Huawei Releases New Atlas 900 SuperCluster for Rise AI Computing Cluster, Supporting Training of Large Models with Over a Trillion Parameters

Huawei (brand)At the recent Fully Connected Conference, the official launch of the newrise AI computing cluster, the Atlas 900 SuperCluster, which utilizes a newHuawei (brand)StarHub AI Intelligent Computing Switch CloudEngine XH16800, and with its high-density 800GE port capability, realizes ultra-large-scale non-convergent cluster networking. The new cluster uses an innovative supernode architecture that greatly improves large model training capabilities.

In addition, Huawei released the more open and easy-to-use CANN 7.0 Heterogeneous Computing Architecture, which is fully compatible with the industry's AI frameworks, acceleration libraries, and mainstream big models, and deeply opens up the underlying capabilities so that AI frameworks and acceleration libraries can more directly invoke and manage computational resources, enabling developers to customize high-performance arithmetic and improve the competitiveness of big models.

Huawei Releases New Atlas 900 SuperCluster for Rise AI Computing Cluster, Supporting Training of Large Models with Over a Trillion Parameters

Huawei has also upgraded the Ascend C programming language to simplify the logic of operator implementation with more efficient programming, significantly shorten the development cycle of fusion operators, and empower the rapid development of AI models and applications.

Huawei's cloud website has also officially launched the "Hundreds of Models and Thousands of Styles" area of the Rise AI cloud service, which includes the industry's mainstream open-source models and is adapted and optimized based on the Rise AI cloud service. At the same time, Huawei provides a tool chain for application development, and the development tools have all been cloud-based, eliminating the need for a cumbersome configuration process and realizing one-click access and instant use.

Huawei Releases New Atlas 900 SuperCluster for Rise AI Computing Cluster, Supporting Training of Large Models with Over a Trillion Parameters

It is reported that the Rise AI cluster has already supported 25 cities in China.artificial intelligence (AI)Seven city public computing platforms have been selected as the first batch of national "New Generation of Artificial Intelligence Public Computing Power Open Innovation Platforms". Meanwhile, Rise AI has developed more than 30 hardware partners and 1,200 ISVs, and jointly launched more than 2,500 industry AI solutions, serving operators, the Internet, finance and other industries.

Huawei's new Atlas 900 SuperCluster, the new Rising AI compute cluster released this time, is a technological breakthrough that elevates the large model training capability to a new level, and at the same time, by being fully compatible with the industry's AI frameworks, acceleration libraries, and mainstream large models, and by providing a heterogeneous compute architecture that deeply opens up the underlying capabilities, it makes it easier for developers to invoke and manage compute resources, thus greatly simplifying the development process. In addition, Huawei upgraded the Ascend C programming language to improve development efficiency, and launched the Ascend AI Cloud Service "Hundreds of Models and Thousands of Styles" zone, which provides comprehensive and optimized industry AI solutions for global enterprises and developers. The launch of this technology will further promote the development and application of AI and help enterprises and developers differentiate themselves in various fields.

This article comes from users or anonymous contributions, does not represent the position of Mass Intelligence; all content (including images, videos, etc.) in this article are copyrighted by the original author. Please refer to this site for the relevant issues involvedstatement denying or limiting responsibilityPlease contact the operator of this website for any infringement of rights (Contact Us) We will handle this as stated. Link to this article: https://dzzn.com/en/2023/1417.html

Like (0)
Previous September 24, 2023 at 12:58 pm
Next September 25, 2023 am7:07 am

Recommended

Leave a Reply

Please Login to Comment