According to the official account of “Huawei China Government and Enterprise”, Huawei and the Industrial and Commercial Bank carried out joint innovation and successfully implemented a serverless NPU flexible computing power scheduling technology solution. Actual measurement results show that compared with traditional computing power scheduling technology solutions, the serverless NPU flexible computing power scheduling technology solution can shorten the startup time of 100 billion MoE large model inference services to 100 seconds, increase startup efficiency by more than 10 times, and transform the computing power supply model from “long-term binding” to “on-demand use”.

According to the official account of “Huawei China Government and Enterprise”, Huawei and the Industrial and Commercial Bank carried out joint innovation and successfully implemented a serverless NPU flexible computing power scheduling technology solution. Actual measurement results show that compared with traditional computing power scheduling technology solutions, the serverless NPU flexible computing power scheduling technology solution can shorten the startup time of 100 billion MoE large model inference services to 100 seconds, increase startup efficiency by more than 10 times, and transform the computing power supply model from “long-term binding” to “on-demand use”.