Task scheduling method, scheduling module, reasoning node and collaborative operation system

Abstract

The invention provides a task scheduling method, a scheduling module, reasoning nodes and a collaborative operation system.The method is applied to the scheduling module and comprises the following steps: to-be-processed information of all the reasoning nodes is obtained, and all the reasoning nodes work collaboratively; the to-be-processed information comprises the number of to-be-processed tasks under the corresponding reasoning node and / or the task type of each to-be-processed task; based on the to-be-processed information of each reasoning node and the importance of each reasoning node in the collaborative operation, target nodes are selected from each reasoning node, and the sum of calculation resources required by all target nodes for task processing is smaller than or equal to the rated calculation amount; and the task processing instruction is sent to the target node to trigger the target node to perform task processing. Requirements for overall throughput and task response time are met, the problem that a scheduling reasoning scheme of a single neural network cannot perform scheduling reasoning on tasks cooperatively working by a plurality of neural networks is solved, and scheduling reasoning in a complex scene is realized.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a task scheduling method, a scheduling module, a reasoning node and a collaborative operation system. Background technique [0002] With the extensive development of cloud multi-channel neural network application scenarios, the application of cloud multi-channel neural network has extended from the initial offline recognition to real-time online, and has changed from the application of a single network to the collaborative work of multiple networks. In this process, it is not only necessary to meet the requirements of throughput, but also the requirements of task response time. [0003] At present, for the application scenario of a single neural network, the number of data splicing units of the task is usually increased to improve the efficiency of equipment usage and meet the requirements of the overall task throughput. The tasks of the network are effective, and it is imposs...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a task scheduling method, a scheduling module, a reasoning node, and a cooperative operation system to solve the defect in the prior art that it is impossible to schedule reasoning tasks for multiple neural networks working together

Images