Model training method, device, equipment and system and storage medium

The invention provides a model training method. According to the method, a model training stage is divided into a plurality of continuous training stages; in each training stage, both iterative updating of model network parameters and optimization search for specific hyper-parameters are emphasized, so that the model after training in each training stage can have the optimal capability of specifichyper-parameters; after the training of each training stage is finished, the optimal model obtained by the training of the current training stage is automatically used as the initial model of the next training stage; training is carried out in a hierarchical progressive mode; therefore, the optimal model obtained by training in the last training stage integrates the optimal capabilities of all specific hyper-parameters; and the hyper-parameter optimization process is carried out in the normal training process of the model, so that the training time and the training effect can be well evaluated without consuming too m