Neural network model deployment method, prediction method and related device

A neural network model prediction method, applied to a terminal device, the method comprising:

• obtaining prediction input data through a data input apparatus of the terminal device;
• invoking a neural network model pre-deployed in the terminal device;
• and
• using the prediction input data as an input of the neural network model, and processing the prediction input data by using the neural network model to obtain a prediction result.

A neural network model deployment apparatus, applied to a terminal device, the apparatus comprising:

• a reading module configured to:
• read an initial neural network model;
• and
• obtain a layer definition of each network layer of the initial neural network model;
• obtain an operation parameter of each network layer of the initial neural network model;
• a target network layer implementation module configured to:
• execute, by using a layer class, a target network layer corresponding to each network layer in the terminal device separately according to the layer definition of each network layer, so that each target network layer is inherited from the layer class;
• a network layer connection module configured to:
• apply relational connections amongst the target network layers using a net class;
• a format conversion module configured to:
• convert the operation parameter of each network layer into a predetermined format, to obtain a target operation parameter of each network layer;
• and
• a parameter loading module configured to:
• load a corresponding target operation parameter in the target network layer corresponding to each network layer separately according to the target operation parameter of each network layer;
• and
• obtain a target neural network model deployed in the terminal device based on the target operation parameter of each network layer.

A non-transitory computer-readable medium configured to store instructions that, when executed, causes a processing circuitry to:

• read an initial neural network model;
• and
• obtain a layer definition of each network layer of the initial neural network model;
• obtain an operation parameter of each network layer of the initial neural network model;
• execute, by using a layer class, a target network layer corresponding to each network layer in a terminal device separately according to the layer definition of each network layer, so that each target network layer is inherited from the layer class;
• apply relational connections amongst the target network layers using a net class;
• convert the operation parameter of each network layer into a predetermined format, to obtain a target operation parameter of each network layer;
• load a corresponding target operation parameter in the target network layer corresponding to each network layer separately according to the target operation parameter of each network layer;
• and
• obtain a target neural network model deployed in the terminal device based on the target operation parameter of each network layer.