Model training method and related device

The invention discloses a model training method and a related device, which can be applied to various scenes such as cloud technology, artificial intelligence, intelligent traffic, auxiliary driving, vehicle-mounted scenes and the like, such as content (advertisement) recommendation. And obtaining request object data corresponding to the historical recommendation request and a target sorting queue output by the second model. And constructing a training sample data set according to the target content in the target sorting queue, wherein training samples in the training sample data set have sample marks. And performing feature extraction on the request object data and the training sample to sequentially obtain a first feature vector and a second feature vector, obtaining a prediction result through an initial model corresponding to the first model based on the first feature vector and the second feature vector, and performing optimization training on the initial model according to the prediction result and the