Multi-task execution method and device based on small language model, equipment and medium

By performing Lora fine-tuning and weight merge on small language models, the problems of insufficient semantic understanding capabilities of small language models and high deployment costs of large language models are solved, and efficient multi-task deployment and cost reduction in financial scenarios are achieved.

CN119647459APending Publication Date: 2025-03-18SHENZHEN QIANHAI HUANRONG LIANYI INFORMATION TECHNOLOGY SERVICES CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202411707104.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-25
Publication Date
2025-03-18

AI Technical Summary

Technical Problem

In the prior art, the semantic understanding ability of small language models is limited and it is difficult to meet the needs of complex financial scenarios. The high cost and difficulty of deployment of large language models are not suitable for the deployment needs of large-scale or edge devices.

Method used

By performing Lora fine-tuning on a small language model, Lora weights of different task types are dynamically obtained, and weight calculations are performed separately under the initial weight and Lora weights, and the calculation results are combined to generate the final task execution results.

Benefits of technology

While maintaining task performance, the deployment cost is significantly reduced, and multi-task deployment in financial scenarios is realized, suitable for small-scale or edge devices deployment.

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

Abstract

The invention relates to an artificial intelligence technology, and provides a multi-task execution method and device based on a small language model, equipment and a medium, and the method comprises the steps: obtaining different Lora weights from a main memory of a central processing unit according to different task types, and loading the different Lora weights into a video memory of a graphics processor, according to the method, weight calculation is further executed under the initial weight and the Lora weight of the small language model, calculation results are combined to generate a final task execution result, multi-task deployment in a financial scene can be achieved based on Lora fine adjustment and a small-scale model, the task performance is guaranteed, and the deployment cost is reduced.
Need to check novelty before this filing date? Find Prior Art