Dynamic matching algorithm, system and device based on GPU resource occupation characteristics

Abstract

The invention discloses a dynamic matching algorithm, system and device based on GPU occupied resource characteristics, and the method comprises the steps: obtaining a training model task, carrying out the task analysis of the received training model task, and obtaining a task queue of task priority sorting and the resource occupied share of the task in each time period; allocating GPU resources from the GPU cluster based on the resource occupation share of the task in each time period; and executing a training task according to the obtained task queue with the task priorities sorted and the allocated GPU resources, and storing a model obtained after training. In a deep learning modeling task, a platform user submits a task of a training model to a platform to schedule the task and distribute resources, the resource quantity and the resource occupation time are comprehensively considered, the resource distribution dimensionality is increased, and resource scheduling is planned more reasonably by utilizing the time dimensionality and the space dimensionality; and the average execution efficiency of the whole task is remarkably improved.

Description

technical field [0001] The present invention relates to the field of information technology, in particular to a dynamic matching algorithm, system and equipment based on characteristics of resources occupied by GPUs. Background technique [0002] In the existing technology, in the deep learning modeling task, the platform user submits the task of training the model to the platform to schedule the task and allocate resources. At this time, the traditional scheduling only considers the amount of resources, and schedules based on the amount of resources And allocation, resulting in low rationality of resource allocation and low average execution efficiency of overall tasks. Contents of the invention [0003] In view of this, the present invention provides a dynamic matching algorithm, system and device based on characteristics of resources occupied by GPUs, which at least partly solves the problems existing in the prior art. [0004] In order to solve the above problems, the...

AI Technical Summary

Problems solved by technology

[0002] In the existing technology, in the deep learning modeling task, the platform user submits the task of training the model to the platform to schedule the task and allocate resources. At this time, the traditional scheduling only considers the amount of resources, and schedules based on the amount of resources and allocation, resulting in low rationality of resource allocation and low average execution efficiency of overall tasks

Images