Title:
MODEL WEIGHT ACQUISITION METHOD AND RELATED SYSTEM
Document Type and Number:
WIPO Patent Application WO/2023/078009
Kind Code:
A1
Abstract:
The present application provides a model weight acquisition method, comprising: acquiring structure information of a first neural network model; according to the structure information of the first neural network model, searching for a reference weight from a weight library storing a plurality of groups of historical weights, the reference weight being the weight of the second neural network model having a similar structure to the first neural network model in the plurality of groups of historical weights; and converting the reference weight to acquire the weight of the first neural network model. In the method, the weight of a neural network model having a similar structure to a neural network model to be trained in the weight library is searched, and the weight is converted to acquire an available weight of the neural network model to be trained; it is unnecessary to perform model training from the beginning, and the model training efficiency is improved.
More Like This:
Inventors:
ZHOU WEIMIN (CN)
LIU YIFENG (CN)
LI YI (CN)
DAI ZONGHONG (CN)
LIU YIFENG (CN)
LI YI (CN)
DAI ZONGHONG (CN)
Application Number:
PCT/CN2022/122500
Publication Date:
May 11, 2023
Filing Date:
September 29, 2022
Export Citation:
Assignee:
HUAWEI CLOUD COMPUTING TECH CO LTD (CN)
International Classes:
G06N20/00; G06N3/08
Domestic Patent References:
| WO2020160787A1 | 2020-08-13 |
Foreign References:
| CN112949711A | 2021-06-11 | |||
| CN111126617A | 2020-05-08 | |||
| CN113128678A | 2021-07-16 | |||
| CN109783713A | 2019-05-21 |
Attorney, Agent or Firm:
SHENPAT INTELLECTUAL PROPERTY AGENCY (CN)
Download PDF: