EfficientNets are a family of state-of-the-art image classification models based on efficiently scaled convolutional neural networks. Currently, EfficientNets can take on the order of days to train; for example, training an EfficientNet-B0 model takes 23 hours on a Cloud TPU v2-8 node. In this paper, we explore techniques to scale up the training of EfficientNets on TPU-v3 Pods with 2048 cores, motivated by speedups that can be achieved when training at such scales. We discuss optimizations required to scale training to a batch size of 65536 on 1024 TPU-v3 cores, such as selecting large batch optimizers and learning rate schedules as well as utilizing distributed evaluation and batch normalization techniques. Additionally, we present timing and performance benchmarks for EfficientNet models trained on the ImageNet dataset in order to analyze the behavior of EfficientNets at scale. With our optimizations, we are able to train EfficientNet on ImageNet to an accuracy of 83% in 1 hour and 4 minutes.
This paper introduces EfficientNetV2, a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop this family of models, we use a combination of training-aware neural architecture search and scaling, to jointly optimize training speed and parameter efficiency. The models were searched from the search space enriched with new ops such as Fused-MBConv. Our experiments show that EfficientNetV2 models train much faster than state-of-the-art models while being up to 6.8x smaller. Our training can be further sped up by progre…
Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. To go even …
The convolutional neural networks (CNNs) trained on ILSVRC12 ImageNet were the backbone of various applications as a generic classifier, a feature extractor or a base model for transfer learning. This paper describes automated heuristics based on model consensus, explainability and confident learning to correct labeling mistakes and remove ambiguous images from this dataset. After making these changes on the training and validation sets, the ImageNet-Clean improves the model performance by 2-2.4 % for SqueezeNet and EfficientNet-B0 models. The results support the importance of larger image cor…
Embedding Convolutional Neural Network (CNN) into edge devices for inference is a very challenging task because such lightweight hardware is not born to handle this heavyweight software, which is the common overhead from the modern state-of-the-art CNN models. In this paper, targeting at reducing the overhead with trading the accuracy as less as possible, we propose a novel of Network Candidate Search (NCS), an alternative way to study the trade-off between the resource usage and the performance through grouping concepts and elimination tournament. Besides, NCS can also be generalized across a…
Tsinghua University（清华大学）；Huawei Singapore Research Center（华为新加坡研究中心）
Recently, it was found that many real-world examples without intentional modifications can fool machine learning models, and such examples are called "natural adversarial examples". ImageNet-A is a famous dataset of natural adversarial examples. By analyzing this dataset, we hypothesized that large, cluttered and/or unusual background is an important reason why the images in this dataset are difficult to be classified. We validated the hypothesis by reducing the background influence in ImageNet-A examples with object detection techniques. Experiments showed that the object detection models wit…
This paper provides an extensive analysis of the performance of the EfficientNet image classifiers with several recent training procedures, in particular one that corrects the discrepancy between train and test images. The resulting network, called FixEfficientNet, significantly outperforms the initial architecture with the same number of parameters. For instance, our FixEfficientNet-B0 trained without additional training data achieves 79.3% top-1 accuracy on ImageNet with 5.3M parameters. This is a +0.5% absolute improvement over the Noisy student EfficientNet-B0 trained with 300M unlabeled i…
Google Research（谷歌研究院）；Carnegie Mellon University（卡内基梅隆大学）
We present Noisy Student Training, a semi-supervised learning approach that works well even when labeled data is abundant. Noisy Student Training achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. Noisy Student Training extends the idea of self-training and distillation with the use of equal-or-larger…
Neural Architecture Search (NAS) yields state-of-the-art neural networks that outperform their best manually-designed counterparts. However, previous NAS methods search for architectures under one set of training hyper-parameters (i.e., a training recipe), overlooking superior architecture-recipe combinations. To address this, we present Neural Architecture-Recipe Search (NARS) to search both (a) architectures and (b) their corresponding training recipes, simultaneously. NARS utilizes an accuracy predictor that scores architecture and training recipes jointly, guiding both sample selection and…
University of Maryland（马里兰大学）；University of Southern California（南加州大学）
Large-scale training of modern deep learning models heavily relies on publicly available data on the web. This potentially unauthorized usage of online data leads to concerns regarding data privacy. Recent works aim to make unlearnable data for deep learning models by adding small, specially designed noises to tackle this issue. However, these methods are vulnerable to adversarial training (AT) and/or are computationally heavy. In this work, we propose a novel, model-free, Convolution-based Unlearnable DAtaset (CUDA) generation technique. CUDA is generated using controlled class-wise convoluti…
North South University（北南大学）；Concordia University（康考迪亚大学）
Melanoma is regarded as the most threatening among all skin cancers. There is a pressing need to build systems which can aid in the early detection of melanoma and enable timely treatment to patients. Recent methods are geared towards machine learning based systems where the task is posed as image recognition, tag dermoscopic images of skin lesions as melanoma or non-melanoma. Even though these methods show promising results in terms of accuracy, they are computationally quite expensive to train, that questions the ability of these models to be deployable in a clinical setting or memory constr…
University of Chinese Academy of Sciences（中国科学院大学）；Microsoft Research（微软研究院）；City University of Hong Kong（香港城市大学）；Stony Brook University（石溪大学）
Differentiable ARchiTecture Search, i.e., DARTS, has drawn great attention in neural architecture search. It tries to find the optimal architecture in a shallow search network and then measures its performance in a deep evaluation network. The independent optimization of the search and evaluation networks, however, leaves room for potential improvement by allowing interaction between the two networks. To address the problematic optimization issue, we propose new joint optimization objectives and a novel Cyclic Differentiable ARchiTecture Search framework, dubbed CDARTS. Considering the structu…
Recent convolutional neural network (CNN) development continues to advance the state-of-the-art model accuracy for various applications. However, the enhanced accuracy comes at the cost of substantial memory bandwidth and storage requirements and demanding computational resources. Although in the past the quantization methods have effectively reduced the deployment cost for edge devices, it suffers from significant information loss when processing the biased activations of contemporary CNNs. In this paper, we hence introduce an adaptive high-performance quantization method to resolve the issue…
The University of Sydney（悉尼大学）；Beijing University of Posts and Telecommunications（北京邮电大学）；SenseTime Research Centre（商汤科技研究中心）；Tsinghua University（清华大学）
Searching for a more compact network width recently serves as an effective way of channel pruning for the deployment of convolutional neural networks (CNNs) under hardware constraints. To fulfill the searching, a one-shot supernet is usually leveraged to efficiently evaluate the performance \wrt~different network widths. However, current methods mainly follow a \textit{unilaterally augmented} (UA) principle for the evaluation of each width, which induces the training unfairness of channels in supernet. In this paper, we introduce a new supernet called Bilaterally Coupled Network (BCNet) to add…
Much recent research has been dedicated to improving the efficiency of training and inference for image classification. This effort has commonly focused on explicitly improving theoretical efficiency, often measured as ImageNet validation accuracy per FLOP. These theoretical savings have, however, proven challenging to achieve in practice, particularly on high-performance training accelerators. In this work, we focus on improving the practical efficiency of the state-of-the-art EfficientNet models on a new class of accelerator, the Graphcore IPU. We do this by extending this family of models i…
Massachusetts Institute of Technology（麻省理工学院）；MIT-IBM Watson AI Lab（麻省理工学院-IBM沃森人工智能实验室）
We address the challenging problem of efficient inference across many devices and resource constraints, especially on edge devices. Conventional approaches either manually design or use neural architecture search (NAS) to find a specialized neural network and train it from scratch for each case, which is computationally prohibitive (causing $CO_2$ emission as much as 5 cars' lifetime) thus unscalable. In this work, we propose to train a once-for-all (OFA) network that supports diverse architectural settings by decoupling training and search, to reduce the cost. We can quickly get a specialized…
Deep learning is extremely computationally intensive, and hardware vendors have responded by building faster accelerators in large clusters. Training deep learning models at petaFLOPS scale requires overcoming both algorithmic and systems software challenges. In this paper, we discuss three systems-related optimizations: (1) distributed batch normalization to control per-replica batch sizes, (2) input pipeline optimizations to sustain model throughput, and (3) 2-D torus all-reduce to speed up gradient summation. We combine these optimizations to train ResNet-50 on ImageNet to 76.3% accuracy in…
Many mission-critical systems are based on GPU for inference. It requires not only high recognition accuracy but also low latency in responding time. Although many studies are devoted to optimizing the structure of deep models for efficient inference, most of them do not leverage the architecture of \textbf{modern GPU} for fast inference, leading to suboptimal performance. To address this issue, we propose a general principle for designing GPU-efficient networks based on extensive empirical studies. This design principle enables us to search for GPU-efficient network structures effectively by …
Convolutional layers are one of the basic building blocks of modern deep neural networks. One fundamental assumption is that convolutional kernels should be shared for all examples in a dataset. We propose conditionally parameterized convolutions (CondConv), which learn specialized convolutional kernels for each example. Replacing normal convolutions with CondConv enables us to increase the size and capacity of a network, while maintaining efficient inference. We demonstrate that scaling networks with CondConv improves the performance and inference cost trade-off of several existing convolutio…
City University of Hong Kong（香港城市大学）；Tsinghua University（清华大学）；Microsoft Research Asia（微软亚洲研究院）；Chinese Academy of Sciences（中国科学院）
One-shot weight sharing methods have recently drawn great attention in neural architecture search due to high efficiency and competitive performance. However, weight sharing across models has an inherent deficiency, i.e., insufficient training of subnetworks in hypernetworks. To alleviate this problem, we present a simple yet effective architecture distillation method. The central idea is that subnetworks can learn collaboratively and teach each other throughout the training process, aiming to boost the convergence of individual models. We introduce the concept of prioritized path, which refer…
Current state-of-the-art Neural Architecture Search (NAS) methods neither efficiently scale to multiple hardware platforms, nor handle diverse architectural search-spaces. To remedy this, we present DONNA (Distilling Optimal Neural Network Architectures), a novel pipeline for rapid, scalable and diverse NAS, that scales to many user scenarios. DONNA consists of three phases. First, an accuracy predictor is built using blockwise knowledge distillation from a reference model. This predictor enables searching across diverse networks with varying macro-architectural parameters such as layer types …
Stony Brook University（石溪大学）；Microsoft Research Asia（微软亚洲研究院）
Despite remarkable progress achieved, most neural architecture search (NAS) methods focus on searching for one single accurate and robust architecture. To further build models with better generalization capability and performance, model ensemble is usually adopted and performs better than stand-alone models. Inspired by the merits of model ensemble, we propose to search for multiple diverse models simultaneously as an alternative way to find powerful models. Searching for ensembles is non-trivial and has two key challenges: enlarged search space and potentially more complexity for the searched…
Microsoft Research（微软研究院）；Tsinghua University（清华大学）
Neural Architecture Search (NAS) has shown promising performance in the automatic design of vision transformers (ViT) exceeding 1G FLOPs. However, designing lightweight and low-latency ViT models for diverse mobile devices remains a big challenge. In this work, we propose ElasticViT, a two-stage NAS approach that trains a high-quality ViT supernet over a very large search space that supports a wide range of mobile devices, and then searches an optimal sub-network (subnet) for direct deployment. However, prior supernet training methods that rely on uniform sampling suffer from the gradient conf…
When deployed for risk-sensitive tasks, deep neural networks must include an uncertainty estimation mechanism. Here we examine the relationship between deep architectures and their respective training regimes, with their corresponding selective prediction and uncertainty estimation performance. We consider some of the most popular estimation performance metrics previously proposed including AUROC, ECE, AURC as well as coverage for selective accuracy constraint. We present a novel and comprehensive study of selective prediction and the uncertainty estimation performance of 523 existing pretrain…
Fine-tuning from pre-trained ImageNet models has been a simple, effective, and popular approach for various computer vision tasks. The common practice of fine-tuning is to adopt a default hyperparameter setting with a fixed pre-trained model, while both of them are not optimized for specific tasks and time constraints. Moreover, in cloud computing or GPU clusters where the tasks arrive sequentially in a stream, faster online fine-tuning is a more desired and realistic strategy for saving money, energy consumption, and CO2 emission. In this paper, we propose a joint Neural Architecture Search a…
Do Newer Lightweight CNNs Perform Better Under Resource Constraints? A Controlled Multigenerational Study of Architecture, Initialization, Training Budget, and Efficiency
Batch normalization is a key component of most image classification models, but it has many undesirable properties stemming from its dependence on the batch size and interactions between examples. Although recent work has succeeded in training deep ResNets without normalization layers, these models do not match the test accuracies of the best batch-normalized networks, and are often unstable for large learning rates or strong data augmentations. In this work, we develop an adaptive gradient clipping technique which overcomes these instabilities, and design a significantly improved class of Nor…
Large-batch training is an efficient approach for current distributed deep learning systems. It has enabled researchers to reduce the ImageNet/ResNet-50 training from 29 hours to around 1 minute. In this paper, we focus on studying the limit of the batch size. We think it may provide a guidance to AI supercomputer and algorithm designers. We provide detailed numerical optimization instructions for step-by-step comparison. Moreover, it is important to understand the generalization and optimization performance of huge batch training. Hoffer et al. introduced "ultra-slow diffusion" theory to larg…
It is commonly believed that networks cannot be both accurate and robust, that gaining robustness means losing accuracy. It is also generally believed that, unless making networks larger, network architectural elements would otherwise matter little in improving adversarial robustness. Here we present evidence to challenge these common beliefs by a careful study about adversarial training. Our key observation is that the widely-used ReLU activation function significantly weakens adversarial training due to its non-smooth nature. Hence we propose smooth adversarial training (SAT), in which we re…
For the past 5 years, the ILSVRC competition and the ImageNet dataset have attracted a lot of interest from the Computer Vision community, allowing for state-of-the-art accuracy to grow tremendously. This should be credited to the use of deep artificial neural network designs. As these became more complex, the storage, bandwidth, and compute requirements increased. This means that with a non-distributed approach, even when using the most high-density server available, the training process may take weeks, making it prohibitive. Furthermore, as datasets grow, the representation learning potentia…
Decentralized Parallel SGD (D-PSGD) and its asynchronous variant Asynchronous Parallel SGD (AD-PSGD) is a family of distributed learning algorithms that have been demonstrated to perform well for large-scale deep learning tasks. One drawback of (A)D-PSGD is that the spectral gap of the mixing matrix decreases when the number of learners in the system increases, which hampers convergence. In this paper, we investigate techniques to accelerate (A)D-PSGD based training by improving the spectral gap while minimizing the communication cost. We demonstrate the effectiveness of our proposed technique…
Facebook AI Research（Facebook人工智能研究院）；LORIA（洛里亚计算机科学实验室）；Univ Lyon（里昂大学）；Inria（法国国家信息与自动化研究所）；CNRS（法国国家科学研究中心）；ENS de Lyon（里昂高等师范学校）；UCB Lyon 1（里昂第一大学）；LIP UMR 5668（里昂第一大学计算机科学实验室）
We tackle the problem of producing compact models, maximizing their accuracy for a given model size. A standard solution is to train networks with Quantization Aware Training, where the weights are quantized during training and the gradients approximated with the Straight-Through Estimator. In this paper, we extend this approach to work beyond int8 fixed-point quantization with extreme compression methods where the approximations introduced by STE are severe, such as Product Quantization. Our proposal is to only quantize a different random subset of weights during each forward, allowing for un…
Tsinghua University（清华大学）；Stanford University（斯坦福大学）；East China Normal University（华东师范大学）；Fuzhou Institute of Data Technology（福州数据技术研究院）
Training multiple deep neural networks (DNNs) and averaging their outputs is a simple way to improve the predictive performance. Nevertheless, the multiplied training cost prevents this ensemble method to be practical and efficient. Several recent works attempt to save and ensemble the checkpoints of DNNs, which only requires the same computational cost as training a single network. However, these methods suffer from either marginal accuracy improvements due to the low diversity of checkpoints or high risk of divergence due to the cyclical learning rates they adopted. In this paper, we propose…
