Model design method, deployment method and device for microcontroller

The invention provides a model design method, deployment method and device for a microcontroller, and belongs to the technical field of deep learning, and the model design method of one embodiment comprises the steps: carrying out the super-network search of a pre-established super-network structure pool, and determining an optimal super-network structure in the super-network structure pool; performing single-path search on the optimal super-network structure, and determining an optimal single-path model of the optimal super-network structure; performing training of a first preset period number on the optimal single-path model to obtain a trained optimal single-path model; and carrying out iteration pruning operation and quantization operation of a first preset number of times on the trained single-path model to generate an optimal sparse single-path model. According to the technical scheme of the invention, the sparse model is introduced into the field of the microcontroller, so that the deployment of a mode