Unlock instant, AI-driven research and patent intelligence for your innovation.

A method and device for training a model in a distributed system

A distributed system and training model technology, applied in the transmission system, computing model, resource allocation, etc., can solve the problems of master node burden and heavy workload of the master node, and achieve the effect of reducing burden, improving efficiency and avoiding bottlenecks

Active Publication Date: 2021-06-22
HUAWEI TECH CO LTD
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] It can be seen that the master node needs to perform multiple model updates and parameter delivery. For large-scale training scenarios, the workload of the master node is relatively heavy, which brings a greater burden to the master node, and it is easy to make the master node a part of the entire training process. The bottleneck of the scene

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A method and device for training a model in a distributed system
  • A method and device for training a model in a distributed system
  • A method and device for training a model in a distributed system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0033] In order to make the objects, technical solutions, and advantages of the present invention more clearly, the technical solutions in the embodiments of the present invention will be described in contemplation in the embodiments of the present invention, and will be described, and the embodiments described herein will be described. It is a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained without making creative labor premises without making creative labor premises.

[0034] It should also be understood that various components may be described herein, various components may be described herein, but these terms are only used to distinguish the elements from each other. "Multiple" in the embodiment of the present invention refers to two or more than two or more. "And / or" describes the association relationship of the associated object, indicating that there are three r...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

A method and device for training a model in a distributed system, which are used to reduce the burden of the master node when performing the model. The method includes: the parameter server in the first slave node receives the training result sent by the parameter client in at least one slave node in the distributed system; wherein, the first slave node is any in the distributed system A slave node, the parameter client of each slave node obtains the training result by executing the training task corresponding to the submodel stored on the parameter server of the slave node; the parameter server in the first slave node obtains the training result according to the received training result Update its stored submodel.

Description

Technical field [0001] The present invention relates to the field of machine learning techniques, and more particularly to methods and apparatus for training models in a distributed system. Background technique [0002] Building a model in Machining Learning, ML is a key step in data mining (DM) tasks. Taking a generic parallel frame (Spark) as an example, when building a model, the master node can be issued to a plurality of slave (SLAVE) execution, generally need to experience multi-wheel iteration operation when performing tasks. After each cycle is over, each from the node needs to report the result of the iterative operation to the primary node, updated by the primary node, and send the updated parameter to each slave node, each starting from the node Execute the next round iterative operation. [0003] It can be seen that the main node needs to perform multiple model updates and parameters. For large-scale training scenarios, the workload of the main node is more heavy, bri...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F9/50G06N20/00
CPCG06N20/00G06F9/5027G06F2209/5017H04L41/0803H04L67/10
Inventor 张友华涂丹丹
Owner HUAWEI TECH CO LTD