Method and apparatus for generating operator

An apparatus for generating an operator, comprising:

• at least one processor;
• and
• a memory storing instructions, wherein the instructions when executed by the at least one processor, cause the at least one processor to perform operations, the operations comprising:
• constructing a group of basic application programming interfaces for providing one of following basic functions:
• an access function, a storage function, and a computing function;
• constructing a kernel application programming interface for invoking the group of basic application programming interfaces to implement an operator logic, wherein constructing the group of basic application programming interfaces comprises:
• acquiring a category of a hardware architecture of a target kernel operator of a deep neural network framework, wherein the hardware architecture comprises at least one of a graphics processing unit (GPU) or an embedded neural network processing unit (NPU), and
• constructing the group of basic application programming interfaces adapted to the hardware architecture based on the category of the hardware architecture;
• and
• generating the target kernel operator adapted to the hardware architecture based on the group of basic application programming interfaces adapted to the hardware architecture and the kernel application programming interface.

A non-transitory computer readable storage medium storing computer instructions, wherein the computer instructions are used for causing a computer to execute operations comprising:

• constructing a group of basic application programming interfaces for providing one of following basic functions:
• an access function, a storage function, and a computing function;
• constructing a kernel application programming interface for invoking the basic application programming interfaces to implement an operator logic, wherein constructing the group of basic application programming interfaces comprises;
• acquiring a category of a hardware architecture of a target kernel operator of a deep neural network framework, wherein the hardware architecture comprises at least one of a graphics processing unit (GPU) or an embedded neural network processing unit (NPU), and
• constructing the group of basic application programming interfaces adapted to the hardware architecture based on the category of the hardware architecture;
• and
• generating the target kernel operator adapted to the hardware architecture based on the group of basic application programming interfaces adapted to the hardware architecture and the kernel application programming interface.