Methods and devices for power management based on synthetic machine learning benchmarks
A method for power management based on synthetic machine learning benchmarks, including generating a record of synthetic machine learning benchmarks for synthetic machine learning models that are obtained by changing machine learning network topology parameters, receiving hardware information from a client device executing a machine learning program or preparing to execute a machine learning program, selecting a synthetic machine learning benchmark based on the correlation of the hardware information with the synthetic machine learning models, and determining work schedules based on the selected synthetic machine learning benchmark.
1. A method for power management based on synthetic machine learning benchmarks, comprising:
generating a record of synthetic machine learning benchmarks for synthetic machine learning models that are obtained by changing machine learning network topology parameters, wherein generating the record of synthetic machine learning benchmarks for synthetic machine learning models comprises:
generating synthetic machine learning models based on controlling input size, filter size, number of filters, stride of filters, and number of layers of models, and
obtaining performance data for each generated synthetic machine learning model based on each of the generated synthetic machine learning models and a power consumption associated with each of the generated synthetic machine learning network models;
receiving hardware information from a client device executing a machine learning program or preparing to execute a machine learning program;
selecting a synthetic machine learning benchmark based on the hardware information and the synthetic machine learning models; and
determining a work schedule for the client device based on the selected synthetic machine learning benchmark.
2. The method of claim 1 , wherein selecting a synthetic machine learning benchmark based on the hardware information and the synthetic machine learning models comprises analyzing the hardware information received from the client device and correlating the hardware information with the synthetic machine learning models based on the analysis of the hardware information.
3. The method of claim 2 , wherein analyzing hardware information received from the client device comprises predicting a machine learning model based on the hardware information.
4. The method of claim 2 , wherein correlating hardware information with the synthetic machine learning models comprises comparing the results of the analysis of the hardware information with the generated synthetic machine learning models.
5. The method of claim 1 , wherein the received hardware information comprises at least one of:
a power consumption measurement of the client device;
an executable instruction measurement of the client device;
layer information of the machine learning program being executed on the client device;
power limits of components of the client device; or
a memory usage measurement of the client device.
6. The method of claim 1 , wherein selecting the synthetic machine learning benchmark comprises choosing the machine learning benchmark from the record that corresponds with the machine learning model that correlates best with results of the analysis of the hardware information.
7. A device for power management based on synthetic machine learning benchmarks, comprising:
a memory configured to store a set of instructions; and
a processor configured to execute the set of instructions to cause the device to:
generate a record of synthetic machine learning benchmarks for synthetic machine learning models that are obtained by changing machine learning network topology parameters, wherein generation of the record of synthetic machine learning benchmarks for synthetic machine learning models comprises:
generate synthetic machine learning models based on controlling input size, filter size, number of filters, stride of filters, and number of layers of models, and
obtain performance data for each generated synthetic machine learning model based on each of the generated synthetic machine learning models and a power consumption associated with each of the generated synthetic machine learning network models;
receive hardware information from a client device executing a machine learning program or preparing to execute a machine learning program;
select a synthetic machine learning benchmark based on the correlation of the hardware information with the synthetic machine learning models; and
determine work schedules based on the selected synthetic machine learning benchmark.
8. The device of claim 7 , wherein selecting a synthetic machine learning benchmark based on the hardware information and the synthetic machine learning models comprises analyzing the hardware information received from the client device and correlating the hardware information with the synthetic machine learning models based on the analysis of the hardware information.
9. The device of claim 8 , wherein analyzing hardware information received from the client device comprises predicting a machine learning model based on the hardware information.
10. The device of claim 8 , wherein correlating hardware information with the synthetic machine learning models comprises comparing the results of the analysis of the hardware information with the generated synthetic machine learning models.
11. The device of claim 7 , wherein the received hardware information comprises at least one of:
a power consumption measurement of the client device;
an executable instruction measurement of the client device;
layer information of the machine learning program being executed on the client device;
power limits of components of the client device; and/or
a memory usage measurement of the client device.
12. The device of claim 7 , wherein selecting a synthetic machine learning benchmark comprises choosing the machine learning benchmark from the record that corresponds with the machine learning model that correlates best with results of the analysis of the hardware information.
13. A non-transitory computer-readable medium that stores a set of computer executable instructions that are executable by at least one processor of a device for power management to cause the device to perform a method for power management based on synthetic machine learning benchmarks, the method comprising:
generating a record of synthetic machine learning benchmarks for synthetic machine learning models that are obtained by changing machine learning network topology parameters, wherein generating the record of synthetic machine learning benchmarks for synthetic machine learning models comprises:
generating synthetic machine learning models based on controlling input size, filter size, number of filters, stride of filters, and number of layers of models, and
obtaining performance data for each generated synthetic machine learning model based on each of the generated synthetic machine learning models and a power consumption associated with each of the generated synthetic machine learning network models;
receiving hardware information from a client device executing a machine learning program or preparing to execute a machine learning program;
selecting a synthetic machine learning benchmark based on the correlation of the hardware information with the synthetic machine learning models; and
determining work schedules based on the selected synthetic machine learning benchmark.
14. The computer readable medium according to claim 13 , wherein selecting a synthetic machine learning benchmark based on the hardware information and the synthetic machine learning models comprises analyzing the hardware information received from the client device and correlating the hardware information with the synthetic machine learning models based on the analysis of the hardware information.
15. The computer readable medium according to claim 14 , wherein analyzing hardware information received from the client device comprises predicting a machine learning model based on the hardware information.
16. The computer readable medium according to claim 14 , wherein correlating hardware information with the one or more synthetic machine learning models comprises comparing the results of the analysis of the hardware information with the generated machine learning models.
17. The computer readable medium according to claim 13 , wherein the received hardware information comprises at least one of:
a power consumption measurement of the client device;
an executable instruction measurement of the client device;
layer information of the machine learning program being executed on the client device;
power limits of components of the client device; and/or
a memory usage measurement of the client device.
18. The computer readable medium according to claim 13 , wherein selecting a synthetic machine learning benchmark comprises choosing the machine learning benchmark from the record that corresponds with the machine learning model that correlates best with results of the analysis of the hardware information.