Baidu Smart Cloud recently successfully illuminated the first self-developed Kunlun Chip 3rd Generation Wanka cluster in China. This breakthrough marks an important step in Baidu's field of artificial intelligence computing power. With the popularization of AI technology, the demand for computing power is growing. Baidu not only solves the problem of its own computing power supply through the construction of self-developed chips and Wanka clusters, but also provides the industry with new development ideas.
Baidu Smart Cloud recently successfully illuminated the first self-developed Kunlun Core third-generation Wanka cluster in China. This milestone breakthrough not only marks an important step in Baidu's field of artificial intelligence computing power, but also provides new development ideas for the entire industry. With the continuous advancement of technology, the improvement of computing power is particularly critical to supporting the training and application of large-scale models.
In the past year, with the popularization of AI technology, many companies have faced the problem of computing power shortage, which directly led to the high cost of using large models. Baidu said that through the construction of self-developed chips and Wanka clusters, they not only effectively solved their own computing power supply problems, but also provided reference and support for other companies. The Wanka cluster has the ability of super-large-scale parallel computing, which can significantly improve training efficiency, especially when training complex 100 billion parameter models, it can greatly shorten the training cycle.
The application of Wanka cluster will meet the needs of rapid iteration of AI native applications, and can also support the processing of trillion-domain parameter models and multimodal data, providing powerful impetus for the development of Sora-like applications. In addition, the multi-task concurrency capability of the Wanka cluster allows it to train multiple lightweight models simultaneously through dynamic resource segmentation, thereby achieving efficient utilization of computing power. This innovation of Baidu Smart Cloud not only improves the comprehensive utilization rate of the cluster, but also greatly reduces the cost of unit computing power.
However, problems such as multi-core mixed training and surge in failure rates in the past have become the main challenges in the deployment of Wanka clusters. In order to solve these problems, Baidu launched the upgraded version of Baige AI heterogeneous computing platform 4.0 in September 2024, which played a crucial role in the construction of Wanka cluster. Through model optimization, parallel strategies and dynamic resource allocation, Baidu Intelligent Cloud is promoting the effective utilization of computing power and laying the foundation for future AI applications.
The success of Baidu Smart Cloud not only demonstrates the strength of independent research and development, but also injects new impetus into the vigorous development of domestic large models. In the future, with the continuous expansion and optimization of the Wanka cluster, we look forward to more innovative AI applications being implemented and bringing new opportunities for industry development.
The success of Baidu Smart Cloud not only demonstrates the strength of independent research and development, but also injects new impetus into the vigorous development of domestic large models. In the future, with the continuous expansion and optimization of the Wanka cluster, we look forward to more innovative AI applications being implemented and bringing new opportunities for industry development.