Shulou(Shulou.com)11/24 Report--

CTOnews.com July 7 news, at today's Huawei developer Conference 2023, Huawei Cloud CEO Zhang Ping an announced the official release of Pangu Great Model 3.0.

▲ Tuyuanhua is the official live broadcast of Cloud, and the following is the same as Zhang Ping an. Pangu Model 3.0 is a series of large models that are completely industry-oriented, including 5+N+X three-tier architecture. The CTOnews.com is summarized as follows:

The first layer L0 is Pangu's five basic large models, including natural language model, visual model, multimodal model, predictive model and scientific computing model, which provide a variety of skills to meet the industry scene.

Pangu 3.0 provides customers with a series of basic models with 10 billion parameters, 38 billion parameters, 71 billion parameters and 100 billion parameters, which match the diversified needs of customers in different scenarios, different delays and different response speeds. At the same time, it provides a new set of capabilities, including NLP large model knowledge question and answer, copy generation, code generation, as well as multimodal large model image generation, image understanding and other capabilities, these skills can be directly used by customers and partner enterprises. No matter how large the parameter size of the large model, Pangu provides a consistent capability set.

The second layer L1 is N industry large models, which can not only provide industry general models for training using industry open data, including government affairs, finance, manufacturing, mining, meteorology, etc., but also can be based on the industry customers' own data. On the L0 and L1 of Pangu, train their own proprietary large models for customers.

The third layer L2 layer is to provide customers with more detailed scenarios of the model, it is more focused on a specific application scenario or specific business, to provide customers with out-of-the-box model services.

Pangu large model uses a complete hierarchical decoupling design, which can quickly adapt and quickly meet the changing needs of the industry. Customers can load independent datasets for their large models, upgrade the underlying model separately, or upgrade the capability set separately.

On the basis of L0 and L1 models, Huawei Cloud also provides customers with a large model industry development kit, through the secondary training of customer-owned data, customers can have their own industry models. At the same time, according to customers' different data security and compliance requirements, Pangu model also provides a variety of deployment forms of public cloud, large model cloud zone and hybrid cloud.

According to reports, Pangu can do web search, consult experts, refer to cases, consult books, and so on, so that each developer has his own programming assistant. It can even achieve a dialogue code generation, a keystroke use case test, a click automatic comment, and an instruction intelligent deployment, greatly simplifying the development process.

In the field of meteorology, the Pangu meteorological model is claimed to be the first AI prediction model whose accuracy is higher than that of traditional numerical forecasting methods, and the prediction speed has also been greatly improved. It used to take five hours to predict the path of a typhoon in the next 10 days on a high-performance computer cluster of 3000 servers. Now based on the pre-trained Pangu meteorological model, through AI reasoning, researchers only need a single card configuration on a single server, and more accurate prediction results can be obtained within 10 seconds.

According to Zhang Ping an, Pangu Model 3.0 continues to build its core competitiveness around the three major directions of "industry reshaping, technology taking root, opening up and flying together".

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.