US 12,288,051 B2
Method and apparatus for processing mirror images of ai platform
Jiyu Wang, Jiangsu (CN)
Assigned to INSPUR SUZHOU INTELLIGENT TECHNOLOGY CO., LTD., Jiangsu (CN)
Appl. No. 18/012,934
Filed by INSPUR SUZHOU INTELLIGENT TECHNOLOGY CO., LTD., Jiangsu (CN)
PCT Filed Jul. 29, 2021, PCT No. PCT/CN2021/109216
§ 371(c)(1), (2) Date Dec. 25, 2022,
PCT Pub. No. WO2022/052657, PCT Pub. Date Mar. 17, 2022.
Claims priority of application No. 202010955468.1 (CN), filed on Sep. 11, 2020.
Prior Publication US 2023/0266956 A1, Aug. 24, 2023
Int. Cl. G06K 9/00 (2022.01); G06F 3/01 (2006.01); G06F 8/61 (2018.01); G06F 16/11 (2019.01); G06K 9/62 (2022.01); G06N 20/00 (2019.01)
CPC G06F 8/63 (2013.01) [G06F 16/116 (2019.01); G06N 20/00 (2019.01)] 17 Claims
OG exemplary drawing
 
1. A method for processing mirror images of an artificial intelligence (AI) training platform, comprising:
scanning, in response to receiving a mirror image building request, a file used for building a mirror image to detect whether a format and syntax of the file are correct, and building the mirror image according to the file in response to detecting that the format and syntax of the file are correct;
selecting and installing, according to a type of the mirror image and component information, a component required for mirror image training in response to receiving a request for importing the mirror image to the training platform;
adjusting queue positions of mirror images to be imported according to priorities of the mirror images, and importing the mirror images according to the queue positions;
in response to receiving a model training request, classifying and selecting mirror images according to information about mirror images and information about training tasks, and performing a model training according to the mirror images selected;
in response to detecting that a storage of mirror images in a local disk space reaches a preset cleaning threshold, screening, according to usage information of the mirror images in the local disk space, the mirror images in the local disk space, and cleaning mirror images screened out;
monitoring a health state of a mirror image repository in real time, and determining whether a standby mirror image repository is provided in response to monitoring that the health state is abnormal;
in response to determining that the standby mirror image repository is provided, transferring traffic in the mirror image repository to the standby mirror image repository;
in response to determining that the standby mirror image repository is not provided, detecting the health state of the mirror image repository at a preset time interval, and sending abnormality prompt information in response to a number of detections reaching a preset number and determining that the health state is still abnormal;
displaying a progress and queue positions of the mirror image being operated, and log information in real time; deleting tasks in queue state; and terminating mirror images not requiring to be built and imported; or any combination thereof;
wherein the classifying and selecting mirror images according to information about mirror images and information about training tasks comprises: in response to the selected mirror images in the mirror image repository being deleted, displaying a prompt and recommending a mirror image similar to the selected mirror images; and
wherein the detecting the health state of the mirror image repository comprises:
rebooting the mirror image repository;
detecting the health state of the mirror image repository after waiting for a first period; and
in response to detecting that the mirror image repository is abnormal, rebooting the mirror image repository after a second period, the second period is longer than the first period.