Resource allocation method and device for model training and electronic equipment

By selecting wireless communication modules and constructing deep learning models based on task type and complexity, the problem of model training being unable to adapt to different tasks in existing technologies is solved, thereby improving the reliability and adaptability of task processing.

CN119045992APending Publication Date: 2024-11-29QINGDAO HAIER TECH +3
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411030542.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

In existing technologies, training deep learning models based on the same LoRa module strategy cannot adapt to different tasks, affecting the reliability of task processing.

Method used

Based on the type and complexity of the task to be processed, the target wireless communication modules of the target type and their activation number are determined from the wireless communication modules. A deep learning model is constructed, and the model's adaptability to the task is improved through model training and fine-tuning optimization.

Benefits of technology

It improves the reliability of task processing and the adaptability of deep learning, ensures that the model architecture is adapted to the task, and improves the accuracy of model training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119045992A_ABST
    Figure CN119045992A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of deep learning, and discloses a resource allocation method for model training, which comprises the following steps: receiving a to-be-processed task, and determining the type of the to-be-processed task; according to the type of the to-be-processed task, determining target wireless communication modules with a target type from the wireless communication modules and a target activation number of each target wireless communication module; according to the target wireless communication module with the target type and the target activation number of each target wireless communication module, constructing a deep learning model corresponding to the to-be-processed task; and carrying out model training on the deep learning model to obtain a model evaluation result. The method can improve the reliability of task processing and the adaptive ability of deep learning. The invention further discloses a resource configuration device for model training and electronic equipment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of deep learning technology, for example, to a resource configuration method and device, and electronic equipment for model training. Background Art

[0002] At present, deep learning is a computational model that simulates the mechanism of biological neural networks (especially the visual cortex) to automatically abstract the data to be processed in layers. Deep learning has achieved many results in other related fields such as search technology, data mining, machine learning, natural language processing, etc. It has solved many complex pattern recognition problems and made great progress in artificial intelligence technology. With the rapid development of artificial intelligence technology, how to build and train neural networks in the task processing process has become a technical problem that needs to be solved urgently.

[0003] In order to solve the above technical problems, the related technology sets the same Lora module strategy; when receiving different tasks to be processed, based on the Lora module strategy, the set parameters and the current system performance are used as input and the deep learning model is globally or locally optimized and trained based on the above input.

[0004] In the process of implementing the embodiments of the present disclosure, it is found that there are at least the following problems in the related art:

[0005] Based on different tasks to be processed, the related technology uses the same Lora module strategy to train the deep learning model. However, the Lora module strategies adapted to different tasks to be processed are not all the same. The model training method of the related technology cannot be adapted to each task to be processed, affecting the reliability of task processing.

[0006] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present application, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0007] In order to provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. The summary is not an extensive review, nor is it intended to identify key / critical components or delineate the scope of protection of these embodiments, but rather serves as a prelude to the detailed description that follows.

[0008] The embodiments of the present disclosure provide a resource configuration method and device, and an electronic device for model training to improve the reliability of task processing and the adaptability of deep learning.

[0009] In some embodiments, the method includes: receiving a task to be processed and determining the type of the task to be processed; determining a target wireless communication module with a target type and a target activation number of each target wireless communication module from the wireless communication module according to the type of the task to be processed; constructing a deep learning model corresponding to the task to be processed according to the target wireless communication module with the target type and the target activation number of each target wireless communication module; and performing model training on the deep learning model to obtain a model evaluation result.

[0010] In some embodiments, according to the type of task to be processed, target wireless communication modules with target types and the target activation number of each target wireless communication module are determined from the wireless communication modules, including: determining the complexity of the task to be processed according to the type of task to be processed; determining the target wireless communication modules with target types corresponding to the subtasks included in the task to be processed; obtaining the priority of each wireless communication module; and determining the target activation number of each target wireless communication module according to the complexity of the task to be processed and the priority of each wireless communication module.

[0011] Optionally, determining the complexity of the task to be processed according to the type of the task to be processed includes: determining the score of the task to be processed according to the information processing amount and data feature complexity of the task to be processed, and determining the complexity of the task to be processed according to the score, wherein the score is positively correlated with the complexity.

[0012] Optionally, the target activation number of each target wireless communication module is determined according to the complexity of the task to be processed and the priority of each wireless communication module, including: selecting a target wireless communication module from the wireless communication modules according to the complexity of the task to be processed and the priority of each wireless communication module; determining the target activation number of each wireless communication module according to the complexity of the task to be processed; wherein, the higher the complexity, the higher the priority of the selected target wireless communication module; and / or, the higher the complexity, the greater the target activation number of the target wireless communication module.

[0013] In some embodiments, obtaining the priority of each wireless communication module includes: obtaining a module evaluation score of each wireless communication module in a historical deep learning stage; obtaining a module computing strength of each wireless communication module in a historical deep learning stage; and determining the priority of each wireless communication module based on the module evaluation score of each wireless communication module and the module computing strength of each wireless communication module.

[0014] Optionally, a module evaluation score of each wireless communication module in a historical deep learning stage is obtained, including: obtaining the accuracy of each wireless communication module in all historical deep learning stages, and obtaining the sum of the accuracy rates of all historical deep learning stages; and determining the module evaluation score of each wireless communication module in the historical deep learning stage based on a positive correlation between the sum of the accuracy rates and the module evaluation scores.

[0015] Optionally, the module computing strength of each wireless communication module in the historical deep learning stage is obtained, including: obtaining the computing amount of each wireless communication module in all historical deep learning stages, and obtaining the sum of the computing amounts in all historical deep learning stages; determining the module computing strength of each wireless communication module in the historical deep learning stage based on the positive correlation between the sum of the computing amount and the module computing strength.

[0016] In some embodiments, it also includes: performing model training on the deep learning model, and after obtaining the model evaluation results, performing model fine-tuning operations on the deep learning model according to the model evaluation results.

[0017] In some embodiments, based on the model evaluation results, a model fine-tuning operation is performed on the deep learning model, including: when the model evaluation results do not meet the accuracy threshold, a model optimization operation is performed on the deep learning model; wherein the model optimization operation includes updating the number of neurons of the target wireless communication module.

[0018] In some embodiments, it also includes: after performing a model optimization operation on the deep learning model, performing model training on the optimized deep learning model again and obtaining a model evaluation optimization result; when the model evaluation optimization result does not meet the accuracy threshold, adding a new wireless communication module to the optimized deep learning model, and updating the optimized deep learning model based on the new wireless communication module.

[0019] In some embodiments, the new wireless communication module is obtained from the target wireless communication module through transfer learning.

[0020] In some embodiments, after performing a model optimization operation on the deep learning model, the optimized deep learning model is again model trained and a model evaluation optimization result is obtained, including: obtaining the priorities of all target wireless communication modules, adjusting the number of neurons of the wireless communication module with the highest priority and updating the deep learning model; performing model training on the updated deep learning model and obtaining a model evaluation result again; if the model evaluation result obtained again does not meet the accuracy threshold, adjusting the number of neurons of the wireless communication module with the second highest priority and updating the deep learning model again until the number of neurons of all target wireless communication modules is updated; and determining that the model training result after updating the number of neurons of all target wireless communication modules is the model evaluation optimization result.

[0021] Optionally, the model optimization operation also includes: adjusting the learning rate of the deep learning model; and / or adjusting the activation function of the deep learning model.

[0022] In some embodiments, the device includes a processor and a memory storing program instructions, wherein the processor is configured to execute the resource configuration method for model training as described above when running the program instructions.

[0023] In some embodiments, an electronic device includes: a device body; and a resource configuration device for model training as described above, installed on the device body.

[0024] The resource configuration method, device, and electronic device for model training provided by the embodiments of the present disclosure can achieve the following technical effects:

[0025] The disclosed embodiments can construct and generate a corresponding deep learning model based on the received task to be processed, and perform model training and obtain model evaluation results based on the constructed deep learning model, so that the deep learning model architecture is adapted to the task to be processed, thereby improving the reliability of task processing and the adaptability of deep learning.

[0026] The above general description and the following description are exemplary and explanatory only and are not intended to limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] One or more embodiments are exemplarily described by corresponding drawings, which do not limit the embodiments. Elements with the same reference numerals in the drawings are shown as similar elements, and the drawings do not constitute a scale limitation, and wherein:

[0028] Figure 1 is a schematic diagram of a resource configuration method for model training provided by an embodiment of the present disclosure;

[0029] Figure 2 is a schematic diagram of another resource configuration method for model training provided by an embodiment of the present disclosure;

[0030] Figure 3 is a schematic diagram of another resource configuration method for model training provided by an embodiment of the present disclosure;

[0031] Figure 4 is a schematic diagram of another resource configuration method for model training provided by an embodiment of the present disclosure;

[0032] Figure 5 is a schematic diagram of another resource configuration method for model training provided by an embodiment of the present disclosure;

[0033] Figure 6 is a schematic diagram of a resource configuration device for model training provided by an embodiment of the present disclosure;

[0034] Figure 7 It is a schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] In order to be able to understand the features and technical contents of the embodiments of the present disclosure in more detail, the implementation of the embodiments of the present disclosure is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not used to limit the embodiments of the present disclosure. In the following technical description, for the convenience of explanation, a full understanding of the disclosed embodiments is provided through multiple details. However, one or more embodiments can still be implemented without these details. In other cases, to simplify the drawings, well-known structures and devices can be simplified for display.

[0036] The terms "first", "second", etc. in the specification and claims of the embodiments of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged where appropriate, so that the embodiments of the embodiments of the present disclosure described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0037] Unless otherwise stated, the term "plurality" means two or more.

[0038] In the embodiment of the present disclosure, the character " / " indicates that the preceding and following objects are in an "or" relationship. For example, A / B indicates: A or B.

[0039] The term "and / or" is a description of the association relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or, A and B.

[0040] The term "correspondence" may refer to an association relationship or a binding relationship. The correspondence between A and B means that there is an association relationship or a binding relationship between A and B.

[0041] In the embodiments of the present disclosure, smart home appliances refer to home appliances that are formed by introducing microprocessors, sensor technology, and network communication technology into home appliances. They have the characteristics of intelligent control, intelligent perception, and intelligent application. The operation process of smart home appliances often relies on the application and processing of modern technologies such as the Internet of Things, the Internet, and electronic chips. For example, smart home appliances can be connected to electronic devices to enable users to remotely control and manage the smart home appliances.

[0042] In the disclosed embodiments, the terminal device refers to an electronic device with a wireless connection function. The terminal device can communicate with the above-mentioned smart home appliances by connecting to the Internet, or can directly communicate with the above-mentioned smart home appliances through Bluetooth, Wi-Fi, etc. In some embodiments, the terminal device is, for example, a mobile device, a computer, or a vehicle-mounted device built into a hover car, or any combination thereof. Mobile devices may include, for example, mobile phones, smart home devices, wearable devices, smart mobile devices, virtual reality devices, etc., or any combination thereof, wherein wearable devices include, for example: smart watches, smart bracelets, pedometers, etc.

[0043] In the embodiments of the present disclosure, the electronic device includes the above-mentioned smart home appliance or the above-mentioned terminal device.

[0044] In the disclosed embodiment, the wireless communication module may be a LoRa module. The LoRa module is a radio frequency module capable of long-distance low-power communication, and uses the LoRa technology to achieve wireless communication. The LoRa module includes a LoRa modem, a radio frequency transceiver, and a microcontroller, and can communicate with other devices through a serial port or SPI (Serial Peripheral Interface). The LoRa module is widely used in the fields of the Internet of Things, smart homes, and wireless data acquisition technology.

[0045] Combination Figure 1 As shown, the embodiment of the present disclosure provides a resource configuration method for model training, including:

[0046] S01, the electronic device receives a task to be processed and determines the type of the task to be processed.

[0047] S02: The electronic device determines, from the wireless communication modules, target wireless communication modules having a target type and a target activation number of each target wireless communication module according to the type of the task to be processed.

[0048] S03, the electronic device builds a deep learning model corresponding to the task to be processed according to the target wireless communication modules with the target type and the target activation number of each target wireless communication module.

[0049] S04, the electronic device performs model training on the deep learning model to obtain a model evaluation result.

[0050] Using the resource configuration method for model training provided by the embodiment of the present disclosure, the embodiment of the present disclosure receives the task to be processed and determines the type of the task to be processed. Then, according to the type of the task to be processed, the target wireless communication module with the target type and the target activation number of each target wireless communication module are determined from the wireless communication module. Then, according to the target wireless communication module of the target type and the target activation number of each target wireless communication module, a deep learning model corresponding to the task to be processed is constructed. Finally, the embodiment of the present disclosure performs model training on the deep learning model to obtain a model evaluation result. It can be seen that the embodiment of the present disclosure can construct and generate a deep learning model corresponding to the received task to be processed, and perform model training according to the constructed deep learning model and obtain a model evaluation result, so that the deep learning model architecture is adapted to the task to be processed, thereby improving the reliability of task processing and the adaptive ability of deep learning.

[0051] Optionally, combined Figure 2 As shown, the electronic device determines a target wireless communication module with a target type and a target activation quantity of each target wireless communication module from the wireless communication modules according to the type of the task to be processed, including:

[0052] S11, the electronic device determines the complexity of the task to be processed according to the type of the task to be processed.

[0053] In this step, the electronic device determines the complexity of the task to be processed according to the type of the task to be processed, including: the electronic device determines the score of the task to be processed according to the information processing amount and data feature complexity of the task to be processed, and determines the complexity of the task to be processed according to the score. The score is positively correlated with the complexity.

[0054] As an example, the score of the task to be processed is score = α × A + β × B. A and B represent the target information processing volume and target data feature complexity of the task to be processed, respectively, and α and β represent the weight coefficient of the information processing volume and the weight coefficient of the data feature complexity, respectively. The target information processing volume is obtained by normalizing the information processing volume of the task to be processed, and the target data feature complexity is obtained by normalizing the data feature complexity of the task to be processed.

[0055] S12, the electronic device determines a target wireless communication module with a target type corresponding to the subtask included in the task to be processed.

[0056] In this step, the electronic device determines the target wireless communication module with the target type corresponding to the subtask according to the subtask included in the task to be processed, including: the electronic device determines the function of the subtask according to the subtask; the electronic device determines the target wireless communication module with the target type according to the function of the subtask, and the target type corresponds to the function of the subtask. As an example, the function of the subtask included in the task to be processed is image recognition, and the electronic device selects the target wireless communication module with the target type corresponding to the image recognition.

[0057] S13, the electronic device obtains the priority of each wireless communication module.

[0058] S14, the electronic device determines a target activation quantity of each target wireless communication module according to the complexity of the task to be processed and the priority of each wireless communication module.

[0059] In this step, the electronic device determines the target activation number of each target wireless communication module according to the complexity of the task to be processed and the priority of each wireless communication module, including: the electronic device selects the target wireless communication module from the wireless communication modules according to the complexity of the task to be processed and the priority of each wireless communication module; the electronic device determines the target activation number of each wireless communication module according to the complexity of the task to be processed. The higher the complexity, the higher the priority of the selected target wireless communication module; and / or the higher the complexity, the greater the target activation number of the target wireless communication module.

[0060] As an example, the task to be processed is a speech recognition task, which includes two subtasks: accent recognition and interesting speech comprehension. The wireless communication module includes a first wireless communication module with a priority level of 8 and a second wireless communication module with a priority level of 6. The electronic device determines that the target wireless communication module with the target type corresponding to the subtask is the first wireless communication module based on the subtasks included in the speech recognition task, and then determines that the target activation number of the first wireless communication module is 8 based on the complexity of the speech recognition task and the priority of each wireless communication module.

[0061] As another example, the task to be processed is an image processing task, and the task to be processed includes three subtasks: simple icon recognition, color analysis, and character face recognition. Among them, the complexity of the image processing task is lower than that of the speech recognition task. The wireless communication module includes a first wireless communication module with a priority level of 8 and a second wireless communication module with a priority level of 6. The electronic device determines that the target wireless communication module with the target type corresponding to the subtask is the second wireless communication module based on the subtasks included in the image processing task, and then determines the target activation number of the second wireless communication module to be 6 based on the complexity of the image processing task and the priority of each wireless communication module.

[0062] In this way, the disclosed embodiment can determine the target wireless communication module with the target type corresponding to the subtask contained in the task to be processed, and then determine the target activation number of each target wireless communication module according to the complexity of the task to be processed and the priority of each wireless communication module. Thus, the fine-tuned deep learning model architecture is adapted to the task to be processed, thereby improving the reliability of task processing and the adaptive ability of deep learning.

[0063] Optionally, combined Figure 3 As shown, the electronic device obtains the priority of each wireless communication module, including:

[0064] S21, the electronic device obtains the module evaluation score of each wireless communication module in the historical deep learning stage.

[0065] In this step, the module evaluation score can be determined based on the accuracy of the wireless communication module in the historical deep learning stage. As an example, the electronic device obtains the module evaluation score of each wireless communication module in the historical deep learning stage, including: the electronic device obtains the accuracy of each wireless communication module in all historical deep learning stages, and obtains the sum of the accuracy of all historical deep learning stages; the electronic device determines the module evaluation score of each wireless communication module in the historical deep learning stage based on the positive correlation between the sum of the accuracy and the module evaluation score.

[0066] S22, the electronic device obtains the module computing intensity of each wireless communication module in the historical deep learning stage.

[0067] In this step, the electronic device obtains the module computing strength of each wireless communication module in the historical deep learning stage, including: the electronic device obtains the computing amount of each wireless communication module in all historical deep learning stages, and obtains the sum of the computing amounts in all historical deep learning stages; the electronic device determines the module computing strength of each wireless communication module in the historical deep learning stage based on the positive correlation between the sum of the computing amount and the module computing strength.

[0068] S23, the electronic device determines the priority of each wireless communication module according to the module evaluation score of each wireless communication module and the module operation strength of each wireless communication module.

[0069] In this way, the embodiment of the present disclosure obtains the module evaluation score of each wireless communication module in the historical deep learning stage and the module operation status in the historical deep learning stage, and comprehensively determines the priority of each wireless communication module based on the module evaluation score and the module operation strength, so as to indirectly reflect the historical module evaluation status and / or operation intensity of the wireless communication module through the priority, and provide accurate data support for determining the type and activation number of the wireless communication module for the task to be processed.

[0070] Optionally, the resource configuration method for model training further includes:

[0071] The electronic device performs model training on the deep learning model, obtains a model evaluation result, and then performs a model fine-tuning operation on the deep learning model according to the model evaluation result.

[0072] In this way, after obtaining the model evaluation results, the embodiment of the present disclosure performs model fine-tuning operations on the deep learning model according to the model evaluation results, so that the deep learning magic core architecture after model fine-tuning is more compatible with the task to be processed, further improving the reliability of task processing.

[0073] Optionally, the electronic device performs a model fine-tuning operation on the deep learning model according to the model evaluation result, including:

[0074] When the model evaluation result does not meet the accuracy threshold, the electronic device performs a model optimization operation on the deep learning model, wherein the model optimization operation includes updating the number of neurons of the target wireless communication module.

[0075] In this way, when the model evaluation result of the embodiment of the present disclosure does not meet the accurate threshold, it indicates that the performance of the deep learning module is poor. At this time, it is necessary to perform model optimization operations on the deep learning model. Therefore, the embodiment of the present disclosure optimizes the model by updating the number of neurons of the target wireless communication module.

[0076] The model evaluation result includes the model accuracy. The model evaluation result does not meet the accuracy threshold, including: the model accuracy is less than the accuracy threshold. The accuracy threshold is greater than or equal to 85% and less than or equal to 90%. Preferably, the accuracy threshold is equal to 85%.

[0077] Combination Figure 4 As shown, the embodiment of the present disclosure provides a resource configuration method for model training, including:

[0078] S31, the electronic device receives a task to be processed and determines the type of the task to be processed.

[0079] S32, the electronic device determines, from the wireless communication modules, target wireless communication modules having a target type and a target activation quantity of each target wireless communication module according to the type of the task to be processed.

[0080] S33, the electronic device builds a deep learning model corresponding to the task to be processed according to the target wireless communication modules with the target type and the target activation number of each target wireless communication module.

[0081] S34, when the model evaluation result does not meet the accuracy threshold, the electronic device performs a model optimization operation on the deep learning model, wherein the model optimization operation includes updating the number of neurons of the target wireless communication module.

[0082] S35, after the electronic device performs model optimization operation on the deep learning model, it performs model training on the optimized deep learning model again and obtains a model evaluation optimization result.

[0083] S36, when the model evaluation optimization result does not meet the accuracy threshold, the electronic device adds a new wireless communication module to the optimized deep learning model, and updates the optimized deep learning model based on the new wireless communication module.

[0084] Using the resource configuration method for model training provided by the embodiment of the present disclosure, the embodiment of the present disclosure receives a task to be processed and determines the type of the task to be processed. Then, according to the type of the task to be processed, a target wireless communication module with a target type and a target activation number of each target wireless communication module are determined from the wireless communication module. Then, according to the target wireless communication module of the target type and the target activation number of each target wireless communication module, a deep learning model corresponding to the task to be processed is constructed. Finally, the embodiment of the present disclosure performs model training on the deep learning model, obtains a model evaluation result, and performs a model fine-tuning operation on the deep learning model according to the model evaluation result. Specifically, in the embodiment of the present disclosure, when the model evaluation result does not meet the accuracy threshold, it indicates that the performance of the deep learning module is poor. At this time, it is necessary to perform a model optimization operation on the deep learning model. Therefore, the embodiment of the present disclosure optimizes the model by updating the number of neurons of the target wireless communication module, and after the model optimization operation, the optimized deep learning model is trained again and a model evaluation optimization result is obtained. If the model optimization result still does not meet the accuracy threshold, it indicates that updating the number of neurons of the target wireless communication module cannot improve the model accuracy of the deep learning model after model training. At this time, it is necessary to optimize the optimized deep learning model again. Specifically, the disclosed embodiment adds a new wireless communication module to the optimized deep learning model and updates the optimized deep learning model based on the new wireless communication module, so that the new wireless communication module works well with the original target wireless communication module.

[0085] Optionally, the new wireless communication module is obtained from the target wireless communication module through transfer learning.

[0086] In this way, a new wireless communication module is obtained through transfer learning, so that the new wireless communication module shares system resources with the target wireless communication module, and the module configuration and module structure of the original target wireless communication module are used as much as possible to ensure that the new wireless communication module can work well with other target wireless communication modules of the deep learning model. In addition, within the feasible range, the new wireless communication module shares the weights and other information of the trained process of the target wireless communication module, which can improve the training efficiency and improve the model accuracy of the deep learning model.

[0087] Combination Figure 5 As shown, after the electronic device performs a model optimization operation on the deep learning model, it performs model training on the optimized deep learning model again and obtains a model evaluation optimization result, including:

[0088] S41, the electronic device obtains the priorities of all target wireless communication modules, adjusts the number of neurons of the wireless communication module with the highest priority, and updates the deep learning model.

[0089] S42, the electronic device performs model training on the updated deep learning model and obtains the model evaluation result again.

[0090] S43, when the model evaluation result obtained again does not meet the accuracy threshold, the electronic device adjusts the number of neurons of the wireless communication module with the second highest priority and updates the deep learning model again until the number of neurons of all target wireless communication modules is updated.

[0091] S44, the electronic device determines that the model training result after updating the number of neurons of all target wireless communication modules is the model evaluation optimization result.

[0092] In this way, the embodiment of the present disclosure obtains the priorities of all target wireless communication modules, adjusts the number of neurons of the wireless communication module with the highest priority, and updates the deep learning model. Afterwards, the embodiment of the present disclosure performs model training on the updated deep learning model and obtains the model evaluation result again. When the model evaluation result obtained again does not meet the accuracy threshold, the embodiment of the present disclosure adjusts the number of neurons of the wireless communication module with the second highest priority and updates the deep learning model again until the number of neurons of all wireless communication modules is updated. In this way, the embodiment of the present disclosure updates the deep learning model in sequence according to the priorities of all target wireless communication modules and determines whether the model accuracy is improved according to the model evaluation results, which is conducive to quickly optimizing the effect of the deep learning model.

[0093] Optionally, the electronic device adjusts the number of neurons of the wireless communication module, including: the electronic device increases the number of neurons of the wireless communication module.

[0094] Optionally, the electronic device performs a model fine-tuning operation on the deep learning model according to the model evaluation result, including:

[0095] When the model evaluation result does not meet the accuracy threshold, the electronic device performs a model optimization operation on the deep learning model. The model optimization operation includes updating the number of neurons of the target wireless communication module, and the model optimization operation also includes adjusting the learning rate of the deep learning model and / or adjusting the activation function of the deep learning model.

[0096] In this way, when the model evaluation result of the embodiment of the present disclosure does not meet the accurate threshold, it indicates that the performance of the deep learning module is poor. At this time, it is necessary to perform model optimization operations on the deep learning model. Therefore, the embodiment of the present disclosure optimizes the model in a variety of ways. The model optimization method can be to update the number of neurons of the target wireless communication module, or to adjust the learning rate and / or activation function of the deep learning model, which is conducive to quickly optimizing the deep learning model.

[0097] It should be noted that the specific details of adjusting the learning rate of the deep learning model and adjusting the activation function of the deep learning model are determined by the task characteristics of the task to be processed and the model evaluation results before optimization.

[0098] Combination Figure 6 As shown, the embodiment of the present disclosure provides a resource configuration device 200 for model training, including a processor (processor) 700 and a memory (memory) 701. Optionally, the device 70 may also include a communication interface (Communication Interface) 702 and a bus 703. Among them, the processor 700, the communication interface 702, and the memory 701 can communicate with each other through the bus 703. The communication interface 702 can be used for information transmission. The processor 700 can call the logic instructions in the memory 701 to execute the resource configuration method for model training of the above embodiment.

[0099] In addition, the logic instructions in the memory 701 described above may be implemented in the form of software functional units and when sold or used as independent products, may be stored in a computer-readable storage medium.

[0100] The memory 701 is a computer-readable storage medium that can be used to store software programs and computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present disclosure. The processor 100 executes the functional application and data processing by running the program instructions / modules stored in the memory 701, that is, the resource configuration method for model training in the above embodiment is implemented.

[0101] The memory 701 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 701 may include a high-speed random access memory and may also include a non-volatile memory.

[0102] Combination Figure 7 As shown, an embodiment of the present disclosure provides an electronic device 100, including: a device body, and the above-mentioned resource configuration device 200 for model training. The resource configuration device 70 for model training is installed on the device body. The installation relationship described here is not limited to placement inside the device body, but also includes installation connections with other components of the electronic device 100, including but not limited to physical connections, electrical connections, or signal transmission connections. It can be understood by those skilled in the art that the resource configuration device 70 for model training can be adapted to a feasible device body, thereby realizing other feasible embodiments.

[0103] An embodiment of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to execute the resource configuration method for model training.

[0104] The technical solution of the embodiment of the present disclosure can be embodied in the form of a software product, which is stored in a storage medium and includes one or more instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiment of the present disclosure. The aforementioned storage medium may be a non-transient storage medium, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes.

[0105] The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure so that those skilled in the art can practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible changes. Unless explicitly required, separate components and functions are optional, and the order of operation may vary. The parts and features of some embodiments may be included in or replace the parts and features of other embodiments. Moreover, the words used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and the claims, unless the context clearly indicates, the singular forms of "a", "an" and "the" are intended to include plural forms as well. Similarly, the term "and / or" as used in this application refers to any and all possible combinations of listings containing one or more associated ones. In addition, when used in the present application, the term "comprise" and its variants "comprises" and / or comprising refer to the presence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof. In the absence of further restrictions, the elements defined by the sentence "comprising a ..." do not exclude the presence of other identical elements in the process, method or device comprising the elements. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the various embodiments may refer to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can refer to the description of the method part.

[0106] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software may depend on the specific application and design constraints of the technical solution. The technicians may use different methods for each specific application to implement the described functions, but such implementations should not be considered to exceed the scope of the embodiments of the present disclosure. The technicians may clearly understand that, for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above may refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0107] In the embodiments disclosed herein, the disclosed methods and products (including but not limited to devices, equipment, etc.) can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units can be only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to implement this embodiment. In addition, each functional unit in the embodiment of the present disclosure may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit.

[0108] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A resource configuration method for model training, characterized in that: include: Receive tasks to be processed and determine the type of tasks to be processed; Determining, from the wireless communication modules, target wireless communication modules having a target type and a target activation number of each target wireless communication module according to the type of the task to be processed; Building a deep learning model corresponding to the task to be processed according to the target wireless communication module having the target type and the target activation number of each target wireless communication module; Perform model training on the deep learning model and obtain model evaluation results.

2. The resource allocation method according to claim 1, characterized in that: According to the type of the task to be processed, determining a target wireless communication module having a target type and a target activation number of each target wireless communication module from the wireless communication modules includes: Determine the complexity of the tasks to be processed based on the types of the tasks to be processed; Determine, according to the subtasks included in the task to be processed, a target wireless communication module with a target type corresponding to the subtask; Get the priority of each wireless communication module; According to the complexity of the task to be processed and the priority of each wireless communication module, the target activation quantity of each target wireless communication module is determined.

3. The resource allocation method according to claim 2, characterized in that: Get the priority of each wireless communication module, including: Obtain module evaluation scores for each wireless communication module in the historical deep learning stage; Obtain the module computing intensity of each wireless communication module in the historical deep learning stage; The priority of each wireless communication module is determined according to the module evaluation score of each wireless communication module and the module operation strength of each wireless communication module.

4. The resource allocation method according to claim 1, characterized in that: Also includes: After training the deep learning model and obtaining the model evaluation results, fine-tune the deep learning model based on the model evaluation results.

5. The resource allocation method according to claim 4, characterized in that: Based on the model evaluation results, fine-tune the deep learning model, including: When the model evaluation result does not meet the accuracy threshold, a model optimization operation is performed on the deep learning model; wherein the model optimization operation includes updating the number of neurons of the target wireless communication module.

6. The resource allocation method according to claim 5, characterized in that: Also includes: After performing model optimization operations on the deep learning model, the optimized deep learning model is trained again and a model evaluation optimization result is obtained; When the model evaluation optimization result does not meet the accuracy threshold, a new wireless communication module is added to the optimized deep learning model, and the optimized deep learning model is updated based on the new wireless communication module.

7. The resource allocation method according to claim 6, characterized in that: The new wireless communication module is obtained by transfer learning from the target wireless communication module.

8. The resource allocation method according to claim 6, characterized in that: After the deep learning model is optimized, the optimized deep learning model is trained again and the model evaluation optimization results are obtained, including: Obtain the priorities of all target wireless communication modules, adjust the number of neurons of the wireless communication module with the highest priority, and update the deep learning model; Perform model training on the updated deep learning model and obtain model evaluation results again; When the model evaluation result obtained again does not meet the accuracy threshold, the number of neurons of the wireless communication module with the second highest priority is adjusted and the deep learning model is updated again until the number of neurons of all target wireless communication modules is updated; The model training result after determining and updating the number of neurons of all target wireless communication modules is the model evaluation optimization result.

9. A resource configuration device for model training, comprising a processor and a memory storing program instructions, characterized in that: The processor is configured to execute the resource configuration method for model training as described in any one of claims 1 to 8 when running the program instructions.

10. An electronic device, characterized in that: include: Equipment body; The resource configuration device for model training as described in claim 9 is installed on the device body.