Huawei Cloud Pangu Large Model 3.0 is released and AI cloud services are launched at the same time: 20 billion petaflops performance

On the afternoon of July 7, Kuai Technology announced at the Huawei Developer Conference 2023 (Cloud) that Huawei Managing Director and Huawei Cloud CEO Zhang Pingan announced that Huawei Cloud Pangu Large Model 3.0 was officially released.

Zhang Pingan said that Pangu Large Model 3.0 is a large model series completely oriented to the industry, including 5 N Its core competitiveness will reshape thousands of industries.

The first layer L0 is Pangu’s five basic large models, including natural language large model, visual large model, multi-modal large model, prediction large model, and scientific computing large model. They provide a variety of models that meet industry scenarios. Skill.

The second layer, L1, is N large industry models. We can provide industry-wide large models trained using industry public data, including government affairs, finance, manufacturing, mining, meteorology, etc.; we can also provide self-owned models based on industry customers. Data is used on Pangu’s L0 and L1 to train its own proprietary large models for customers.

The third layer, the L2 layer, is a model that provides customers with more detailed scenarios. It focuses more on a specific application scenario or a specific business and provides customers with out-of-the-box model services.

In addition to Pangu Big Model 3.0, Huawei Cloud also announced the launch of Shengteng AI cloud service, which provides long-term and stable AI computing power, can reach 2000P Flops, which is 20 billion billion times of performance. Currently, a single The cluster's Shengteng AI cloud service was launched simultaneously in Ulanqab and Gui'an.

Shengteng AI Cloud Service can provide longer and more stable AI computing power services. The long-term stability rate of kilocalorie training reaches 90% for 30 days, and the breakpoint recovery time does not exceed 10 minutes.

In addition to supporting Huawei's AI framework Mindspore, Ascend Cloud Service also supports mainstream AI frameworks such as Pytorch and Tensorflow. 90% of the operators in the framework can be smoothly migrated from GPU to Ascend through Huawei's migration tool.

Huawei Cloud Zhang Pingan said that with the Ascend AI cloud service, everyone does not need to increase the price to grab GPU computing power. Huawei's AI cloud service can satisfy it.

The above is the detailed content of Huawei Cloud Pangu Large Model 3.0 is released and AI cloud services are launched at the same time: 20 billion petaflops performance. For more information, please follow other related articles on the PHP Chinese website!