WO2021223686A1 - Model training task processing method and apparatus, electronic device, and storage medium - Google Patents

Abstract

A model training task processing method and apparatus, an electronic device, and a storage medium, relating to the technical field of mobile terminals. The method comprises: when a mobile terminal receives a model training task to be processed, identifying a model category corresponding to the model training task (S202); determining a computation burden corresponding to the model training task according to the model category corresponding to the model training task and a preset relationship between the model category and the computation burden (S204); and obtaining a real-time use state of the mobile terminal, and controlling the execution of the model training task according to the real-time use state and the computation burden corresponding to the model training task (S206). In this case, the model training task can be reasonably executed by computation resources of the mobile terminal instead of triggering the execution operation when the model training task is received, so that the training efficiency is improved.

Description

Model training task processing method, device, electronic equipment and storage medium 【Technical Field】

This application relates to the technical field of mobile terminals, and in particular to a method, device, electronic device, and storage medium for processing model training tasks.

【Background technique】

With the development of science and technology, various smart devices are used in daily life, generating massive amounts of behavioral data and information resources. Machine learning emerged at the historic moment, bringing earth-shaking changes to people's production and life. Through machine learning methods, it is possible to mine valuable information from massive amounts of data to better serve users and bring convenience to users.

Machine learning can be roughly divided into three parts: determining model, training model and using model. Among them, training model is an important part of machine learning. Traditional model training generally triggers the execution of the training task when there is a new training task, and the calculation amount of model training is generally large, which leads to low training efficiency of the model training task.

[Summary of the invention]

Based on this, it is necessary to provide a model training task processing method, device, electronic device, and storage medium that can improve training efficiency in response to the above technical problems.

A method for processing model training tasks, the method comprising:

When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

In an embodiment, the acquiring the real-time usage status of the mobile terminal includes:

Acquiring temperature information and real-time running process information of the mobile terminal;

According to the temperature information and the real-time running process information, the real-time use status of the mobile terminal is determined.

In an embodiment, the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount includes:

When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Obtaining provided data corresponding to the real-time usage status of the calculation amount;

Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

Call the same process to execute the executable model training task.

A model training task processing device, the device comprising:

The task recognition module is used to recognize the model category corresponding to the model training task when the mobile terminal receives the to-be-processed model training task;

The calculation amount determination module is configured to determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

The task control module is used to obtain the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

In one embodiment, the task control module is also used to obtain temperature information and real-time operation process information of the mobile terminal; and determine the real-time use status of the mobile terminal according to the temperature information and the real-time operation process information .

In one embodiment, the calculation amount determination module is further configured to obtain the calculation amount level corresponding to the model training task as the highest level when the model category corresponding to the model training task is a neural network model; When the model category corresponding to the training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained; according to the obtained calculation amount level and the preset calculation amount level and calculation amount Relationship to determine the amount of calculation corresponding to the model training task.

In one embodiment, the task control module is further configured to obtain the calculation amount provided data corresponding to the real-time use status; compare the calculation amount provided data with the calculation amount corresponding to the model training task, and filter Obtain the model training target task within the range of the data provided by the calculation amount; call the calculation resource corresponding to the data provided by the calculation amount to execute the model training target task.

In one embodiment, the task control module is also used to call a preset first process, and obtain executable model training tasks according to the real-time usage status and the calculation amount corresponding to the model training tasks; respectively call different The preset second process executes the executable model training task, and the second process and the first process are different processes.

In one embodiment, the task control module is also used to call a preset process, and obtain an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task; call the same process to execute The executable model training task.

An electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, a method for processing a model training task is implemented, and the method includes:

When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

In an embodiment, the acquiring the real-time usage status of the mobile terminal includes:

Acquiring temperature information and real-time running process information of the mobile terminal;

According to the temperature information and the real-time running process information, the real-time use status of the mobile terminal is determined.

In one embodiment, the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount includes:

When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Obtaining provided data corresponding to the real-time usage status of the calculation amount;

Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

Call the same process to execute the executable model training task.

A computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, a method for processing a model training task is implemented, the method comprising:

When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

In an embodiment, the acquiring the real-time usage status of the mobile terminal includes:

Acquiring temperature information and real-time running process information of the mobile terminal;

According to the temperature information and the real-time running process information, the real-time use status of the mobile terminal is determined.

In one embodiment, the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount includes:

When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Obtaining provided data corresponding to the real-time usage status of the calculation amount;

Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.

In an embodiment, the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes:

Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

Call the same process to execute the executable model training task.

The above-mentioned model training task processing method, device, electronic equipment and storage medium, when receiving the model training task to be processed, identify the model category corresponding to the model training task; according to the model category corresponding to the model training task and the preset model category and The relationship between the amount of calculations determines the amount of calculation corresponding to the model training task; obtains the real-time use status of the mobile terminal, and controls the execution of the model training task according to the real-time use status and the amount of calculation corresponding to the model training task, instead of receiving the model training task. The execution operation is triggered, so that the model training task can be executed reasonably by the computing resources of the mobile terminal, thereby improving the training efficiency.

【Explanation of the drawings】

FIG. 1 is an application environment diagram of a method for processing model training tasks in an embodiment of this application;

2 is a schematic flowchart of a method for processing model training tasks in an embodiment of this application;

FIG. 3 is a schematic diagram of a master-slave architecture for processing model training tasks in an embodiment of this application;

FIG. 4 is a schematic flowchart of a step of determining the amount of calculation corresponding to a model training task in an embodiment of the application;

FIG. 5 is a schematic flowchart of execution control steps of a model training task in an embodiment of this application;

Fig. 6 is a structural block diagram of a model training task processing device in an embodiment of the application;

Fig. 7 is an internal structure diagram of an electronic device in an embodiment of the application.

【Detailed ways】

In order to make the purpose, technical solutions, and advantages of this application clearer and clearer, the following further describes the application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

The method for processing model training tasks provided in this application can be applied to the application environment as shown in FIG. 1. When the user uses the mobile terminal, the data required for training the preset model is collected. When the collected data reaches the preset condition, the corresponding model training request is triggered, and the model training request carries the model training task to be processed. When the mobile terminal receives the model training task to be processed, it recognizes the model category corresponding to the model training task; according to the model category corresponding to the model training task and the relationship between the preset model category and the amount of calculation, the calculation amount corresponding to the model training task is determined ; Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task. Among them, the mobile terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, and tablet computers. Specifically, the mobile terminal executing the model training task processing method in the embodiment of the present application may be executed by a processor.

In an embodiment, as shown in FIG. 2, a method for processing a model training task is provided. Taking the method applied to the mobile terminal in FIG. 1 as an example for description, the method includes the following steps:

Step 202: When the mobile terminal receives the to-be-processed model training task, identify the model category corresponding to the model training task.

The model training task refers to the data training process that needs to be completed to realize the model function, and the model category refers to the attribute information of the model, such as whether the model is a neural network model or a non-neural network model. The model training task processing architecture can be built on the processor of the mobile terminal, such as the master-slave architecture, as shown in Figure 3, through the training management module to manage all model training tasks, all model training tasks need to be carried out in the training management module Registration, the model category is provided when the model training task is registered, and different model categories corresponding to the model training task are distinguished by the model category.

Step 204: Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount.

Different model categories correspond to different calculation amounts. According to the preset relationship between model categories and calculation amounts, for example, the calculation amount corresponding to model category A is 1 unit, and the calculation amount corresponding to model category B is 2 units, and so on. According to the model category corresponding to the model training task and the preset correspondence relationship, the calculation amount corresponding to the model training task is obtained.

Step 206: Obtain the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

The real-time usage status of the mobile terminal is used to characterize the usage scenario of the mobile terminal. In different usage scenarios, different mobile terminal usage states are defined. For example, when the screen off time of the mobile terminal exceeds a preset period of time, it is defined that the mobile terminal is in a sleep state at this time. When the mobile terminal is in different usage scenarios, the execution of model training tasks with different calculation amounts is arranged, and the execution of the model training tasks to be processed is controlled according to the calculation power provided by the current situation of the mobile terminal.

In the above-mentioned model training task processing method, when a model training task to be processed is received, the model category corresponding to the model training task is identified; the model training is determined according to the model category corresponding to the model training task and the relationship between the preset model category and the amount of calculation The calculation amount corresponding to the task; obtain the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task, instead of triggering the execution operation when the model training task is received, so that the model training The task can be executed reasonably by the computing resources of the mobile terminal, thereby improving training efficiency.

In one embodiment, acquiring the real-time usage status of the mobile terminal includes: acquiring temperature information and real-time running process information of the mobile terminal; and determining the real-time usage status of the mobile terminal according to the temperature information and real-time running process information. The temperature information of the mobile terminal can be collected through the temperature sensor, and specifically can be the real-time temperature of the processor of the mobile terminal. Real-time process information refers to the working state of the mobile terminal processor, such as whether the processor is in a working state or an idle state, and the load status in the working state. Specifically, as shown in Table 1, five usage states of the mobile terminal and corresponding state explanations are given.

Table 1 Definition of mobile terminal usage status

In one embodiment, as shown in FIG. 4, according to the model category corresponding to the model training task and the relationship between the preset model category and the calculation amount, determining the calculation amount corresponding to the model training task includes: step 402, when the model training task corresponds to When the model category of is a neural network model, the calculation level corresponding to the model training task is obtained as the highest level; step 404, when the model category corresponding to the model training task is a non-neural network model, the model training task is obtained by running the model training task Corresponding operation amount level; Step 406: Determine the operation amount corresponding to the model training task according to the obtained operation amount level and the relationship between the preset operation amount level and the operation amount. When the model training task is registered with the training management module, the model category corresponding to the task is provided, such as distinguishing between xCNN (Convolutional Neural Network) model types and other types. The xCNN model represents a derivative model of the CNN model. The convolutional neural network model is often used in the deep neural network architecture of the image type. It is characterized by a deeper network and relatively large computing resources. For the xCNN model, it is directly classified as an ultra-high computing level. For other types of models, the training management module detects the calculation amount of the model training task when the screen is turned off, and classifies the calculation amount according to the detection result, the specific calculation amount level and the corresponding calculation power consumption unit As shown in table 2. Among them, the computing power consumption unit refers to the computing power required for the mobile terminal to perform a model training task of a certain computing level. It can be seen that the higher the computing level, the more computing power is consumed. Take TOPS (Tera Operations Per Second, which executes 10 to the 12th power of operations per second) as an example of a processor's computing power unit. 1TOPS means that the processor can perform 10 to 12th power of operations per second.

Table 2 The calculation level and the corresponding calculation power consumption unit

Computing level	Computing power consumption unit
Low	1
middle	2
high	4
Super high	8

For model training tasks, SGD (Stochastic Gradient Descent, Stochastic Gradient Descent) is generally used for training, that is, samples of a preset batch size are selected in the training set for each training. In the training process, one iteration means training once with samples of the preset batch size in the training set, and one period means training once with all samples in the training set. For example, the training set corresponding to the model training task includes 1000 samples, and the preset batch size is batchsize=10, then 100 iterations and one period are required to train the entire sample set. As shown in Table 3 below, Table 3 is based on the relationship between the level of computation obtained by a certain model of processor and the iteration time.

Table 3 Operational level and corresponding training time

Computing level	Training time for 1000 iterations
Low	Less than 10 seconds
middle	10 seconds to 1 minute
high	More than 1 minute
Super high	More than 10 minutes

In one embodiment, as shown in FIG. 5, according to the real-time usage status and the calculation amount corresponding to the model training task, controlling the execution of the model training task includes: Step 502: Obtain the calculation amount provided data corresponding to the real-time usage status; Step 504 , Compare the calculation amount provided data with the calculation amount corresponding to the model training task, and filter the model training target tasks within the range of the calculation amount provided data bearing; step 506, call the calculation amount to provide the corresponding calculation resources of the data, and perform model training Target task. The data provided by the amount of computing refers to the computing power that the mobile terminal can provide when it is in a certain use state. Different use states of the mobile terminal correspond to different use scenarios. In different use scenarios, the mobile terminal can provide different computing power. The termination conditions for training tasks are also different. The training management module recognizes the usage scenarios of the mobile terminal, and arranges and runs model training tasks with different computing levels in different usage scenarios. When the conditions corresponding to a certain usage scenario are met, the execution of the model training tasks to be processed is controlled according to the computing power provided by the usage scenario, as shown in Table 4. For example, when the real-time usage status of the mobile terminal satisfies the first usage scenario in the table, at this time, the computing power that the mobile terminal can provide is 8 units. According to Table 2, it can be seen that a model training task with an ultra-high computing level can be performed. , Or perform two high-level model training tasks...and so on.

Table 4

In one embodiment, controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes: invoking the preset first process, and obtaining the available data according to the real-time usage status and the calculation amount corresponding to the model training task. Performed model training tasks; different preset second processes are respectively called to execute executable model training tasks, and the second process and the first process are different processes. The training management module and the execution of multiple model training tasks run in different processes, that is, the control of the model training task and the execution of the model training task are processed through different processes, which is easy to maintain and can improve stability. The model training task processing method in this application is executed by using a master-slave architecture to build a model training task processing structure on a mobile terminal. Under this structure, different model training tasks run in different processes and will not affect each other. Abnormal termination of the training process will not affect other training processes.

In one embodiment, controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task includes: calling a preset process, and obtaining the executable according to the real-time usage status and the calculation amount corresponding to the model training task Model training tasks; call the same process to execute executable model training tasks. The training management module and the execution of multiple model training tasks are integrated and run in the same process, and different model training tasks are transformed into multiple different threads for execution, providing another way to implement the model training task processing in this application Methods to meet diverse needs.

It should be understood that although the steps in the flowcharts of FIGS. 2 and 4-5 are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless specifically stated in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least part of the steps in Figures 2, 4-5 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. These sub-steps Or the execution order of the stages is not necessarily carried out sequentially, but may be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

In one embodiment, as shown in FIG. 6, a model training task processing device is provided. The model training task processing device includes a task identification module 602, a calculation amount determination module 604, and a task control module 606. The task recognition module 602 is configured to recognize the model category corresponding to the model training task when the mobile terminal receives the model training task to be processed. The calculation amount determination module 604 is configured to determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount. The task control module 606 is used to obtain the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.

In one embodiment, the task control module 606 is also used to obtain temperature information and real-time running process information of the mobile terminal; determine the real-time use status of the mobile terminal according to the temperature information and real-time running process information.

In one embodiment, the calculation amount determining module 604 is also used to obtain the calculation amount level corresponding to the model training task as the highest level when the model category corresponding to the model training task is a neural network model; when the model category corresponding to the model training task is In the case of non-neural network models, by running the model training task, the calculation amount level corresponding to the model training task is obtained; the calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the relationship between the preset calculation amount level and the calculation amount.

In one embodiment, the task control module 606 is also used to obtain the calculation amount provided data corresponding to the real-time usage status; compare the calculation amount provided data with the calculation amount corresponding to the model training task, and filter to obtain the calculation amount provided data bearing The model training target tasks within the scope; the calculation amount is called to provide the computing resources corresponding to the data, and the model training target tasks are performed.

In one embodiment, the task control module 606 is also used to call the preset first process, obtain executable model training tasks according to the real-time usage status and the calculation amount corresponding to the model training tasks; respectively call different preset first processes. The second process executes executable model training tasks, and the second process and the first process are different processes.

In one embodiment, the task control module 606 is also used to call a preset process to obtain an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task; call the same process to perform executable model training Task.

For the specific limitation of the model training task processing device, please refer to the above limitation of the model training task processing method, which will not be repeated here. Each module in the above-mentioned model training task processing device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

In one embodiment, an electronic device is provided. The electronic device may be a mobile terminal. FIG. 7 provides an internal structure diagram of the mobile terminal. The mobile terminal includes a processor, a memory, and a display screen connected through a system bus. Among them, the processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The method for processing model training tasks performed by the mobile terminal in FIG. 1 may be specifically completed by the processor of the mobile terminal, that is, a method for processing model training tasks is implemented when a computer program is executed by the processor.

Those skilled in the art can understand that the structure shown in FIG. 7 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the terminal to which the solution of the present application is applied. The specific mobile terminal may include More or fewer parts than shown in the figure, or some parts are combined, or have a different arrangement of parts.

In one embodiment, an electronic device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program: when a mobile terminal receives a model training task to be processed, Identify the model category corresponding to the model training task; determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the relationship between the preset model category and the amount of calculation; obtain the real-time usage status of the mobile terminal according to the real-time usage status And the amount of calculation corresponding to the model training task, to control the execution of the model training task.

In one embodiment, the processor further implements the following steps when executing the computer program: acquiring temperature information and real-time running process information of the mobile terminal; and determining the real-time usage status of the mobile terminal according to the temperature information and real-time running process information.

In one embodiment, the processor further implements the following steps when executing the computer program: when the model category corresponding to the model training task is a neural network model, the calculation level corresponding to the model training task is the highest level; when the model training task corresponds to When the model type is a non-neural network model, run the model training task to obtain the calculation level corresponding to the model training task; according to the obtained calculation level and the relationship between the preset calculation level and the calculation volume, determine the corresponding model training task Computation.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the computing amount provided data corresponding to the real-time use status; comparing the computing amount provided data with the computing amount corresponding to the model training task, and filtering to obtain the The volume provides the target tasks of model training within the scope of data bearing; the call volume provides the computing resources corresponding to the data, and the target tasks of model training are performed.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: call the preset first process, obtain executable model training tasks according to the real-time usage status and the calculation amount corresponding to the model training tasks; respectively call different The preset second process executes executable model training tasks, and the second process and the first process are different processes.

In one embodiment, the processor further implements the following steps when executing the computer program: calling a preset process, obtaining an executable model training task according to the real-time usage status and the amount of calculation corresponding to the model training task; calling the same process to execute the executable The model training task performed.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are also implemented: when the mobile terminal receives a model training task to be processed, the model is recognized The model category corresponding to the training task; the calculation amount corresponding to the model training task is determined according to the model category corresponding to the model training task and the relationship between the preset model category and the amount of calculation; the real-time usage status of the mobile terminal is obtained, based on the real-time usage status and model The calculation amount corresponding to the training task controls the execution of the model training task.

In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: acquiring temperature information and real-time running process information of the mobile terminal; and determining the real-time use status of the mobile terminal according to the temperature information and real-time running process information.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the model category corresponding to the model training task is a neural network model, the calculation level corresponding to the model training task is the highest level; when the model training task corresponds to When the model type of is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained; according to the obtained calculation amount level and the relationship between the preset calculation amount level and the calculation amount, the corresponding model training task is determined The amount of computing.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtain the computing amount provided data corresponding to the real-time use status; The amount of calculation provides the target tasks of model training within the range of data bearing; the amount of calculation is called to provide the calculation resources corresponding to the data, and the target tasks of model training are performed.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: call the preset first process, obtain executable model training tasks according to the real-time usage status and the calculation amount corresponding to the model training tasks; respectively call different The preset second process executes executable model training tasks, and the second process and the first process are different processes.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: call a preset process, obtain an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task; call the same process to execute Executable model training tasks.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage. In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered as the range described in this specification.

The above-mentioned embodiments only express several implementation manners of the present application, and their description is relatively specific and detailed, but they should not be interpreted as a limitation on the scope of the patent application. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can be made, and these all fall within the protection scope of this application. Therefore, the scope of protection of the patent of this application shall be subject to the appended claims.

Claims (24)

1. A method for processing model training tasks, the method comprising:

   When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

   Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

   Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.
2. The method according to claim 1, wherein said obtaining the real-time usage status of the mobile terminal comprises:

   Acquiring temperature information and real-time running process information of the mobile terminal;

   According to the temperature information and the real-time running process information, the real-time use status of the mobile terminal is determined.
3. The method according to claim 1, wherein the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount comprises:

   When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

   When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

   The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.
4. The method according to claim 1, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

   Obtaining provided data corresponding to the real-time usage status of the calculation amount;

   Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

   Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.
5. The method according to claim 1, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

   Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

   Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.
6. The method according to claim 1, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

   Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

   Call the same process to execute the executable model training task.
7. A model training task processing device, the device comprising:

   The task recognition module is used to recognize the model category corresponding to the model training task when the mobile terminal receives the to-be-processed model training task;

   The calculation amount determination module is configured to determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

   The task control module is used to obtain the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.
8. 7. The device according to claim 7, wherein the task control module is further configured to obtain temperature information and real-time running process information of the mobile terminal; determine the mobile terminal according to the temperature information and the real-time running process information The real-time usage status of the terminal.
9. 8. The device according to claim 7, wherein the calculation amount determining module is further configured to obtain that the calculation amount level corresponding to the model training task is the highest level when the model category corresponding to the model training task is a neural network model ; When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained; according to the obtained calculation amount level and the preset calculation amount The relationship between the level and the amount of calculation determines the amount of calculation corresponding to the model training task.
10. 7. The device according to claim 7, wherein the task control module is further configured to obtain the calculation amount provided data corresponding to the real-time use status; The model training target tasks within the data bearing range provided by the calculation amount are filtered to obtain the model training target tasks; the calculation resources corresponding to the data provided by the calculation amount are called to execute the model training target tasks.
11. 8. The device according to claim 7, wherein the task control module is further configured to call a preset first process, and obtain executable model training according to the real-time usage status and the calculation amount corresponding to the model training task Tasks; respectively calling different preset second processes to execute the executable model training task, and the second process and the first process are different processes.
12. 8. The device according to claim 7, wherein the task control module is further configured to call a preset process, and obtain an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task; Call the same process to execute the executable model training task.
13. An electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, a method for processing a model training task is implemented, and the method includes:

    When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

    Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

    Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.
14. The electronic device according to claim 13, wherein said acquiring the real-time usage status of the mobile terminal comprises:

    Acquiring temperature information and real-time running process information of the mobile terminal;

    According to the temperature information and the real-time running process information, the real-time usage status of the mobile terminal is determined.
15. The electronic device according to claim 13, wherein the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount comprises:

    When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

    When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

    The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.
16. The electronic device according to claim 13, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Obtaining provided data corresponding to the real-time usage status of the calculation amount;

    Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

    Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.
17. The electronic device according to claim 13, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

    Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.
18. The electronic device according to claim 13, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

    Call the same process to execute the executable model training task.
19. A computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, a method for processing a model training task is implemented, the method comprising:

    When the mobile terminal receives the model training task to be processed, identifying the model category corresponding to the model training task;

    Determine the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the preset relationship between the model category and the calculation amount;

    Acquire the real-time usage status of the mobile terminal, and control the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task.
20. The computer-readable storage medium according to claim 19, wherein said obtaining the real-time usage status of the mobile terminal comprises:

    Acquiring temperature information and real-time running process information of the mobile terminal;

    According to the temperature information and the real-time running process information, the real-time use status of the mobile terminal is determined.
21. The computer-readable storage medium according to claim 19, wherein the determining the calculation amount corresponding to the model training task according to the model category corresponding to the model training task and the relationship between the preset model category and the calculation amount comprises :

    When the model category corresponding to the model training task is a neural network model, the calculation amount level corresponding to the model training task is obtained as the highest level;

    When the model category corresponding to the model training task is a non-neural network model, by running the model training task, the calculation amount level corresponding to the model training task is obtained;

    The calculation amount corresponding to the model training task is determined according to the obtained calculation amount level and the preset relationship between the calculation amount level and the calculation amount.
22. The computer-readable storage medium according to claim 19, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Obtaining provided data corresponding to the real-time usage status of the calculation amount;

    Comparing the data provided by the calculation amount with the calculation amount corresponding to the model training task respectively, and filter to obtain the model training target tasks within the range of the data provided by the calculation amount;

    Invoking the computing resources to provide computing resources corresponding to the data to execute the model training target task.
23. The computer-readable storage medium according to claim 19, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Calling a preset first process, and obtaining an executable model training task according to the real-time usage status and the calculation amount corresponding to the model training task;

    Different preset second processes are respectively called to execute the executable model training task, and the second process and the first process are different processes.
24. The computer-readable storage medium according to claim 19, wherein the controlling the execution of the model training task according to the real-time usage status and the calculation amount corresponding to the model training task comprises:

    Calling a preset process, and obtaining an executable model training task according to the real-time use status and the calculation amount corresponding to the model training task;

    Call the same process to execute the executable model training task.

Images