Shared encoder generation method and device and electronic equipment

Abstract

The invention provides a shared encoder generation method and device and electronic equipment, and belongs to the technical field of computers. The method comprises the following steps that: enablinga main node to send a shared encoder training instruction to each sub-node, so that each sub-node acquires a training sample according to the type of a target shared encoder included in the training instruction; after a confirmation message returned by each sub-node is obtained, sending an initial parameter set of a to-be-trained target shared encoder to each sub-node, so that each sub-node trainsthe initial parameter set of the target shared encoder by using respective training samples; obtaining an update parameter set of the target shared encoder returned by each child node; and determining a target parameter set corresponding to the target shared encoder according to a first preset rule and the update parameter set of the target shared encoder returned by each child node. Therefore, through the shared encoder generation method, the acquisition difficulty and cost of the multi-field training corpus are reduced, and the performance of the shared encoder is improved.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to the field of big data technology, and proposes a shared encoder generation method, device and electronic equipment. Background technique [0002] The pre-training model has been proven effective in the fields of computer vision and natural language processing. It has become a popular practice to perform a pre-training under big data and then fine-tune the model parameters according to specific tasks. [0003] However, in related technologies, it is difficult and costly to obtain multi-domain training corpus, which makes it difficult to implement a pre-training model. Contents of the invention [0004] The shared encoder generation method, device and electronic equipment proposed in this application are used to solve the problem in the related art that the acquisition of multi-domain training corpus is difficult and costly, which leads to the difficulty of implementi...

AI Technical Summary

Problems solved by technology

[0004] The shared encoder generation method, device and electronic equipment proposed by this application are used to solve the problem of difficulty and high cost in obtaining multi-field training corpus in related technologies, thus The problem that makes the realization of the pre-training model difficult

Images