Method, electronic device, and computer program product for data processing utilizing separable multi-stage data processing model in edge-cloud network
Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for data processing. In a method for data processing, a first electronic device processes data based on a first data processing model to generate an initial result. A data size of the initial result is smaller than a data size of the data. The first electronic device sends the initial result to a second electronic device. The initial result is adjusted at the second electronic device and based on a second data processing model to generate an adjusted result. The second electronic device has more computing resources than the first electronic device, the second data processing model occupies more computing resources than the first data processing model, and an accuracy of the adjusted result is higher than that of the initial result.
1. A method for data processing, comprising:
processing data at a first electronic device and based on a first data processing model to generate an initial result, a data size of the initial result being smaller than a data size of the data, the first electronic device comprising an edge device of an edge portion of an edge-cloud network, the edge device being coupled between a user device and a cloud portion of the edge-cloud network; and
sending the initial result to a second electronic device, the initial result being adjusted at the second electronic device and based on a second data processing model to generate an adjusted result, the second electronic device comprising a cloud device of the cloud portion of the edge-cloud network;
wherein the second electronic device has more computing resources than the first electronic device, the second data processing model occupies more computing resources than the first data processing model, and an accuracy of the adjusted result is higher than that of the initial result; and
wherein the first and second electronic devices are deployed in the respective edge and cloud portions of the edge-cloud network and implement the respective first and second data processing models as respective first and second sequential stages of a separable multi-stage data processing model, the first sequential stage comprising a sensing stage of the separable multi-stage data processing model and the second sequential stage comprising a fine-adjusting stage of the separable multi-stage data processing model.
2. The method according to claim 1 , wherein sending the initial result to the second electronic device comprises:
sending the initial result to the second electronic device if it is determined that a confidence degree of the initial result is less than a threshold.
3. The method according to claim 1 , wherein processing the data based on the first data processing model comprises:
processing the data to generate the initial result in a word form.
4. The method according to claim 1 , further comprising:
generating in the first electronic device at least one signal characterizing content of the data; and
sending the at least one signal from the first electronic device to the second electronic device over a communication link.
5. The method according to claim 1 , further comprising:
receiving the adjusted result from the second electronic device.
6. The method according to claim 1 , further comprising:
receiving a final result from the second electronic device, the final result being generated by the second electronic device based on the adjusted result.
7. The method according to claim 1 , wherein the first data processing model and the second data processing model are obtained by joint training.
8. The method according to claim 1 , wherein the first data processing model comprises a compressed model, and the second data processing model comprises a language model.
9. The method according to claim 1 , wherein the data comprises at least one of an image, a video, and a voice.
10. A computer program product that is tangibly stored on a non-transitory computer-readable medium and comprises machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to execute steps of the method according to claim 1 .
11. A method for data processing, comprising:
receiving, from a first electronic device at a second electronic device, an initial result generated through processing data by the first electronic device based on a first data processing model, a data size of the initial result being smaller than a data size of the data, the first electronic device comprising an edge device of an edge portion of an edge-cloud network, the edge device being coupled between a user device and a cloud portion of the edge-cloud network; and
adjusting the initial result based on a second data processing model at the second electronic device to generate an adjusted result, the second electronic device comprising a cloud device of the cloud portion of the edge-cloud network;
wherein the second electronic device has more computing resources than the first electronic device, the second data processing model occupies more computing resources than the first data processing model, and an accuracy of the adjusted result is higher than that of the initial result; and
wherein the first and second electronic devices are deployed in the respective edge and cloud portions of the edge-cloud network and implement the respective first and second data processing models as respective first and second sequential stages of a separable multi-stage data processing model, the first sequential stage comprising a sensing stage of the separable multi-stage data processing model and the second sequential stage comprising a fine-adjusting stage of the separable multi-stage data processing model.
12. The method according to claim 11 , wherein the initial result is in a word form.
13. The method according to claim 11 , wherein adjusting the initial result based on the second data processing model comprises:
receiving word information for describing content of the data from the first electronic device; and
generating the adjusted result based on the word information and the initial result.
14. The method according to claim 11 , further comprising:
sending the adjusted result to the first electronic device.
15. The method according to claim 11 , further comprising:
processing the adjusted result based on the second data processing model to generate a final result, and sending the final result to the first electronic device.
16. The method according to claim 11 , wherein the first data processing model and the second data processing model are obtained by joint training.
17. The method according to claim 11 , wherein the first data processing model comprises a compressed model, and the second data processing model comprises a language model.
18. A computer program product that is tangibly stored on a non-transitory computer-readable medium and comprises machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to execute steps of the method according to claim 11 .
19. A first electronic device, comprising:
at least one processor; and
at least one memory storing computer program instructions, the at least one memory and the computer program instructions being configured to cause, together with the at least one processor, the first electronic device to:
process data based on a first data processing model to generate an initial result, a data size of the initial result being smaller than a data size of the data, the first electronic device comprising an edge device of an edge portion of an edge-cloud network, the edge device being coupled between a user device and a cloud portion of the edge-cloud network; and
send the initial result to a second electronic device, the initial result being adjusted at the second electronic device and based on a second data processing model to generate an adjusted result, the second electronic device comprising a cloud device of the cloud portion of the edge-cloud network;
wherein the second electronic device has more computing resources than the first electronic device, the second data processing model occupies more computing resources than the first data processing model, and an accuracy of the adjusted result is higher than that of the initial result; and
wherein the first and second electronic devices are deployed in the respective edge and cloud portions of the edge-cloud network and implement the respective first and second data processing models as respective first and second sequential stages of a separable multi-stage data processing model, the first sequential stage comprising a sensing stage of the separable multi-stage data processing model and the second sequential stage comprising a fine-adjusting stage of the separable multi-stage data processing model.