Deep learning compiler test acceleration method based on message passing neural network

The invention belongs to the field of software testing, and relates to a deep learning compiler test acceleration method based on a message passing neural network, which can be used for optimizing the execution sequence of deep learning compiler test cases so as to improve the test efficiency. The method is composed of a data preprocessing module, a data set construction module, a predictor module and a scheduler module. The data preprocessing module is responsible for extracting operator features, edge features and structural features from the deep learning model and converting the operator features, the edge features and the structural features into input of a message passing neural network; the data set construction module generates a large number of deep learning models by using model generation tools such as NNSmith, extracts features of each model by using the data preprocessing module, and judges whether each model triggers an error of a target compiler or not; the predictor is responsible for predicti