Mode conversion method and device in deep learning framework and computer storage medium

• Main Authors: ZHENG HUIHUANG, LIU HONGYU, WANG HAIFENG, YU DIANHAI, LI YAMEI, ZHANG LIUJIE, WU TIAN, LAN XIANG, MA YANJUN
• Format: Patent
• Language: Chinese; English
• Published: 26.01.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

According to an example embodiment of the present disclosure, a mode conversion method and device in a deep learning framework, a computer storage medium a computer program product are provided, whichcan be used for deep learning framework construction. The mode conversion method in the deep learning framework comprises the following steps: updating a first mode based on first syntax elements inthe first mode and a context relationship among the first syntax elements in the first mode to obtain an updated first mode; generating second syntax elements corresponding to the updated first syntaxelements based on a mapping relationship between the updated first syntax elements in the updated first mode and the second syntax elements in the second mode system; and combining the second syntaxelements according to the updated context relationship of the first syntax elements to generate a second mode. According to the scheme disclosed by the invention, the conversion between the modes canbe efficiently executed. 根据本公