Task training method and device, equipment and storage medium

Abstract

The invention discloses a task training method and device, equipment and a storage medium, and the method comprises the steps: carrying out the virtualization processing of a plurality of physical network cards of an artificial intelligence server, so as to obtain a plurality of virtual network cards; wherein a virtual relationship exists between each physical network card and the corresponding virtual network card; all the virtual network cards are grouped, so that different resource groups are obtained; wherein the virtual relationship between the virtual network card and the physical network card in each resource group is not repeated; and providing network resources for a target task on the artificial intelligence server by using the virtual network card in a target resource group so as to train the target task. In the training process, when any network card in the resource group is abnormal, other network cards without abnormity in the resource group can be utilized to continuously provide network resources, so that high-performance network cards are reasonably allocated to the training task, high-performance network card multiplexing is realized, and the task training efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a task training method, device, equipment and storage medium. Background technique [0002] In artificial intelligence (AI, Artificial Intelligence) training scenarios, there will be a large amount of data transmission, which is limited by the traditional operating system's mechanism for receiving and sending network data packets, which cannot meet the training tasks of downloading large data sets and transferring large amounts of training parameters. For this reason, a large number of AI training platforms have gradually begun to support high-performance networks, such as Infiniband and Roce. In the prior art, when an AI server has multiple GPU cards, in order to obtain the best training effect, sometimes a single AI server is equipped with 2, 4, or 8 ROCE network cards or Infiniband network cards. For traditional CPU servers, in order to ensure the stability an...

AI Technical Summary

Problems solved by technology

However, for the Roce network card and Infiniband network card in the AI ​​server, due to the situation that the bond cannot be performed and the resources are wasted, two high-performance network cards are generally not bonded. When the high-performance network card is abnormal, the normal operation of the training task cannot be guaranteed.

Images