Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

The release of the general AI model of Pengcheng mind: 200 billion parameters, domestic computing platform

2025-04-06 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

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

CTOnews.com September 21 news, this morning, Gao Wen, director of Pengcheng Laboratory and academician of the Chinese Academy of Engineering, officially released the "Pengcheng mind" (PengCheng Mind) general AI model at the 2023 Huawei full connection meeting.

Director Gao Wen said that the general AI model of "Pengcheng mind" realizes 200 billion parameters with a dense framework, relies on the localized AI computing platform of "Pengcheng Yunnao II" for full pre-training, adopts the localized deep learning framework of MindSpore, improves the large-scale parallel training strategy, the performance of underlying operators and fault-tolerant mechanism, and significantly improves the training efficiency of domestic computing platform. It also constructs a set of diversified corpus data set and data quality assessment tool set covering Chinese, English and more than 50 countries and regions along "Belt and Road Initiative".

This release is the basic version of "Pengcheng mind" with Chinese as the core, which will be open source to the whole society in the OpenI Qizhi community.

The official account of ▲ Tu Yuan Pengcheng Laboratory, together with Director Gao Wen, stressed that Pengcheng Lab is simultaneously developing the "China Computing Network, ClearNet" national integrated computing collaborative computing and dispatching platform, with Shenzhen as the general dispatching center, and has gathered national collaborative computing on a scale of 3e. " Pengcheng mind "big model" will further rely on the collaborative computing power connected by "China Computing Network", with the combination of "computing power dispatching + model empowerment", and join hands with the national new generation AI public computing platform to promote the continuous evolution and deployment and application of domestic AI model, and finally create a digital base of AI model and application ecology in various fields based on "China Computing Network".

At the same time, the laboratory will take the lead in the establishment of the "Pengcheng mind" open source consortium, a wide range of enterprises, universities and scientific research institutes. Through this new open source collective intelligence cooperation model, the "package" provides modeling tools covering compression, deployment, fine-tuning and value alignment, and strengthens the open source management and resource sharing of large models.

▲ "Pengcheng mind" big model promotion plan

▲ "Pengcheng mind" big model vision CTOnews.com with "Pengcheng mind" big model (basic version) open source link: https://openi.pcl.ac.cn/extension/mind

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.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report