clusters generated by the vertex cut algorithms onto a non-uniform memory access multi-core platform. Experimental results demonstrate that our proposed WB-Libra algorithm provides performance improvements of 1.56x and 1.86x over existing state-of-the-art approaches for 8 and 1024 clusters running on a multi-core platform, respectively.

Driven by the tremendous effort in researching novel deep learning (DL) algorithms, the training cost of developing new models increases staggeringly in recent years. We analyze GPU cluster usage statistics from a top research institute for more insights into the hardware efficiency achieved by typical DL training jobs. Our study reveals that single-accelerator training jobs can dominate the cluster-wide resource consumption when launched repetitively (e.g., for hyper-parameter tuning) while severely under-utilizing the hardware. Fortunately, we observe that such workloads have the following unique characteristics: (i) the models among jobs often have the same types of operators with the same shapes, and (ii) the inter-model horizontal fusion of such operators is mathematically equivalent to other already well-optimized operators. Thus, to help DL researchers and practitioners effectively improve the hardware utilization of their novel DL training workloads, we propose Horizontally Fused Training Array (HFTA). HFTA is a new DL framework extension library that horizontally fuses the models from different repetitive jobs deeply down to operators and then trains them simultaneously on a shared accelerator. To show the generality of our solution, we apply HFTA to six DL models training on state-of-the-art accelerators (GPUs and TPUs). Our results indicate that HFTA is highly effective in improving hardware utilization and achieves up to 15.1x higher training throughput vs. the standard practice of running each job on a separate accelerator.

DNNs have revolutionized across a wide range of applications, such as image classification, speech recognition and robotics control. As DNN models become more computationally expensive to train, parallel execution with multiple accelerators (e.g. GPUs) is adopted. System efficiency is a big issue when scaling out. However, as computation power increases, GPUs are under-utilized mainly due to limited local memory size. To address this memory bound, we present Wavelet, an efficient and generic approach that can fully utilize all the available on-device memory among GPUs involved in the distributed training job. Wavelet achieves near optimal on-device memory usage by adopting a simple scheduling scheme called Tick-Tock, which interleaves waves of peak memory usage among the accelerators. Evaluations on a variety of DNN models and tasks show that, Wavelet trains models up to 6.7x faster than commonly used parallelism techniques.

Manual debugging is a common productivity drain in the machine learning (ML) lifecycle. Identifying underperforming training jobs requires constant developer attention and deep domain expertise. As state-of-the-art models grow in size and complexity, debugging becomes increasingly difficult. Just as unit tests boost traditional software development, an automated ML debugging library can save time and money. We present Amazon SageMaker Debugger, a machine learning feature that automatically identifies and stops underperforming training jobs. Debugger is a new feature of Amazon SageMaker that automatically captures relevant data during training and evaluation and presents it for online and offline inspection. Debugger helps users define a set of conditions, in the form of built-in or custom rules, that are applied to this data, thereby enabling users to catch training issues as well as monitor and debug ML model training in real-time. These rules save time and money by alerting the developer and terminating a problematic training job early.