Network model reasoning acceleration method and device, equipment and storage medium

The invention relates to the technical field of artificial intelligence, and provides a network model reasoning acceleration method and device, equipment and a storage medium, and the method comprises the steps: detecting whether a graphic processor exists or not when a network model starts to execu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: JIANG JIAJUN, CHENG JIEFENG
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 relates to the technical field of artificial intelligence, and provides a network model reasoning acceleration method and device, equipment and a storage medium, and the method comprises the steps: detecting whether a graphic processor exists or not when a network model starts to execute a reasoning task; when the graphics processor exists, determining a target calculation structure in the file of the network model; the target calculation structure comprises matrix multiplication; determining a target tensor associated with the target calculation structure; when the source program format requirement of the target tensor does not meet the preset calculation unified device architecture parallel program format requirement, modifying the source program format of the target tensor according to the preset calculation unified device architecture parallel program format requirement; and performing predictive reasoning on the reasoning task according to the modified source program of the target tensor ba