您当前的位置是:  首页 > technology > Corporate communications > technology > voice communication > Technical dynamics >
  Home >technology > Corporate communications > technology > voice communication > Technical dynamics > Shen Jian Technology Speech Recognition Acceleration System Launches AWS

Shen Jian Technology Speech Recognition Acceleration System Launches AWS

- Provide Cloud Acceleration Services to Global Users

2018-03-06 14:00:04 Author: Source:CTI ForumComment:0  Click:


  Became the first Chinese artificial intelligence company to go online with FPGA platform acceleration solution
On March 5, 2018, SZM announced that it has officially launched the DEEPE (DeePhi Descartes Efficient Speech Recognition Engine) based on the deep Descartes sparse neural network processing architecture on Amazon AWS, and began to provide accelerated trial services to users around the world. The release of the DDESE automatic speech recognition acceleration engine at AWS marks the beginning of Shenzhen Jianmei Technology becoming the first Chinese artificial intelligence company to launch an AWS-based FPGA-based acceleration solution.
 
FPGA chips can be customized and versatile to meet the needs of cloud computing, convenient and economical, and easy to expand. Since 2016, major cloud service providers have deployed FPGA cloud acceleration services. The cooperation between Amazon Web Services (AWS), the world's largest cloud provider, and Xilinx, the leading FPGA manufacturer, has fully demonstrated that cloud computing data centers will be an important market for FPGA application release potential. At the same time, through the influence and service capabilities of large-scale data centers, FPGAs affect hundreds of thousands of application developers and enterprise users, and quickly become the favored application trend of public cloud services.
Deeply aware of the DDESE focused speech recognition scenario released by the company, Xilinx built an automatic speech recognition system acceleration engine based on FPGA as the computing platform, which can provide users with end-to-end speech recognition services on AWS. Based on the FPGA platform, combined with deep learning technology algorithms, software and hardware co-designed deep learning acceleration solutions, the current DDESE initial version is twice as fast as the GPU computing speed in the calculation of speech recognition. This time, Shenzhen Sci-Tech acquired AWS system service authorization and deployed an FPGA speech recognition acceleration solution in the cloud, which can provide accelerated services to global users, and support it in reducing development and deployment costs, eliminating hardware barriers, and through collaborative optimization of hardware and software. Get more efficient cloud computing capabilities.
  
The deep LSTM acceleration engine can effectively reduce the latency of AI timing applications such as speech recognition and machine translation. Deep CTO said: "The purpose of releasing DDESE on Amazon AWS is to demonstrate AI acceleration capability through Xilinx FPGA-based ESE engine, and let more customers understand the deep voice recognition full-process solution through cloud promotion. We are very pleased to work with Xilinx and AWS to build AI Cloud Acceleration Ecology."
During the technical testing phase, using the same network and input data, the end-to-end delay of the automatic speech recognition engine compares to GPU (Tesla P4 + cudnn6020), CPU (Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz The acceleration ratios of 8 and 9 are respectively 2.06 times and 6.59 times, and the acceleration ratios of the LSTM parts are 2.82 times and 8.64 times, respectively. Shen Jian will also release a new version of DDESE soon. In the future, the combination of FPGA and cloud services will continue to expand and deepen. DeepSense Technology will continue to optimize the power consumption performance of the technology, expand the acceleration characteristics of FPGAs, and accelerate the speed of public cloud services.
 
About Shenjian Technology
Shen Jian Technology was founded by Stanford and Tsinghua University's deep learning hardware acceleration researchers in China. It has always been a leader in the field of FPGA technology and deep learning of AI-specific chip development speed. Collaborating with FPGA giant Xilinx for a long time, DDESE's core technology, the ESE (Highly Efficient Speech Recognition Engine), was first researched in 2017 as the top conference FPGA chip field FPGA 2017 Best Paper Award. Shenjian Technology looks forward to in-depth cooperation with AWS to output more convenient, more efficient and more economical AI cloud acceleration services.
About Xilinx
Xilinx (NASDAQ:XLNX) is the world's leading supplier of All Programmable FPGAs, SoCs, MPSoCs, RFSoCs, and 3D ICs, uniquely enabling applications that are both software-defined and hardware-optimized. The development of industries such as cloud computing, 5G wireless, embedded vision and industrial Internet of things.
[Disclaimer] This article only represents the author's own opinion and has nothing to do with the CTI Forum. The CTI Forum maintains its neutrality in the presentation of statements and opinions, and does not provide any express or implied warranty for the accuracy, reliability or completeness of the contents contained therein. Readers are for reference only, and please bear full responsibility for yourself.

Topics