Multi-modal data label model training method and device and electronic equipment

The invention provides a multi-modal data label model training method and device and electronic equipment, and the method comprises the steps: constructing a to-be-trained candidate multi-modal data label model, and obtaining a multi-modal training sample; performing feature extraction and resampling on the multi-modal training sample to obtain a multi-modal splicing feature of the multi-modal training sample; obtaining a candidate multi-modal fusion prompt vector and a candidate label prompt vector corresponding to the multi-modal training sample; and obtaining a training loss set of the candidate multi-modal data label model based on the candidate multi-modal fusion prompt vector, the candidate label prompt vector and the multi-modal splicing feature, so as to carry out model training on the candidate multi-modal data label model until the training is finished, and obtaining a trained target multi-modal data label model. The stability of the candidate multi-modal data label model is improved, the noise filt