DISTRIBUTED TRAINING CONTAINER SCHEDULING FOR INTELLIGENT COMPUTING

Title:

DISTRIBUTED TRAINING CONTAINER SCHEDULING FOR INTELLIGENT COMPUTING

Document Type and Number:

WIPO Patent Application WO/2024/007849

Kind Code:

A1

Abstract:

A method for distributed training container scheduling for intelligent computing. The method comprises: splitting a target model to obtain a plurality of sub-models; according to the plurality of sub-models, determining at least one computing node for deploying the plurality of sub-models, and creating a plurality of containers on the at least one computing node, so as to respectively deploy the plurality of sub-models into the plurality of containers; using sample data to execute a model training task, so as to train the plurality of sub-models deployed in the plurality of containers; according to load data of the at least one computing node and operation durations respectively corresponding to the plurality of containers, determining a computing node with a container distribution needing to be adjusted, and taking same as a target node; with an adjustment target of the operation durations that respectively correspond to the plurality of containers being close to each other, adjusting the distribution of containers in the target node; and executing a training task for the target model on the basis of the computing node after the container distribution adjustment.

More Like This:

WO/2021/238305	UNIVERSAL DISTRIBUTED GRAPH PROCESSING METHOD AND SYSTEM BASED ON REINFORCEMENT LEARNING
JP2023046498	MACHINE LEARNING DEVICE, MACHINE LEARNING METHOD AND MACHINE LEARNING PROGRAM
JP7384267	Processing equipment, processing method and program

Inventors:

LI YONG (CN)
CHENG WEN (CN)
WU YUNXIANG (CN)
CHEN GUANG (CN)
ZHU SHIQIANG (CN)
ZENG LINGFANG (CN)

Application Number:

PCT/CN2023/101093

Publication Date:

January 11, 2024

Filing Date:

June 19, 2023

Export Citation:

Click for automatic bibliography generation Help

Assignee:

ZHEJIANG LAB (CN)

International Classes:

G06N20/00; G06F9/455

Foreign References:

CN116167463A	2023-05-26
CN115248728A	2022-10-28
CN115827253A	2023-03-21
CN113110914A	2021-07-13
CN114091536A	2022-02-25
CN114787833A	2022-07-22
US20220207716A1	2022-06-30

Attorney, Agent or Firm:

BEIJING BESTIPR INTELLECTUAL PROPERTY LAW CORPORATION (CN)

Download PDF:

View/Download PDF PDF Help

Previous Patent: RGBWCLA-BASED SEVEN COLOR-IN-ONE LED FULL-SPECTRUM COLOR MIXING METHOD

Next Patent: ANTI-INFLUENZA B VIRUS ANTIBODY, AND REAGENT AND KIT FOR DETECTING INFLUENZA B VIRUS