Lightweight distributed multi-task collaborative framework

Abstract

The invention discloses a lightweight distributed multi-task cooperation framework, and aims to provide a multi-task cooperation framework which is high in training efficiency and small in iterative computation amount. According to the technical scheme, a distributed neural network training framework is constructed in a multi-machine lightweight mode, a user performs data reading and writing and operation control on the framework through an interface, each object writes description information of the object into an object information recording database, and operation state information of the object is written into an object state recording database; the object information recording database and the object state recording database are in communication and cooperative control with each otherthrough a RESTful interface; and a training task described by the task object is started, a factory object is scheduled to generate a shift captain object, the description information of the task object is read again from the object information recording database, the factory object is scheduled to generate a plurality of worker objects, and the plurality of worker objects generated by the shiftcaptain object are scheduled to complete the training task described by the task object in a distributed cooperation manner.

Description

technical field [0001] The invention belongs to the field of information systems, and in particular relates to a lightweight distributed multi-task collaborative framework in the field of deep learning, in particular to a lightweight distributed multi-task collaborative operation framework involved in the field of distributed training. Background technique [0002] Neural network (Neural Network, NN) is a complex network system formed by a large number of simple processing units (referred to as neurons) through extensive interconnection, reflecting many basic characteristics of human brain functions, and is a highly complex network system. Nonlinear Dynamical Systems. Neural network has the ability of large-scale parallelism, distributed storage and processing, self-organization, self-adaptation and self-learning, etc. It is especially suitable for dealing with practical problems that need to consider many factors and conditions at the same time, and contain imprecise and fu...

AI Technical Summary

Problems solved by technology

Although significant progress has been made in computing chip GPU hardware, network architecture, and training methods in recent years, the fact is that on a single machine, the time required for network training is still unrealistically long

Images