Distributed machine learning training method, device, electronic equipment and storage medium

Abstract

An embodiment of the present invention provides a distributed machine learning training method, device, electronic equipment, and storage medium, wherein the method includes: determining the machine learning model to be trained, and the communication priority of the data sent to each computing node, said The communication priority of the data sent to each computing node is determined based on the model structure of the machine learning model and / or the processing speed of each computing node running the corresponding training task in the machine learning model; based on the data sent to The data communication priority of each computing node is to transmit the data flow generated by the machine learning model during the training process, so that each computing node can train the machine learning model based on the received data flow. The method, device, electronic equipment, and storage medium provided by the embodiments of the present invention improve the computing speed of machine learning model training and improve the utilization rate of computing resources in a distributed system.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a distributed machine learning training method, device, electronic equipment and storage medium. Background technique [0002] Currently, iterative synchronization applications represented by distributed machine learning training tasks are very popular in data centers. In iterative synchronization applications, multiple computing nodes in a distributed system perform computing tasks iteratively, and perform global synchronization on their respective computing results in each iteration. Only after the global synchronization ends, these computing nodes can start downloading. One round of iterative calculation. [0003] However, as the complexity of applications increases, for example, when the number of parameters of a machine learning model reaches hundreds of millions, the computing speed of distributed systems is slow and the utilization of computing resources is low....

AI Technical Summary

Problems solved by technology

[0006] Embodiments of the present invention provide a distributed machine learning training method, device, electronic equipment, and storage medium to solve the problems of slow computing speed and low utilization rate of computing resources in distributed systems in the prior art

Images