Network model compression method and device, electronic equipment and storage medium

The invention provides a network model compression method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining an initial weight parameter and an initial bias parameter of a to-be-compressed network model; based on the initial weight parameter and the...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ZHOU WEIXIN, XIAO WAN'ANG
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention provides a network model compression method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining an initial weight parameter and an initial bias parameter of a to-be-compressed network model; based on the initial weight parameter and the initial bias parameter, determining a target reasoning model, the target reasoning model being used for reasoning states of different gates in the to-be-compressed network model at different moments; performing preset gate clipping on the target reasoning model, and determining a target weight parameter and a target offset parameter of the model obtained by clipping; and determining a target network model based on the target weight parameter and the target bias parameter. By using the method of the invention, not only are the defects of increased model parameter quantity and calculation quantity and increased storage difficulty caused by introduction of a sparse matrix avoided, but also the purposes of reducing the