Huawei Technologies Co., Limited

Youhua Zhang

Dandan Tu

A method and a device for training a model in a distributed system are disclosed, so as to reduce load of a master node () during model training. The method includes: receiving, by a parameter server () in a first slave node (), a training result sent by a parameter client () in at least one slave node () in the distributed system, where the first slave node () is any slave node () in the distributed system, and a parameter client () in each slave node () obtains a training result by executing a training task corresponding to a sub-model stored on a parameter server () in the slave node (); and updating, by the parameter server () in the first slave node () based on the received training result, a sub-model stored on the parameter server in the first slave node.

Growing community of inventors

Shenzhen, China

Youhua Zhang