PyTorchKR Papers

최신 논문

총 5,575편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2010.13887공식 코드 · Papers with Code

LightSeq: A High Performance Inference Library for Transformers

Transformer, BERT and their variants have achieved great success in natural language processing. Since Transformer models are huge in size, serving these models is a challenge for real industrial applications. In this paper, we propose LightSeq, a highly efficient inference library for models in the Transformer family. LightSeq includes a series of GPU optimization techniques to to streamline the computation of neural layers and to reduce memory footprint. LightSeq can easily import models trained using PyTorch and Tensorflow. Experimental results on machine translation benchmarks show that LightSeq achieves up to 14x speedup compared with TensorFlow and 1.4x compared with FasterTransformer, a concurrent CUDA implementation. The code is available at https://github.com/bytedance/lightseq.

arXiv:2010.12635

Not Half Bad: Exploring Half-Precision in Graph Convolutional Neural Networks

With the growing significance of graphs as an effective representation of data in numerous applications, efficient graph analysis using modern machine learning is receiving a growing level of attention. Deep learning approaches often operate over the entire adjacency matrix -- as the input and intermediate network layers are all designed in proportion to the size of the adjacency matrix -- leading to intensive computation and large memory requirements as the graph size increases. It is therefore desirable to identify efficient measures to reduce both run-time and memory requirements allowing for the analysis of the largest graphs possible. The use of reduced precision operations within the forward and backward passes of a deep neural network along with novel specialised hardware in modern GPUs can offer promising avenues towards efficiency. In this paper, we provide an in-depth exploration of the use of reduced-precision operations, easily integrable into the highly popular PyTorch framework, and an analysis of the effects of Tensor Cores on graph convolutional neural networks. We perform an extensive experimental evaluation of three GPU architectures and two widely-used graph analysis tasks (vertex classification and link prediction) using well-known benchmark and synthetically generated datasets. Thus allowing us to make important observations on the effects of reduced-precision operations and Tensor Cores on computational and memory usage of graph convolutional neural networks -- often neglected in the literature.

arXiv:2010.13100

Tensor Casting: Co-Designing Algorithm-Architecture for Personalized Recommendation Training

Personalized recommendations are one of the most widely deployed machine learning (ML) workload serviced from cloud datacenters. As such, architectural solutions for high-performance recommendation inference have recently been the target of several prior literatures. Unfortunately, little have been explored and understood regarding the training side of this emerging ML workload. In this paper, we first perform a detailed workload characterization study on training recommendations, root-causing sparse embedding layer training as one of the most significant performance bottlenecks. We then propose our algorithm-architecture co-design called Tensor Casting, which enables the development of a generic accelerator architecture for tensor gather-scatter that encompasses all the key primitives of training embedding layers. When prototyped on a real CPU-GPU system, Tensor Casting provides 1.9-21x improvements in training throughput compared to state-of-the-art approaches.

arXiv:2010.13017공식 코드 · Papers with Code

APB2FaceV2: Real-Time Audio-Guided Multi-Face Reenactment

Audio-guided face reenactment aims to generate a photorealistic face that has matched facial expression with the input audio. However, current methods can only reenact a special person once the model is trained or need extra operations such as 3D rendering and image post-fusion on the premise of generating vivid faces. To solve the above challenge, we propose a novel \emph{R}eal-time \emph{A}udio-guided \emph{M}ulti-face reenactment approach named \emph{APB2FaceV2}, which can reenact different target faces among multiple persons with corresponding reference face and drive audio signal as inputs. Enabling the model to be trained end-to-end and have a faster speed, we design a novel module named Adaptive Convolution (AdaConv) to infuse audio information into the network, as well as adopt a lightweight network as our backbone so that the network can run in real time on CPU and GPU. Comparison experiments prove the superiority of our approach than existing state-of-the-art methods, and further experiments demonstrate that our method is efficient and flexible for practical applications https://github.com/zhangzjn/APB2FaceV2

arXiv:2010.13412

Flexible Piecewise Curves Estimation for Photo Enhancement

This paper presents a new method, called FlexiCurve, for photo enhancement. Unlike most existing methods that perform image-to-image mapping, which requires expensive pixel-wise reconstruction, FlexiCurve takes an input image and estimates global curves to adjust the image. The adjustment curves are specially designed for performing piecewise mapping, taking nonlinear adjustment and differentiability into account. To cope with challenging and diverse illumination properties in real-world images, FlexiCurve is formulated as a multi-task framework to produce diverse estimations and the associated confidence maps. These estimations are adaptively fused to improve local enhancements of different regions. Thanks to the image-to-curve formulation, for an image with a size of 512*512*3, FlexiCurve only needs a lightweight network (150K trainable parameters) and it has a fast inference speed (83FPS on a single NVIDIA 2080Ti GPU). The proposed method improves efficiency without compromising the enhancement quality and losing details in the original image. The method is also appealing as it is not limited to paired training data, thus it can flexibly learn rich enhancement styles from unpaired data. Extensive experiments demonstrate that our method achieves state-of-the-art performance on photo enhancement quantitively and qualitatively.

arXiv:2010.13294

Global Image Segmentation Process using Machine Learning algorithm & Convolution Neural Network method for Self- Driving Vehicles

In autonomous Vehicles technology Image segmentation was a major problem in visual perception. This image segmentation process is mainly used in medical applications. Here we adopted an image segmentation process to visual perception tasks for predicting the agents on the surrounding environment, identifying the road boundaries and tracking the line markings. Main objective of the paper is to divide the input images using the image segmentation process and Convolution Neural Network method for efficient results of visual perception. For Sampling assume a local city data-set samples and validation process done in Jupyter Notebook using Python language. We proposed this image segmentation method planning to standard and further the development of state-of-the art methods for visual inspection system understanding. The experimental results achieves 73% mean IOU. Our method also achieves 90 FPS inference speed and using a NVDIA GeForce GTX 1050 GPU.

arXiv:2010.13972공식 코드 · Papers with Code

GPUTreeShap: Massively Parallel Exact Calculation of SHAP Scores for Tree Ensembles

SHAP (SHapley Additive exPlanation) values provide a game theoretic interpretation of the predictions of machine learning models based on Shapley values. While exact calculation of SHAP values is computationally intractable in general, a recursive polynomial-time algorithm called TreeShap is available for decision tree models. However, despite its polynomial time complexity, TreeShap can become a significant bottleneck in practical machine learning pipelines when applied to large decision tree ensembles. Unfortunately, the complicated TreeShap algorithm is difficult to map to hardware accelerators such as GPUs. In this work, we present GPUTreeShap, a reformulated TreeShap algorithm suitable for massively parallel computation on graphics processing units. Our approach first preprocesses each decision tree to isolate variable sized sub-problems from the original recursive algorithm, then solves a bin packing problem, and finally maps sub-problems to single-instruction, multiple-thread (SIMT) tasks for parallel execution with specialised hardware instructions. With a single NVIDIA Tesla V100-32 GPU, we achieve speedups of up to 19x for SHAP values, and speedups of up to 340x for SHAP interaction values, over a state-of-the-art multi-core CPU implementation executed on two 20-core Xeon E5-2698 v4 2.2 GHz CPUs. We also experiment with multi-GPU computing using eight V100 GPUs, demonstrating throughput of 1.2M rows per second -- equivalent CPU-based performance is estimated to require 6850 CPU cores.

arXiv:2010.13821

Wavelet Flow: Fast Training of High Resolution Normalizing Flows

Normalizing flows are a class of probabilistic generative models which allow for both fast density computation and efficient sampling and are effective at modelling complex distributions like images. A drawback among current methods is their significant training cost, sometimes requiring months of GPU training time to achieve state-of-the-art results. This paper introduces Wavelet Flow, a multi-scale, normalizing flow architecture based on wavelets. A Wavelet Flow has an explicit representation of signal scale that inherently includes models of lower resolution signals and conditional generation of higher resolution signals, i.e., super resolution. A major advantage of Wavelet Flow is the ability to construct generative models for high resolution data (e.g., 1024 x 1024 images) that are impractical with previous models. Furthermore, Wavelet Flow is competitive with previous normalizing flows in terms of bits per dimension on standard (low resolution) benchmarks while being up to 15x faster to train.

arXiv:2010.14109

Out-of-core Training for Extremely Large-Scale Neural Networks With Adaptive Window-Based Scheduling

While large neural networks demonstrate higher performance in various tasks, training large networks is difficult due to limitations on GPU memory size. We propose a novel out-of-core algorithm that enables faster training of extremely large-scale neural networks with sizes larger than allotted GPU memory. Under a given memory budget constraint, our scheduling algorithm locally adapts the timing of memory transfers according to memory usage of each function, which improves overlap between computation and memory transfers. Additionally, we apply virtual addressing technique, commonly performed in OS, to training of neural networks with out-of-core execution, which drastically reduces the amount of memory fragmentation caused by frequent memory transfers. With our proposed algorithm, we successfully train ResNet-50 with 1440 batch-size with keeping training speed at 55%, which is 7.5x larger than the upper bound of physical memory. It also outperforms a previous state-of-the-art substantially, i.e. it trains a 1.55x larger network than state-of-the-art with faster execution. Moreover, we experimentally show that our approach is also scalable for various types of networks.

논문 ID: document/7995966

Efficient ConvNet for Real-time Semantic Segmentation

Semantic segmentation is a task that covers most of the perception needs of intelligent vehicles in an unified way. ConvNets excel at this task, as they can be trained end-to-end to accurately classify multiple object categories in an image at the pixel level. However, current approaches normally involve complex architectures that are expensive in terms of computational resources and are not feasible for ITS applications. In this paper, we propose a deep architecture that is able to run in real-time while providing accurate semantic segmentation. The core of our ConvNet is a novel layer that uses residual connections and factorized convolutions in order to remain highly efficient while still retaining remarkable performance. Our network is able to run at 83 FPS in a single Titan X, and at more than 7 FPS in a Jetson TX1 (embedded GPU). A comprehensive set of experiments demonstrates that our system, trained from scratch on the challenging Cityscapes dataset, achieves a classification performance that is among the state of the art, while being orders of magnitude faster to compute than other architectures that achieve top precision. This makes our model an ideal approach for scene understanding in intelligent vehicles applications.

arXiv:2010.13382공식 코드 · Papers with Code

FastFormers: Highly Efficient Transformer Models for Natural Language Understanding

Transformer-based models are the state-of-the-art for Natural Language Understanding (NLU) applications. Models are getting bigger and better on various tasks. However, Transformer models remain computationally challenging since they are not efficient at inference-time compared to traditional approaches. In this paper, we present FastFormers, a set of recipes to achieve efficient inference-time performance for Transformer-based models on various NLU tasks. We show how carefully utilizing knowledge distillation, structured pruning and numerical optimization can lead to drastic improvements on inference efficiency. We provide effective recipes that can guide practitioners to choose the best settings for various NLU tasks and pretrained models. Applying the proposed recipes to the SuperGLUE benchmark, we achieve from 9.8x up to 233.9x speed-up compared to out-of-the-box models on CPU. On GPU, we also achieve up to 12.4x speed-up with the presented methods. We show that FastFormers can drastically reduce cost of serving 100 million requests from 4,223 USD to just 18 USD on an Azure F16s_v2 instance. This translates to a sustainable runtime by reducing energy consumption 6.9x - 125.8x according to the metrics used in the SustaiNLP 2020 shared task.

arXiv:2010.12138공식 코드 · Papers with Code

Rethinking the competition between detection and ReID in Multi-Object Tracking

Due to balanced accuracy and speed, one-shot models which jointly learn detection and identification embeddings, have drawn great attention in multi-object tracking (MOT). However, the inherent differences and relations between detection and re-identification (ReID) are unconsciously overlooked because of treating them as two isolated tasks in the one-shot tracking paradigm. This leads to inferior performance compared with existing two-stage methods. In this paper, we first dissect the reasoning process for these two tasks, which reveals that the competition between them inevitably would destroy task-dependent representations learning. To tackle this problem, we propose a novel reciprocal network (REN) with a self-relation and cross-relation design so that to impel each branch to better learn task-dependent representations. The proposed model aims to alleviate the deleterious tasks competition, meanwhile improve the cooperation between detection and ReID. Furthermore, we introduce a scale-aware attention network (SAAN) that prevents semantic level misalignment to improve the association capability of ID embeddings. By integrating the two delicately designed networks into a one-shot online MOT system, we construct a strong MOT tracker, namely CSTrack. Our tracker achieves the state-of-the-art performance on MOT16, MOT17 and MOT20 datasets, without other bells and whistles. Moreover, CSTrack is efficient and runs at 16.4 FPS on a single modern GPU, and its lightweight version even runs at 34.6 FPS. The complete code has been released at https://github.com/JudasDie/SOTS.

arXiv:2010.12496

ResNet or DenseNet? Introducing Dense Shortcuts to ResNet

ResNet or DenseNet? Nowadays, most deep learning based approaches are implemented with seminal backbone networks, among them the two arguably most famous ones are ResNet and DenseNet. Despite their competitive performance and overwhelming popularity, inherent drawbacks exist for both of them. For ResNet, the identity shortcut that stabilizes training also limits its representation capacity, while DenseNet has a higher capacity with multi-layer feature concatenation. However, the dense concatenation causes a new problem of requiring high GPU memory and more training time. Partially due to this, it is not a trivial choice between ResNet and DenseNet. This paper provides a unified perspective of dense summation to analyze them, which facilitates a better understanding of their core difference. We further propose dense weighted normalized shortcuts as a solution to the dilemma between them. Our proposed dense shortcut inherits the design philosophy of simple design in ResNet and DenseNet. On several benchmark datasets, the experimental results show that the proposed DSNet achieves significantly better results than ResNet, and achieves comparable performance as DenseNet but requiring fewer computation resources.

arXiv:2010.11305

Mixed-Precision Embedding Using a Cache

In recommendation systems, practitioners observed that increase in the number of embedding tables and their sizes often leads to significant improvement in model performances. Given this and the business importance of these models to major internet companies, embedding tables for personalization tasks have grown to terabyte scale and continue to grow at a significant rate. Meanwhile, these large-scale models are often trained with GPUs where high-performance memory is a scarce resource, thus motivating numerous work on embedding table compression during training. We propose a novel change to embedding tables using a cache memory architecture, where the majority of rows in an embedding is trained in low precision, and the most frequently or recently accessed rows cached and trained in full precision. The proposed architectural change works in conjunction with standard precision reduction and computer arithmetic techniques such as quantization and stochastic rounding. For an open source deep learning recommendation model (DLRM) running with Criteo-Kaggle dataset, we achieve 3x memory reduction with INT8 precision embedding tables and full-precision cache whose size are 5% of the embedding tables, while maintaining accuracy. For an industrial scale model and dataset, we achieve even higher >7x memory reduction with INT4 precision and cache size 1% of embedding tables, while maintaining accuracy, and 16% end-to-end training speedup by reducing GPU-to-host data transfers.

arXiv:2010.11297

Performance Prediction for Convolutional Neural Networks in Edge Devices

Running Convolutional Neural Network (CNN) based applications on edge devices near the source of data can meet the latency and privacy challenges. However due to their reduced computing resources and their energy constraints, these edge devices can hardly satisfy CNN needs in processing and data storage. For these platforms, choosing the CNN with the best trade-off between accuracy and execution time while respecting Hardware constraints is crucial. In this paper, we present and compare five (5) of the widely used Machine Learning based methods for execution time prediction of CNNs on two (2) edge GPU platforms. For these 5 methods, we also explore the time needed for their training and tuning their corresponding hyperparameters. Finally, we compare times to run the prediction models on different platforms. The utilization of these methods will highly facilitate design space exploration by providing quickly the best CNN on a target edge GPU. Experimental results show that eXtreme Gradient Boosting (XGBoost) provides a less than 14.73% average prediction error even for unexplored and unseen CNN models' architectures. Random Forest (RF) depicts comparable accuracy but needs more effort and time to be trained. The other 3 approaches (OLS, MLP and SVR) are less accurate for CNN performances estimation.

arXiv:2010.10903공식 코드 · Papers with Code

Visual Navigation in Real-World Indoor Environments Using End-to-End Deep Reinforcement Learning

Visual navigation is essential for many applications in robotics, from manipulation, through mobile robotics to automated driving. Deep reinforcement learning (DRL) provides an elegant map-free approach integrating image processing, localization, and planning in one module, which can be trained and therefore optimized for a given environment. However, to date, DRL-based visual navigation was validated exclusively in simulation, where the simulator provides information that is not available in the real world, e.g., the robot's position or image segmentation masks. This precludes the use of the learned policy on a real robot. Therefore, we propose a novel approach that enables a direct deployment of the trained policy on real robots. We have designed visual auxiliary tasks, a tailored reward scheme, and a new powerful simulator to facilitate domain randomization. The policy is fine-tuned on images collected from real-world environments. We have evaluated the method on a mobile robot in a real office environment. The training took ~30 hours on a single GPU. In 30 navigation experiments, the robot reached a 0.3-meter neighborhood of the goal in more than 86.7% of cases. This result makes the proposed method directly applicable to tasks like mobile manipulation.

arXiv:2010.11113공식 코드 · Papers with Code

One Model to Reconstruct Them All: A Novel Way to Use the Stochastic Noise in StyleGAN

Generative Adversarial Networks (GANs) have achieved state-of-the-art performance for several image generation and manipulation tasks. Different works have improved the limited understanding of the latent space of GANs by embedding images into specific GAN architectures to reconstruct the original images. We present a novel StyleGAN-based autoencoder architecture, which can reconstruct images with very high quality across several data domains. We demonstrate a previously unknown grade of generalizablility by training the encoder and decoder independently and on different datasets. Furthermore, we provide new insights about the significance and capabilities of noise inputs of the well-known StyleGAN architecture. Our proposed architecture can handle up to 40 images per second on a single GPU, which is approximately 28x faster than previous approaches. Finally, our model also shows promising results, when compared to the state-of-the-art on the image denoising task, although it was not explicitly designed for this task.

arXiv:2010.10499공식 코드 · Papers with Code

Optimal Subarchitecture Extraction For BERT

We extract an optimal subset of architectural parameters for the BERT architecture from Devlin et al. (2018) by applying recent breakthroughs in algorithms for neural architecture search. This optimal subset, which we refer to as "Bort", is demonstrably smaller, having an effective (that is, not counting the embedding layer) size of $5.5\%$ the original BERT-large architecture, and $16\%$ of the net size. Bort is also able to be pretrained in $288$ GPU hours, which is $1.2\%$ of the time required to pretrain the highest-performing BERT parametric architectural variant, RoBERTa-large (Liu et al., 2019), and about $33\%$ of that of the world-record, in GPU hours, required to train BERT-large on the same hardware. It is also $7.9$x faster on a CPU, as well as being better performing than other compressed variants of the architecture, and some of the non-compressed variants: it obtains performance improvements of between $0.3\%$ and $31\%$, absolute, with respect to BERT-large, on multiple public natural language understanding (NLU) benchmarks.

arXiv:2010.10458

Towards Scalable Distributed Training of Deep Learning on Public Cloud Clusters

Distributed training techniques have been widely deployed in large-scale deep neural networks (DNNs) 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.

arXiv:2010.08855

GOAT: GPU Outsourcing of Deep Learning Training With Asynchronous Probabilistic Integrity Verification Inside Trusted Execution Environment

Machine learning models based on Deep Neural Networks (DNNs) are increasingly deployed in a wide range of applications ranging from self-driving cars to COVID-19 treatment discovery. To support the computational power necessary to learn a DNN, cloud environments with dedicated hardware support have emerged as critical infrastructure. However, there are many integrity challenges associated with outsourcing computation. Various approaches have been developed to address these challenges, building on trusted execution environments (TEE). Yet, no existing approach scales up to support realistic integrity-preserving DNN model training for heavy workloads (deep architectures and millions of training examples) without sustaining a significant performance hit. To mitigate the time gap between pure TEE (full integrity) and pure GPU (no integrity), we combine random verification of selected computation steps with systematic adjustments of DNN hyper-parameters (e.g., a narrow gradient clipping range), hence limiting the attacker's ability to shift the model parameters significantly provided that the step is not selected for verification during its training phase. Experimental results show the new approach achieves 2X to 20X performance improvement over pure TEE based solution while guaranteeing a very high probability of integrity (e.g., 0.999) with respect to state-of-the-art DNN backdoor attacks.