Method for compiling neutral network model electronic device and storage medium

) A non-transitory computer readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, causes the processor to implement a method for compiling a neural network model, the method comprising:

• acquiring a neural network model to be compiled:
• determining, based on the neural network model to be compiled, an initial computation graph corresponding to the neural network model to be compiled, wherein the initial computation graph comprises one or more first computation sub-graphs each including one or more operator nodes and connection relationships between the respective operator nodes;
• determining, from the respective first computation sub-graphs, a second computation sub-graph including a preset operator node, and partitioning, based on an input tensor for the second computation sub-graph, the second computation sub-graph into a first number of target computation sub-graphs;
• establishing, based on a preset dependency rule, dependency relationships between the first number of target computation sub-graphs, to obtain a first dependency graph corresponding to the second computation sub-graph, the dependency relationship representing a computation order among the target computation sub-graphs;
• and
• generating, based on the first dependency graphs corresponding respectively to the respective second computation sub-graphs, a target instruction sequence corresponding to the neural network model to be compiled.

An electronic device, comprising:

• a processor;
• and
• a memory, configured to store processor-executable instructions, wherein
• the processor is configured to read the executable instructions from the memory, and execute the instructions to implement a method for compiling a neural network model, the method comprising:
• acquiring a neural network model to be compiled;
• determining, based on the neural network model to be compiled, an initial computation graph corresponding to the neural network model to be compiled, wherein the initial computation graph comprises one or more first computation sub-graphs each including one or more operator nodes and connection relationships between the respective operator nodes;
• determining, from the respective first computation sub-graphs, a second computation sub-graph including a preset operator node, and partitioning, based on an input tensor for the second computation sub-graph, the second computation sub-graph into a first number of target computation sub-graphs;
• establishing, based on a preset dependency rule, dependency relationships between the first number of target computation sub-graphs, to obtain a first dependency graph corresponding to the second computation sub-graph, the dependency relationship representing a computation order among the target computation sub-graphs;
• and
• generating, based on the first dependency graphs corresponding respectively to the respective second computation sub-graphs, a target instruction sequence corresponding to the neural network model to be compiled.