PyTorchKR Papers

최신 논문

총 2,153편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2108.07049공식 코드 · Papers with Code

Towards Efficient and Data Agnostic Image Classification Training Pipeline for Embedded Systems

Nowadays deep learning-based methods have achieved a remarkable progress at the image classification task among a wide range of commonly used datasets (ImageNet, CIFAR, SVHN, Caltech 101, SUN397, etc.). SOTA performance on each of the mentioned datasets is obtained by careful tuning of the model architecture and training tricks according to the properties of the target data. Although this approach allows setting academic records, it is unrealistic that an average data scientist would have enough resources to build a sophisticated training pipeline for every image classification task he meets in practice. This work is focusing on reviewing the latest augmentation and regularization methods for the image classification and exploring ways to automatically choose some of the most important hyperparameters: total number of epochs, initial learning rate value and it's schedule. Having a training procedure equipped with a lightweight modern CNN architecture (like bileNetV3 or EfficientNet), sufficient level of regularization and adaptive to data learning rate schedule, we can achieve a reasonable performance on a variety of downstream image classification tasks without manual tuning of parameters to each particular task. Resulting models are computationally efficient and can be deployed to CPU using the OpenVINO toolkit. Source code is available as a part of the OpenVINO Training Extensions (https://github.com/openvinotoolkit/training_extensions).

arXiv:2108.05818공식 코드 · Papers with Code

PatrickStar: Parallel Training of Pre-trained Models via Chunk-based Memory Management

The pre-trained model (PTM) is revolutionizing Artificial Intelligence (AI) technology. However, the hardware requirement of PTM training is prohibitively high, making it a game for a small proportion of people. Therefore, we proposed PatrickStar system to lower the hardware requirements of PTMs and make them accessible to everyone. PatrickStar uses the CPU-GPU heterogeneous memory space to store the model data. Different from existing works, we organize the model data in memory chunks and dynamically distribute them in the heterogeneous memory. Guided by the runtime memory statistics collected in a warm-up iteration, chunks are orchestrated efficiently in heterogeneous memory and generate lower CPU-GPU data transmission volume and higher bandwidth utilization. Symbiosis with the Zero Redundancy Optimizer, PatrickStar scales to multiple GPUs on multiple nodes. % using data parallelism. The system can train tasks on bigger models and larger batch sizes, which cannot be accomplished by existing works. Experimental results show that PatrickStar extends model scales 2.27 and 2.5 times of DeepSpeed, and consistently exhibits significantly higher execution speed. PatricStar also successfully runs the 175B GPT3 training task on a 32 GPU cluster. Our code is publicly available at https://github.com/Tencent/PatrickStar.

arXiv:2108.05170

ProAI: An Efficient Embedded AI Hardware for Automotive Applications -- a Benchmark Study

Development in the field of Single Board Computers (SBC) have been increasing for several years. They provide a good balance between computing performance and power consumption which is usually required for mobile platforms, like application in vehicles for Advanced Driver Assistance Systems (ADAS) and Autonomous Driving (AD). However, there is an ever-increasing need of more powerful and efficient SBCs which can run power intensive Deep Neural Networks (DNNs) in real-time and can also satisfy necessary functional safety requirements such as Automotive Safety Integrity Level (ASIL). ProAI is being developed by ZF mainly to run powerful and efficient applications such as multitask DNNs and on top of that it also has the required safety certification for AD. In this work, we compare and discuss state of the art SBC on the basis of power intensive multitask DNN architecture called Multitask-CenterNet with respect to performance measures such as, FPS and power efficiency. As an automotive supercomputer, ProAI delivers an excellent combination of performance and efficiency, managing nearly twice the number of FPS per watt than a modern workstation laptop and almost four times compared to the Jetson Nano. Furthermore, it was also shown that there is still power in reserve for further and more complex tasks on the ProAI, based on the CPU and GPU utilization during the benchmark.

arXiv:2108.04315

FL-MISR: Fast Large-Scale Multi-Image Super-Resolution for Computed Tomography Based on Multi-GPU Acceleration

Multi-image super-resolution (MISR) usually outperforms single-image super-resolution (SISR) under a proper inter-image alignment by explicitly exploiting the inter-image correlation. However, the large computational demand encumbers the deployment of MISR in practice. In this work, we propose a distributed optimization framework based on data parallelism for fast large-scale MISR using multi-GPU acceleration named FL-MISR. The scaled conjugate gradient (SCG) algorithm is applied to the distributed subfunctions and the local SCG variables are communicated to synchronize the convergence rate over multi-GPU systems towards a consistent convergence. Furthermore, an inner-outer border exchange scheme is performed to obviate the border effect between neighboring GPUs. The proposed FL-MISR is applied to the computed tomography (CT) system by super-resolving the projections acquired by subpixel detector shift. The SR reconstruction is performed on the fly during the CT acquisition such that no additional computation time is introduced. FL-MISR is extensively evaluated from different aspects and experimental results demonstrate that FL-MISR effectively improves the spatial resolution of CT systems in modulation transfer function (MTF) and visual perception. Comparing to a multi-core CPU implementation, FL-MISR achieves a more than 50x speedup on an off-the-shelf 4-GPU system.

arXiv:2108.02037

The MIT Supercloud Dataset

Artificial intelligence (AI) and Machine learning (ML) workloads are an increasingly larger share of the compute workloads in traditional High-Performance Computing (HPC) centers and commercial cloud systems. This has led to changes in deployment approaches of HPC clusters and the commercial cloud, as well as a new focus on approaches to optimized resource usage, allocations and deployment of new AI frame- works, and capabilities such as Jupyter notebooks to enable rapid prototyping and deployment. With these changes, there is a need to better understand cluster/datacenter operations with the goal of developing improved scheduling policies, identifying inefficiencies in resource utilization, energy/power consumption, failure prediction, and identifying policy violations. In this paper we introduce the MIT Supercloud Dataset which aims to foster innovative AI/ML approaches to the analysis of large scale HPC and datacenter/cloud operations. We provide detailed monitoring logs from the MIT Supercloud system, which include CPU and GPU usage by jobs, memory usage, file system logs, and physical monitoring data. This paper discusses the details of the dataset, collection methodology, data availability, and discusses potential challenge problems being developed using this data. Datasets and future challenge announcements will be available via https://dcc.mit.edu.

arXiv:2108.00644공식 코드 · Papers with Code

Jointly Optimizing Query Encoder and Product Quantization to Improve Retrieval Performance

Recently, Information Retrieval community has witnessed fast-paced advances in Dense Retrieval (DR), which performs first-stage retrieval with embedding-based search. Despite the impressive ranking performance, previous studies usually adopt brute-force search to acquire candidates, which is prohibitive in practical Web search scenarios due to its tremendous memory usage and time cost. To overcome these problems, vector compression methods have been adopted in many practical embedding-based retrieval applications. One of the most popular methods is Product Quantization (PQ). However, although existing vector compression methods including PQ can help improve the efficiency of DR, they incur severely decayed retrieval performance due to the separation between encoding and compression. To tackle this problem, we present JPQ, which stands for Joint optimization of query encoding and Product Quantization. It trains the query encoder and PQ index jointly in an end-to-end manner based on three optimization strategies, namely ranking-oriented loss, PQ centroid optimization, and end-to-end negative sampling. We evaluate JPQ on two publicly available retrieval benchmarks. Experimental results show that JPQ significantly outperforms popular vector compression methods. Compared with previous DR models that use brute-force search, JPQ almost matches the best retrieval performance with 30x compression on index size. The compressed index further brings 10x speedup on CPU and 2x speedup on GPU in query latency.

arXiv:2107.14688

High-Resolution Depth Maps Based on TOF-Stereo Fusion

The combination of range sensors with color cameras can be very useful for robot navigation, semantic perception, manipulation, and telepresence. Several methods of combining range- and color-data have been investigated and successfully used in various robotic applications. Most of these systems suffer from the problems of noise in the range-data and resolution mismatch between the range sensor and the color cameras, since the resolution of current range sensors is much less than the resolution of color cameras. High-resolution depth maps can be obtained using stereo matching, but this often fails to construct accurate depth maps of weakly/repetitively textured scenes, or if the scene exhibits complex self-occlusions. Range sensors provide coarse depth information regardless of presence/absence of texture. The use of a calibrated system, composed of a time-of-flight (TOF) camera and of a stereoscopic camera pair, allows data fusion thus overcoming the weaknesses of both individual sensors. We propose a novel TOF-stereo fusion method based on an efficient seed-growing algorithm which uses the TOF data projected onto the stereo image pair as an initial set of correspondences. These initial "seeds" are then propagated based on a Bayesian model which combines an image similarity score with rough depth priors computed from the low-resolution range data. The overall result is a dense and accurate depth map at the resolution of the color cameras at hand. We show that the proposed algorithm outperforms 2D image-based stereo algorithms and that the results are of higher resolution than off-the-shelf color-range sensors, e.g., Kinect. Moreover, the algorithm potentially exhibits real-time performance on a single CPU.

arXiv:0508.31905

Optimus: An Efficient Dynamic Resource Scheduler for Deep Learning Clusters

Deep learning workloads are common in today’s production clusters due to the proliferation of deep learning-driven AI services (e.g., speech recognition, machine translation). A deep learning training job is resource-intensive and time-consuming. Efficient resource scheduling is the key to the maximal performance of a deep learning cluster. Existing cluster schedulers are largely not tailored to deep learning jobs, and typically specifying a fixed amount of resources for each job, prohibiting high resource efficiency and job performance. This paper proposes Optimus, a customized job scheduler for deep learning clusters, which minimizes job training time based on online resource-performance models. Optimus uses online fitting to predict model convergence during training and sets up performance models to accurately estimate training speed as a function of allocated resources in each job. Based on the models, a simple yet effective method is designed and used for dynamically allocating resources and placing deep learning tasks to minimize job completion time. We implement Optimus on top of Kubernetes, a cluster manager for container orchestration, and experiment on a deep learning cluster with 7 CPU servers and 6 GPU servers, running 9 training jobs using the MXNet framework. Results show that Optimus outperforms representative cluster schedulers by about 139% and 63% in terms of job completion time and makespan, respectively

arXiv:2107.12421

Parallel Surrogate-assisted Optimization Using Mesh Adaptive Direct Search

We consider computationally expensive blackbox optimization problems and present a method that employs surrogate models and concurrent computing at the search step of the mesh adaptive direct search (MADS) algorithm. Specifically, we solve a surrogate optimization problem using locally weighted scatterplot smoothing (LOWESS) models to find promising candidate points to be evaluated by the blackboxes. We consider several methods for selecting promising points from a large number of points. We conduct numerical experiments to assess the performance of the modified MADS algorithm with respect to available CPU resources by means of five engineering design problems.

arXiv:2107.11167

Dynamic detection of mobile malware using smartphone data and machine learning

Mobile malware are malicious programs that target mobile devices. They are an increasing problem, as seen in the rise of detected mobile malware samples per year. The number of active smartphone users is expected to grow, stressing the importance of research on the detection of mobile malware. Detection methods for mobile malware exist but are still limited. In this paper, we provide an overview of the performance of machine learning (ML) techniques to detect malware on Android, without using privileged access. The ML-classifiers use device information such as the CPU usage, battery usage, and memory usage for the detection of 10 subtypes of Mobile Trojans on the Android Operating System (OS). We use a real-life dataset containing device and malware data from 47 users for a year (2016). We examine which features, i.e. aspects, of a device, are most important to monitor to detect (subtypes of) Mobile Trojans. The focus of this paper is on dynamic hardware features. Using these dynamic features we apply state-of-the-art machine learning classifiers: Random Forest, K-Nearest Neighbour, and AdaBoost. We show classification results on different feature sets, making a distinction between global device features, and specific app features. None of the measured feature sets require privileged access. Our results show that the Random Forest classifier performs best as a general malware classifier: across 10 subtypes of Mobile Trojans, it achieves an F1 score of 0.73 with a False Positive Rate (FPR) of 0.009 and a False Negative Rate (FNR) of 0.380. The Random Forest, K-Nearest Neighbours, and AdaBoost classifiers achieve F1 scores above 0.72, an FPR below 0.02 and, an FNR below 0.33, when trained separately to detect each subtype of Mobile Trojans.

arXiv:2107.10624

LANA: Latency Aware Network Acceleration

We introduce latency-aware network acceleration (LANA) - an approach that builds on neural architecture search techniques and teacher-student distillation to accelerate neural networks. LANA consists of two phases: in the first phase, it trains many alternative operations for every layer of the teacher network using layer-wise feature map distillation. In the second phase, it solves the combinatorial selection of efficient operations using a novel constrained integer linear optimization (ILP) approach. ILP brings unique properties as it (i) performs NAS within a few seconds to minutes, (ii) easily satisfies budget constraints, (iii) works on the layer-granularity, (iv) supports a huge search space $O(10^{100})$, surpassing prior search approaches in efficacy and efficiency. In extensive experiments, we show that LANA yields efficient and accurate models constrained by a target latency budget, while being significantly faster than other techniques. We analyze three popular network architectures: EfficientNetV1, EfficientNetV2 and ResNeST, and achieve accuracy improvement for all models (up to $3.0\%$) when compressing larger models to the latency level of smaller models. LANA achieves significant speed-ups (up to $5\times$) with minor to no accuracy drop on GPU and CPU. The code will be shared soon.

arXiv:2107.09521

On the Design of Complex EM Devices and Systems through the System-by-Design Paradigm -- A Framework for Dealing with the Computational Complexity

The System-by-Design (SbD) is an emerging engineering framework for the optimization-driven design of complex electromagnetic (EM) devices and systems. More specifically, the computational complexity of the design problem at hand is addressed by means of a suitable selection and integration of functional blocks comprising problem-dependent and computationally-efficient modeling and analysis tools as well as reliable prediction and optimization strategies. Thanks to the suitable re-formulation of the problem at hand as an optimization one, the profitable minimum-size coding of the degrees-of-freedom (DoFs), the "smart" replacement of expensive full-wave (FW) simulators with proper surrogate models (SMs), which yield fast yet accurate predictions starting from minimum size/reduced CPU-costs training sets, a favorable "environment" for an optimal exploitation of the features of global optimization tools in sampling wide/complex/nonlinear solution spaces is built. This research summary is then aimed at (i) providing a comprehensive description of the SbD framework and of its pillar concepts and strategies, (ii) giving useful guidelines for its successful customization and application to different EM design problems characterized by different levels of computational complexity, (iii) envisaging future trends and advances in this fascinating and high-interest (because of its relevant and topical industrial and commercial implications) topic. Representative benchmarks concerned with the synthesis of complex EM systems are presented to highlight advantages and potentialities as well as current limitations of the SbD paradigm.

arXiv:2107.09333

StreamBlocks: A compiler for heterogeneous dataflow computing (technical report)

To increase performance and efficiency, systems use FPGAs as reconfigurable accelerators. A key challenge in designing these systems is partitioning computation between processors and an FPGA. An appropriate division of labor may be difficult to predict in advance and require experiments and measurements. When an investigation requires rewriting part of the system in a new language or with a new programming model, its high cost can retard the study of different configurations. A single-language system with an appropriate programming model and compiler that targets both platforms simplifies this exploration to a simple recompile with new compiler directives. This work introduces StreamBlocks, an open-source compiler and runtime that uses the CAL dataflow programming language to partition computations across heterogeneous (CPU/accelerator) platforms. Because of the dataflow model's semantics and the CAL language, StreamBlocks can exploit both thread parallelism in multi-core CPUs and the inherent parallelism of FPGAs. StreamBlocks supports exploring the design space with a profile-guided tool that helps identify the best hardware-software partitions.

arXiv:2107.08199

Dynamic Transformer for Efficient Machine Translation on Embedded Devices

The Transformer architecture is widely used for machine translation tasks. However, its resource-intensive nature makes it challenging to implement on constrained embedded devices, particularly where available hardware resources can vary at run-time. We propose a dynamic machine translation model that scales the Transformer architecture based on the available resources at any particular time. The proposed approach, 'Dynamic-HAT', uses a HAT SuperTransformer as the backbone to search for SubTransformers with different accuracy-latency trade-offs at design time. The optimal SubTransformers are sampled from the SuperTransformer at run-time, depending on latency constraints. The Dynamic-HAT is tested on the Jetson Nano and the approach uses inherited SubTransformers sampled directly from the SuperTransformer with a switching time of <1s. Using inherited SubTransformers results in a BLEU score loss of <1.5% because the SubTransformer configuration is not retrained from scratch after sampling. However, to recover this loss in performance, the dimensions of the design space can be reduced to tailor it to a family of target hardware. The new reduced design space results in a BLEU score increase of approximately 1% for sub-optimal models from the original design space, with a wide range for performance scaling between 0.356s - 1.526s for the GPU and 2.9s - 7.31s for the CPU.

arXiv:2107.06433

A New Parallel Algorithm for Sinkhorn Word-Movers Distance and Its Performance on PIUMA and Xeon CPU

The Word Movers Distance (WMD) measures the semantic dissimilarity between two text documents by computing the cost of optimally moving all words of a source/query document to the most similar words of a target document. Computing WMD between two documents is costly because it requires solving an $O(V^3log(V))$ optimization problem where $V$ is the number of unique words in the document. Fortunately, WMD can be framed as an Earth Mover's Distance (EMD) for which the algorithmic complexity can be reduced to $O(V^2)$ by adding an entropy penalty to the optimization problem and solving it using the Sinkhorn-Knopp algorithm. Additionally, the computation can be made highly parallel by adopting a batching approach, i.e., computing the WMD of a single query document against multiple target documents at once. Sinkhorn WMD is a key kernel used in many ML/NLP applications. and usually gets implemented in Python. However, a straightforward Python implementation may leave significant performance on the table even though it may internally call optimized C++ BLAS routines. We present a new sparse {P}arallel {A}lgorithm for {S}inkhorn-Knopp {W}ord-movers {D}istance to compute the semantic distance of one document to many other documents by adopting the $O(V^2)$ EMD algorithm. We algorithmically transform $O(V^2)$ dense compute-heavy EMD version into an equivalent sparse one using new fused SDDMM-SpMM (sparse selection of dense-dense matrix-, sparse-dense matrix-multiplication) kernels. We implemented and optimized this algorithm for two very different architectures -- the new Intel Programmable Integrated Unified Memory Architecture (PIUMA) and Intel Xeon CPUs. We show that we were able to reach close to peak performance on both platforms.

arXiv:2107.06397공식 코드 · Papers with Code

SurgeonAssist-Net: Towards Context-Aware Head-Mounted Display-Based Augmented Reality for Surgical Guidance

We present SurgeonAssist-Net: a lightweight framework making action-and-workflow-driven virtual assistance, for a set of predefined surgical tasks, accessible to commercially available optical see-through head-mounted displays (OST-HMDs). On a widely used benchmark dataset for laparoscopic surgical workflow, our implementation competes with state-of-the-art approaches in prediction accuracy for automated task recognition, and yet requires 7.4x fewer parameters, 10.2x fewer floating point operations per second (FLOPS), is 7.0x faster for inference on a CPU, and is capable of near real-time performance on the Microsoft HoloLens 2 OST-HMD. To achieve this, we make use of an efficient convolutional neural network (CNN) backbone to extract discriminative features from image data, and a low-parameter recurrent neural network (RNN) architecture to learn long-term temporal dependencies. To demonstrate the feasibility of our approach for inference on the HoloLens 2 we created a sample dataset that included video of several surgical tasks recorded from a user-centric point-of-view. After training, we deployed our model and cataloged its performance in an online simulated surgical scenario for the prediction of the current surgical task. The utility of our approach is explored in the discussion of several relevant clinical use-cases. Our code is publicly available at https://github.com/doughtmw/surgeon-assist-net.

arXiv:2107.05584

Implementing a LoRa Software-Defined Radio on a General-Purpose ULP Microcontroller

Emerging Internet-of-Things sensing applications rely on ultra low-power (ULP) microcontroller units (MCUs) that wirelessly transmit data to the cloud. Typical MCUs nowadays consist of generic blocks, except for the protocol-specific radios implemented in hardware. Hardware radios however slow down the evolution of wireless protocols due to retrocompatiblity concerns. In this work, we explore a software-defined radio architecture by demonstrating a LoRa transceiver running on custom ULP MCU codenamed SleepRider with an ARM Cortex-M4 CPU. In SleepRider MCU, we offload the generic baseband operations (e.g., low-pass filtering) to a reconfigurable digital front-end block and use the Cortex-M4 CPU to perform the protocol-specific computations. Our software implementation of the LoRa physical layer only uses the native SIMD instructions of the Cortex-M4 to achieve real-time transmission and reception of LoRa packets. SleepRider MCU has been fabricated in a 28nm FDSOI CMOS technology and is used in a testbed to experimentally validate the software implementation. Experimental results show that the proposed software-defined radio requires only a CPU frequency of 20 MHz to correctly receive a LoRa packet, with an ultra-low power consumption of 0.42 mW on average.

arXiv:2107.05050공식 코드 · Papers with Code

Neural Waveshaping Synthesis

We present the Neural Waveshaping Unit (NEWT): a novel, lightweight, fully causal approach to neural audio synthesis which operates directly in the waveform domain, with an accompanying optimisation (FastNEWT) for efficient CPU inference. The NEWT uses time-distributed multilayer perceptrons with periodic activations to implicitly learn nonlinear transfer functions that encode the characteristics of a target timbre. Once trained, a NEWT can produce complex timbral evolutions by simple affine transformations of its input and output signals. We paired the NEWT with a differentiable noise synthesiser and reverb and found it capable of generating realistic musical instrument performances with only 260k total model parameters, conditioned on F0 and loudness features. We compared our method to state-of-the-art benchmarks with a multi-stimulus listening test and the Fr\'echet Audio Distance and found it performed competitively across the tested timbral domains. Our method significantly outperformed the benchmarks in terms of generation speed, and achieved real-time performance on a consumer CPU, both with and without FastNEWT, suggesting it is a viable basis for future creative sound design tools.

arXiv:2107.04767

Anomaly Detection in Residential Video Surveillance on Edge Devices in IoT Framework

Intelligent resident surveillance is one of the most essential smart community services. The increasing demand for security needs surveillance systems to be able to detect anomalies in surveillance scenes. Employing high-capacity computational devices for intelligent surveillance in residential societies is costly and not feasible. Therefore, we propose anomaly detection for intelligent surveillance using CPU-only edge devices. A modular framework to capture object-level inferences and tracking is developed. To cope with partial occlusions, posture deformations, and complex scenes, we employed feature encoding and trajectory association governed by two metrices complementing to each other. The elements of an anomaly detection framework are optimized to run on CPU-only edge devices with sufficient frames per second (FPS). The experimental results indicate the proposed method is feasible and achieves satisfactory results in real-life scenarios.

arXiv:2107.03312

SoundStream: An End-to-End Neural Audio Codec

We present SoundStream, a novel neural audio codec that can efficiently compress speech, music and general audio at bitrates normally targeted by speech-tailored codecs. SoundStream relies on a model architecture composed by a fully convolutional encoder/decoder network and a residual vector quantizer, which are trained jointly end-to-end. Training leverages recent advances in text-to-speech and speech enhancement, which combine adversarial and reconstruction losses to allow the generation of high-quality audio content from quantized embeddings. By training with structured dropout applied to quantizer layers, a single model can operate across variable bitrates from 3kbps to 18kbps, with a negligible quality loss when compared with models trained at fixed bitrates. In addition, the model is amenable to a low latency implementation, which supports streamable inference and runs in real time on a smartphone CPU. In subjective evaluations using audio at 24kHz sampling rate, SoundStream at 3kbps outperforms Opus at 12kbps and approaches EVS at 9.6kbps. Moreover, we are able to perform joint compression and enhancement either at the encoder or at the decoder side with no additional latency, which we demonstrate through background noise suppression for speech.