Search AI Products and News

Baidu's deep learning platform, PaddlePaddle, recently announced the official release of its new generation framework, PaddlePaddle 3.0. This release introduces five core technological innovations, including "dynamic and static unified automatic parallelism," aiming to effectively reduce the development and training costs of large models and support the infrastructure construction of the large model era. As the core infrastructure supporting large model training and inference tasks, PaddlePaddle 3.0 demonstrates excellent performance optimization. The framework already supports multiple mainstream large models, including Wenxin 4.5 and Wenxin X1, and through optimization...

Beijing Baidu Network Technology Co., Ltd. and Beijing Zhiyuan Artificial Intelligence Research Institute announced today the formal signing of a strategic cooperation agreement, under which the two sides will engage in deep collaboration in large model and other fields, jointly constructing an AI industry-research collaborative ecosystem. Baidu has been comprehensively laying out its artificial intelligence strategy since 2010 and is one of the few companies globally that has fully integrated its AI efforts, from Kunlun chips, PaddlePaddle deep learning platform, Wenxin large model to applications.

The release of PaddlePaddle Framework 3.0 features a core upgrade to unified automatic parallelism technology, aimed at simplifying the distributed training process for large models and improving development efficiency. The new version supports mixed parallelism techniques from four-dimensional to five-dimensional, employing various parallel methods such as data parallelism, tensor model parallelism, pipeline parallelism, and grouped parameter slicing parallelism, effectively enhancing the training efficiency of large models. The automatic parallelism technology uses tensor slicing syntax to automatically infer distributed slicing states and add communication operators, reducing the difficulty of development. The principles of automatic parallelism include distributed tensor representation and slicing inference.

PaddleX 3.0-beta Ascend version is a low-code development tool launched by PaddlePaddle, aimed at helping developers solve practical industry problems with low cost and zero barriers. This version was released on June 27, 2023, focusing on seven mainstream AI scenarios, featuring 68 high-quality PaddlePaddle models, and building 16 industry-grade model production lines.

The user base of Wenxin Yiyuan has surpassed 100 million, with the number of PaddlePaddle developers reaching 10.7 million. The Wenxin large model 4.0 has increased its effectiveness by 32% in just two months, making it easier to develop AI-native applications. The Baidu Conference unveiled new progress in large models and PaddlePaddle, promoting the customization of personal AI applications for everyone. The Galaxy Community's large model tools center has been launched, providing developers with comprehensive AI application services.

The user base of Wenxin Yiyan has surpassed 100 million since it opened public services on August 31. The volume of user inquiries continues to rise, and the massive data from user questions continuously optimizes the intelligence level of Wenxin Yiyan. The number of developers on the PaddlePaddle platform has reached 10.7 million.