Fine tuning method and device for federal learning large model

The invention discloses a federal learning large model fine tuning method and device. The method comprises the steps that a server side divides a pre-trained model into a first model and a second model; the client constructs a local model consistent with the first model in structure, and local model parameters are fixed according to the parameters of the first model; in each round of communication, the client obtains embedded data and encrypts the embedded data according to a differential privacy mechanism to obtain encrypted embedded data; the server side inputs the encrypted embedded data into the second model to obtain output data; the client side calculates model loss according to the output data so as to obtain a corresponding gradient; the server side calculates the gradient corresponding to the parameter in the second model according to the gradient, updates the second model according to the gradient corresponding to the parameter in the second model, and repeats the communication training until the se