Method for managing artificial intelligence application, device, and program product
The present disclosure relates to a method for managing an artificial intelligence (AI) application, a device, and a program product. One method comprises receiving input data to be processed by the AI application; updating a first version of the AI application with the input data to generate a second version, wherein the first version is deployed at server equipment; compressing the second version of the AI application to generate a third version of the AI application; and deploying the third version of the AI application to terminal equipment to replace a fourth version of the AI application deployed at the terminal equipment, wherein the fourth version of the AI application is used for processing the input data received at the terminal equipment. A device and a computer program product corresponding thereto are provided.
1. A method for managing an artificial intelligence (AI) application, comprising:
collecting, at terminal equipment, input data to be processed by the AI application;
updating a first version of the AI application with the input data to generate a second version of the AI application, the first version of the AI application being deployed at server equipment associated with the terminal equipment;
compressing the second version of the AI application to generate a third version of the AI application; and
deploying the third version of the AI application to the terminal equipment to replace a fourth version of the AI application at the terminal equipment;
wherein updating the first version of the AI application with the input data comprises:
in response at least in part to a detected change of the input data collected at the terminal equipment, relative to previous input data utilized in previous training of the AI application, being greater than a predetermined threshold, updating the first version of the AI application by training the first version of the AI application, utilizing at least portions of the input data, to generate the second version of the AI application; and
wherein at least the first version of the AI application is a full version of the AI application deployed at the server equipment and at least the third version of the AI application deployed at the terminal equipment is a reduced version of the AI application, relative to the full version of the AI application deployed at the server equipment.
2. The method according to claim 1 , wherein updating the first version of the AI application with the input data to generate the second version of the AI application, is further performed in response to at least in part any of the following:
reaching a predetermined time interval; and
an update request for updating the AI application.
3. The method according to claim 1 , wherein the generating of the second version of the AI application comprises: training the first version of the AI application with the input data to acquire the second version of the AI application.
4. The method according to claim 1 , wherein the generating of the third version of the AI application comprises at least any of the following:
deleting redundant parameters in the second version of the AI application;
cutting out redundant branches in the second version of the AI application; and
reducing precision of parameters in the second version of the AI application.
5. The method according to claim 1 , further comprising:
deploying the third version of the AI application at the server equipment;
processing the input data with the third version of the AI application to acquire output data; and
determining performance of the fourth version of the AI application based on the output data.
6. The method according to claim 1 , further comprising:
processing the input data with the second version of the AI application to acquire output data; and
sending the output data to the terminal equipment.
7. The method according to claim 1 , wherein the method is implemented at the server equipment.
8. The method according to claim 1 , wherein the second version of the AI application is a full version of the AI application, and the fourth version of the AI application is a reduced version of the AI application relative to the full version of the AI application deployed at the server equipment.
9. The method according to claim 1 , comprising:
processing the input data with the fourth version of the AI application at the terminal equipment in response to the detected change of the input data being lower than the predetermined threshold.
10. The method according to claim 1 , wherein the collecting of the input data is performed by one or more sensors at the terminal equipment.
11. A method for managing an AI application, comprising:
collecting, at terminal equipment, input data;
transmitting the input data to server equipment deployed with a first version of the AI application to update the first version of the AI application to a second version of the AI application with the input data, the server equipment associated with the terminal equipment;
receiving a third version of the AI application from the server equipment, the third version of the AI application being acquired based on compressing the second version of the AI application; and
deploying the third version of the AI application to the terminal equipment to replace a fourth version of the AI application at the terminal equipment;
wherein transmitting the input data to the server equipment comprises:
in response at least in part to a detected change of the input data collected at the terminal equipment, relative to previous input data utilized in previous training of the AI application being greater than a predetermined threshold, transmitting the input data to the server equipment to allow the server equipment to update the first version of the AI application by training the first version of the AI application, utilizing at least portions of the input data, to generate the second version of the AI application; and
wherein at least the first version of the AI application is a full version of the AI application deployed at the server equipment and at least the third version of the AI application deployed at the terminal equipment is a reduced version of the AI application, relative to the full version of the AI application deployed at the server equipment.
12. The method according to claim 11 , further comprising:
in response to a detected change of the input data collected at the terminal equipment, relative to previous input data utilized in previous training of the AI application, being lower than the predetermined threshold, processing the input data with the fourth version of the AI application.
13. The method according to claim 12 , wherein the transmitting of the input data to the server equipment comprises: transmitting the input data to the server equipment in response to at least in part any of the following:
reaching a predetermined time interval; and
an update request for updating the AI application.
14. The method according to claim 11 , further comprising: receiving output data from the server equipment, the output data being acquired by processing the input data with the second version of the AI application.
15. The method according to claim 11 , wherein the method is implemented at the terminal equipment.
16. The method according to claim 11 , wherein the second version of the AI application is a full version of the AI application, and the fourth version of the AI application is a reduced version of the AI application relative to the full version of the AI application deployed at the server equipment.
17. The method according to claim 11 , wherein the collecting of the input data is performed by one or more sensors at the terminal equipment.
18. An electronic device, comprising:
at least one processor;
a volatile memory; and
a memory coupled with the at least one processor, the memory having instructions stored therein, and the instructions, when executed by the at least one processor, causing the electronic device to execute a method for managing an artificial intelligence (AI) application, the method comprising:
collecting, at terminal equipment, input data to be processed by the AI application;
updating a first version of the AI application with the input data to generate a second version of the AI application, the first version of the AI application being deployed at server equipment associated with the terminal equipment;
compressing the second version of the AI application to generate a third version of the AI application; and
deploying the third version of the AI application to the terminal equipment to replace a fourth version of the AI application at the terminal equipment;
wherein updating the first version of the AI application with the input data comprises:
in response at least in part to a detected change of the input data collected at the terminal equipment, relative to previous input data utilized in previous training of the AI application, being greater than a predetermined threshold, updating the first version of the AI application by training the first version of the AI application, utilizing at least portions of the input data, to generate the second version of the AI application; and
wherein at least the first version of the AI application is a full version of the AI application deployed at the server equipment and at least the third version of the AI application deployed at the terminal equipment is a reduced version of the AI application, relative to the full version of the AI application deployed at the server equipment.
19. The electronic device according to claim 18 , wherein updating the first version of the AI application with the input data to generate the second version of the AI application, is further performed in response to at least in part any of the following:
reaching a predetermined time interval; and
an update request for updating the AI application.
20. The electronic device according to claim 18 , wherein the generating of the second version of the AI application comprises: training the first version of the AI application with the input data to acquire the second version of the AI application.