AI Technology Go to Sea – Aliyun GPU Server Helps Win Four World First

Time:2019-9-11

Abstract:On September 9, 2018, the COCO+Mapillary Joint Object Recognition Challenge Seminar was held as scheduled at the ECCV 2018 Conference, at which the ranking of the competition was announced and the Chinese team won all the championships. Among them, Aliyun GPU Client – Ignoring Science and Technology won 4 world first prizes at one stroke, becoming the biggest winner of this competition.

MSCOCO (Microsoft Common Objects in Context) is one of the most concerned and authoritative competitions in the field of computer vision in the world today. On September 9, the COCO+Mapillary Joint Object Recognition Challenge Seminar was held as scheduled at the ECCV 2018 Conference, at which the ranking of the competition was announced and all the champions were won by the Chinese team. Among them, Aliyun GPU Client – Ignoring Science and Technology won 4 world first prizes at one stroke, becoming the biggest winner of this competition.

AI Technology Go to Sea - Aliyun GPU Server Helps Win Four World First

It is reported that this year’s Joint Challenge includes six challenges, including COCO4 and Mapillary 2. The team participated in three COCO competitions (instance segmentation, panoramic segmentation, human key point detection) and one Mapillary competition (street panoramic segmentation), and all won the championship.

In the field of artificial intelligence, an excellent product can not be separated from the three elements of computing, algorithm and data. Behind the success of science and technology, it is Aliyun’s latest generation of cloud service servers, GN6 family, which includes commercially deployed GN6V and SCCGN6 under public testing, that provide powerful financial support.

GN6 family is a server platform designed to meet the needs of AI and high-performance computing customers. It has many technological breakthroughs.

The GN6 family carries NVIDIA Volta-based Tesla V100 computing cards with a maximum specification of 8 V100 GPUs.

Each GPU has 640 Tensor cores and 125 TeraFlops deep learning and computing capabilities. Compared with the previous generation of GN5 (P100) family, GN6 can provide up to 12 times higher floating point computing capabilities in AI training applications.

Between GPUs, a new generation of NVIDIA NVLink is adopted to communicate with each other at a speed of up to 300 GB/s.

CPU uses Intel Xeon Platinum 8163 (Skylake) processor with 2.5 GHz main frequency.

On storage, it supports the latest generation of cloud disk products ESSD of Aliyun Elastic Computing, providing high reliable cloud disk storage capacity with ultra-low latency.

In terms of network capability, SCCGN6 provides an ultra-low delay RDMA network up to 50Gbps.

In deployment, Aliyun’s GPU cloud service supports automatic installation of one-button CUDA driver, assists developers to quickly access various optimized in-depth learning frameworks with perfect support of NVIDIA NGC container mirroring, and supports GPU host cloud monitoring in operation and maintenance, which can real-time understand the running status of GPU.

Details of the GN6V specification family currently deployed commercially are as follows:

AI Technology Go to Sea - Aliyun GPU Server Helps Win Four World First

Aliyun has a large-scale and technologically advanced public cloud GPU computing cluster in China, providing heterogeneous computing acceleration platform for multi-scenarios, serving more than 1000 enterprise users. With its security, stability, high performance and flexibility, Aliyun GPU cloud server is helping AI and high performance computing enterprises in technology. Continuously innovate and succeed in business!

Regarding ignorance of science and technology:

Ignorance technology is a solution provider of Material Federation with artificial intelligence technology as the core, dedicated to creating the greatest value for customers and society with extraordinary technology. As a representative AI original technology enterprise in China, it has nearly 800 patents applied and authorized at home and abroad, and participated in 19 AI national and industry standards formulation on behalf of leading technology providers. Including MS COCO, Ignorance has defeated the giants such as Google, Facebook and Microsoft many times in various top international AI competitions and won the first place in 22 world technology evaluations. In the commercial field, the face recognition technology, image recognition technology, intelligent video cloud products, intelligent sensor products, intelligent robot products have been widely used in finance, mobile phones, security, logistics, retail and other fields.

Regarding MS COCO:

The full name of MS COCO is Microsoft Common Objects in Context, which originated from Microsoft’s annotated Microsoft COCO dataset in 2014. Like the famous ImageNet contest, the same name contest is regarded as one of the most concerned and authoritative contests in the field of computer vision.

Author: Goose Eagle

Read the original text

This article is the original content of Yunqi Community, which can not be reproduced without permission.

Recommended Today

Hadoop MapReduce Spark Configuration Item

Scope of application The configuration items covered in this article are mainly for Hadoop 2.x and Spark 2.x. MapReduce Official documents https://hadoop.apache.org/doc…Lower left corner: mapred-default.xml Examples of configuration items name value description mapreduce.job.reduce.slowstart.completedmaps 0.05 Resource requests for Reduce Task will not be made until the percentage of Map Task completed reaches that value. mapreduce.output.fileoutputformat.compress false […]