Model training method and device, equipment and storage medium

Abstract

The invention discloses a model training method and device, equipment and a storage medium. The model training method comprises the following steps: performing model training based on a training data set to obtain an intermediate model; based on a preset data acquisition type, acquiring data from the to-be-labeled data set as to-be-tested data, and generating a model reasoning result of the to-be-tested data based on the current intermediate model; based on the dynamic test set and the fixed test set, testing and evaluating the model reasoning result; and if the evaluation result of the fixed test set does not meet the standard condition, adjusting a preset data acquisition type based on the evaluation result of the dynamic test set, performing iterative training on the intermediate model based on the to-be-tested data corresponding to the adjusted preset data acquisition type until the evaluation result of the fixed test set meets the standard condition, and obtaining a target model. Under the driving of small sample annotation data, dynamic model evaluation is carried out, model iteration is rapidly carried out, and then the effect of rapidly generating a model conforming to annotation is achieved.

Description

technical field [0001] Embodiments of the present invention relate to deep learning technology, and in particular to a model training method, device, device and storage medium. Background technique [0002] As an important research direction in the field of machine learning, deep learning is introduced into machine learning to make it closer to the original goal of artificial intelligence. The process of deep learning modeling generally starts from the original data, uses a given multi-layer algorithm, initial parameters and initial weights, conducts data analysis, compares it with the standard results, finds the factors that cause gaps, adjusts relevant parameters and The weights are then subjected to a new round of simulation, and finally the calculation process that minimizes the total tolerance between the calculated results and the actual results. In the practical application of deep learning, if it needs to be integrated into daily business, a basic problem that needs...

AI Technical Summary

Problems solved by technology

[0005] However, the existing model iteration methods based on the attention mechanism or setting the neglect area mainly use the attention mechanism for model iteration. The attention mechanism itself has some shortcomings. For example, the use of the attention mechanism in natural language processing cannot capture the position. information, that is, the order relationship in the sequence cannot be learned

Existing model optimization methods based on parameter merging, this type of method mainly combines the convolution and batch normalization parameters corresponding to the model in the optimization merging method during the optimization process of the deep learning model, which is a model parameter level. The optimization method cannot solve the problem of rapid iteration in a short period of time based on a small amount of data

Existing model optimization methods based on preset ranges mainly use artificially set ranges as the idea of ​​model optimization. Due to the dependence on range setting, there are many inter-class and intra-class problems in the actual model iteration and optimization. Changes will require a large amount of labeled data, the data cannot be used flexibly, and the model cannot be evaluated dynamically

Images