US 12,340,251 B2
Scheduling method and device based on deep learning node computation, and storage medium
Kai Ma, Guangdong (CN); Chao Xiong, Guangdong (CN); Xinyu Niu, Guangdong (CN); and Kuen Hung Tsoi, Guangdong (CN)
Assigned to Shenzhen Corerain Technologies Co., Ltd., Shenzhen (CN)
Appl. No. 17/790,667
Filed by SHENZHEN CORERAIN TECHNOLOGIES CO., LTD., Guangdong (CN)
PCT Filed Dec. 31, 2020, PCT No. PCT/CN2020/142198
§ 371(c)(1), (2) Date Jul. 1, 2022,
PCT Pub. No. WO2021/136512, PCT Pub. Date Jul. 8, 2021.
Claims priority of application No. 202010004459.4 (CN), filed on Jan. 3, 2020.
Prior Publication US 2023/0034881 A1, Feb. 2, 2023
Int. Cl. G06F 9/50 (2006.01); G06F 9/48 (2006.01)
CPC G06F 9/5044 (2013.01) [G06F 9/4881 (2013.01); G06F 9/5016 (2013.01)] 9 Claims
OG exemplary drawing
 
1. A scheduling method based on a deep learning node computation, comprising:
acquiring a to-be-computed node of a preset neural network computation graph;
determining a node type of the to-be-computed node, wherein the node type comprises a hardware computation node and a software computation node;
in a case where the node type is the hardware computation node, scheduling the hardware computation node to a first queue, and determining whether a hardware computing power module corresponding to the hardware computation node is occupied or not; and
in a case where the hardware computing power module is not occupied, inputting the hardware computation node into the hardware computing power module for computing;
wherein the determining the node type of the to-be-computed node comprises determining whether a preset thread number is greater than zero, and determining the node type of the to-be-computed node in a case where the preset thread number is greater than zero.