Edge device model training method and device, equipment and medium

The invention provides a model training method and device for edge equipment, equipment and a medium, and the method comprises the steps: selecting a group of optimal clients based on an improved fireworks algorithm, and transmitting model parameters to each client in the group, enabling each client to train a local model corresponding to the client according to the corresponding model parameter; obtaining a trained local model sent by the client and corresponding model parameters thereof, and aggregating the local model and the corresponding parameters thereof to obtain a trained global model; and when it is judged that the first preset iteration termination condition is not met, returning to the client training step. By adopting the embodiment of the invention, the client combination participating in the training can be preferentially selected during each training based on the improved firework algorithm, and the global model of the cloud platform server side is split into a plurality of local models throug