Communication method, terminal equipment and network equipment

Through the communication method between the terminal device and the network device in the large model fine-tuning system, the terminal device determines the first parameter based on the global model parameters and determines the local model aggregation parameter based on the candidate parameters, solving the problem of large transmission overhead in the large model fine-tuning, and achieving the effect of reducing communication overhead and resource occupation.

CN119732098APending Publication Date: 2025-03-28QUECTEL WIRELESS SOLUTIONS CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202480003356.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-25
Publication Date
2025-03-28

AI Technical Summary

Technical Problem

In the large-model fine-tuning system, due to the large-model parameters, a large number of model parameters need to be transmitted in each fine-tuning round, and the parameters interact frequently, which increases communication overhead.

Method used

By establishing a communication method between the terminal device and the network device, the terminal device receives the global model parameters, determines the first parameter, and determines the local model aggregation parameters based on the multiple candidate parameters to reduce the amount of transmitted data.

Benefits of technology

By selectively transmitting and aggregating local model parameters, the amount of data transmission and aggregation during model training is reduced, and communication overhead and resource usage is reduced.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a communication method, terminal equipment and network equipment. The communication method is applied to a machine learning model, and comprises the following steps: the terminal equipment receives a global model parameter which is sent by network equipment and is related to a first model; the terminal equipment sends a first parameter to the network equipment, wherein the first parameter is determined according to the global model parameter; the terminal equipment receives a plurality of candidate parameters sent by the network equipment; the terminal device determines a local model aggregation parameter sent to the network device according to the plurality of candidate parameters; wherein the terminal device is one of a plurality of terminal devices, the plurality of candidate parameters are determined according to a plurality of first parameters sent by the plurality of terminal devices, and a plurality of local model aggregation parameters of the plurality of terminal devices are used for training the first model.
Need to check novelty before this filing date? Find Prior Art