اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدE-BATCH: Energy-Efficient and High-Throughput RNN Batching
51
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Recurrent Neural Network (RNN) inference exhibits low hardware utilization due to the strict data dependencies across time-steps. Batching multiple requests can increase throughput. However, RNN batching requires a large amount of padding since the batched input sequences may largely differ in length. Schemes that dynamically update the batch every few time-steps avoid padding. However, they require executing different RNN layers in a short timespan, decreasing energy efficiency. Hence, we propose E-BATCH, a low-latency and energy-efficient batching scheme tailored to RNN accelerators. It consists of a runtime system and effective hardware support. The runtime concatenates multiple sequences to create large batches, resulting in substantial energy savings. Furthermore, the accelerator notifies it when the evaluation of a sequence is done, so that a new sequence can be immediately added to a batch, thus largely reducing the amount of padding. E-BATCH dynamically controls the number of time-steps evaluated per batch to achieve the best trade-off between latency and energy efficiency for the given hardware platform. We evaluate E-BATCH on top of E-PUR and TPU. In E-PUR, E-BATCH improves throughput by 1.8x and energy-efficiency by 3.6x, whereas in TPU, it improves throughput by 2.1x and energy-efficiency by 1.6x, over the state-of-the-art.
قيم البحث
اقرأ أيضاً
Grid computing systems require innovative methods and tools to identify cybersecurity incidents and perform autonomous actions i.e. without administrator intervention. They also require methods to isolate and trace job payload activity in order to pr
otect users and find evidence of malicious behavior. We introduce an integrated approach of security monitoring via Security by Isolation with Linux Containers and Deep Learning methods for the analysis of real time data in Grid jobs running inside virtualized High-Throughput Computing infrastructure in order to detect and prevent intrusions. A dataset for malware detection in Grid computing is described. We show in addition the utilization of generative methods with Recurrent Neural Networks to improve the collected dataset. We present Arhuaco, a prototype implementation of the proposed methods. We empirically study the performance of our technique. The results show that Arhuaco outperforms other methods used in Intrusion Detection Systems for Grid Computing. The study is carried out in the ALICE Collaboration Grid, part of the Worldwide LHC Computing Grid.
Sensing systems powered by energy harvesting have traditionally been designed to tolerate long periods without energy. As the Internet of Things (IoT) evolves towards a more transient and opportunistic execution paradigm, reducing energy storage cost
s will be key for its economic and ecologic viability. However, decreasing energy storage in harvesting systems introduces reliability issues. Transducers only produce intermittent energy at low voltage and current levels, making guaranteed task completion a challenge. Existing ad hoc methods overcome this by buffering enough energy either for single tasks, incurring large data-retention overheads, or for one full application cycle, requiring a large energy buffer. We present Julienning: an automated method for optimizing the total energy cost of batteryless applications. Using a custom specification model, developers can describe transient applications as a set of atomically executed kernels with explicit data dependencies. Our optimization flow can partition data- and energy-intensive applications into multiple execution cycles with bounded energy consumption. By leveraging interkernel data dependencies, these energy-bounded execution cycles minimize the number of system activations and nonvolatile data transfers, and thus the total energy overhead. We validate our methodology with two batteryless cameras running energy-intensive machine learning applications. Results demonstrate that compared to ad hoc solutions, our method can reduce the required energy storage by over 94% while only incurring a 0.12% energy overhead.
The advent of experimental science facilities-instruments and observatories, such as the Large Hadron Collider, the Laser Interferometer Gravitational Wave Observatory, and the upcoming Large Synoptic Survey Telescope-has brought about challenging, l
arge-scale computational and data processing requirements. Traditionally, the computing infrastructure to support these facilitys requirements were organized into separate infrastructure that supported their high-throughput needs and those that supported their high-performance computing needs. We argue that to enable and accelerate scientific discovery at the scale and sophistication that is now needed, this separation between high-performance computing and high-throughput computing must be bridged and an integrated, unified infrastructure provided. In this paper, we discuss several case studies where such infrastructure has been implemented. These case studies span different science domains, software systems, and application requirements as well as levels of sustainability. A further aim of this paper is to provide a basis to determine the common characteristics and requirements of such infrastructure, as well as to begin a discussion of how best to support the computing requirements of existing and future experimental science facilities.
Energy proportionality is the key design goal followed by architects of modern multicore CPUs. One of its implications is that optimization of an application for performance will also optimize it for energy. In this work, we show that energy proporti
onality does not hold true for multicore CPUs. This finding creates the opportunity for bi-objective optimization of applications for performance and energy. We propose and study the first application-level method for bi-objective optimization of multithreaded data-parallel applications for performance and energy. The method uses two decision variables, the number of identical multithreaded kernels (threadgroups) executing the application and the number of threads in each threadgroup, with the workload always partitioned equally between the threadgroups. We experimentally demonstrate the efficiency of the method using four highly optimized multithreaded data-parallel applications, 2D fast Fourier transform based on FFTW and Intel MKL, and dense matrix-matrix multiplication using OpenBLAS and Intel MKL. Four modern multicore CPUs are used in the experiments. The experiments show that optimization for performance alone results in the increase in dynamic energy consumption by up to 89% and optimization for dynamic energy alone degrades the performance by up to 49%. By solving the bi-objective optimization problem, the method determines up to 11 globally Pareto-optimal solutions. Finally, we propose a qualitative dynamic energy model employing performance monitoring counters as parameters, which we use to explain the discovered energy nonproportionality and the Pareto-optimal solutions determined by our method. The model shows that the energy nonproportionality in our case is due to the activity of the data translation lookaside buffer (dTLB), which is disproportionately energy expensive.
Batching is an essential technique to improve computation efficiency in deep learning frameworks. While batch processing for models with static feed-forward computation graphs is straightforward to implement, batching for dynamic computation graphs s
uch as syntax trees or social network graphs is challenging due to variable computation graph structure across samples. Through simulation and analysis of a Tree-LSTM model, we show the key trade-off between graph analysis time and batching effectiveness in dynamic batching. Based on this finding, we propose a dynamic batching method as an extension to MXNet Gluons just-in-time compilation (JIT) framework. We show empirically that our method yields up to 6.25 times speed-up on a common dynamic workload, a tree-LSTM model for the semantic relatedness task.
الأسئلة المقترحة
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
