US 12,367,048 B2
Method for loading multiple neural network models and electronic device
Guo-Chin Sun, New Taipei (TW); and Chin-Pin Kuo, New Taipei (TW)
Assigned to HON HAI PRECISION INDUSTRY CO., LTD., New Taipei (TW)
Filed by HON HAI PRECISION INDUSTRY CO., LTD., New Taipei (TW)
Filed on Jan. 11, 2022, as Appl. No. 17/572,867.
Claims priority of application No. 202110036325.5 (CN), filed on Jan. 12, 2021.
Prior Publication US 2022/0222084 A1, Jul. 14, 2022
Int. Cl. G06F 9/445 (2018.01); G06F 9/50 (2006.01); G06F 16/174 (2019.01); G06N 3/045 (2023.01); G06N 3/082 (2023.01)
CPC G06F 9/445 (2013.01) [G06F 9/5027 (2013.01); G06F 16/1744 (2019.01); G06N 3/045 (2023.01); G06N 3/082 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for loading multi neural network model comprising:
compiling at least two neural network models and generating at least two binary model files corresponding to the at least two neural network models;
taking one of the at least two binary model files as the basic model, calculating and recording the difference between each binary model file except the basic model in the at least two binary model files and the basic model using preset difference calculation method, and generating a differences file;
compressing the basic model and the differences file using a preset compression method, and generating an input file; and
inputting the input file in a neural network accelerator, decompressing the input file to obtain the basic model and the differences file, and loading the basic model and the differences file in the neural network accelerator.