Scene model calculation method and apparatus, and electronic device

By compressing the scene model and matching the performance and complexity scores of home appliance equipment, the problem of insufficient computing power in smart home appliance equipment with limited computing power is solved, and efficient distributed computing and big data analysis are achieved.

CN118916710APending Publication Date: 2024-11-08QINGDAO HAIER TECH +3
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411024349.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The prior art is difficult to effectively run large-scale deep learning models on terminal devices with limited computing power, especially in smart home appliances, resulting in insufficient computing power.

Method used

By compressing the calculated scene model, the compression model calculation task is obtained, and the performance score of the home appliance equipment and the complexity score of the compression model calculation task is matched, the compression model calculation task is assigned to the matching home appliance equipment, and distributed computing is realized .

Benefits of technology

The magnitude of the full-size model is reduced, allowing home appliances to run scenario models efficiently, implement large-scale parallel processing and big data analysis, and solve the problem of insufficient computing power.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916710A_ABST
    Figure CN118916710A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of smart home, and discloses a scene model calculation method, which comprises the following steps: compressing a full-size model of a scene model to be calculated to obtain a compression model calculation task corresponding to the scene model; obtaining a performance score of the household appliance and a complexity score of the compression model calculation task, and matching the household appliance with the compression model calculation task according to the performance score of the household appliance and the complexity score of the compression model calculation task; and determining a target household appliance matched with the target compression model calculation task, so that the target household appliance executes the target compression model calculation task. According to the method, the large model can be compressed and allocated to the household appliances for calculation. The invention further discloses a computing device for the scene model 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 smart home technology, for example, to a method and device for calculating a scene model, and an electronic device. Background Art

[0002] At present, with the development of deep learning technology, models are getting larger and larger, but many terminal devices such as smartphones and smart home appliances cannot effectively run these large models due to power consumption and computing power limitations. How to deploy trained large models to devices with limited computing power and perform efficient reasoning operations is an important challenge facing the current deep learning field.

[0003] In order to meet the demand of running large models on terminal devices, the relevant technology discloses a deep learning model training and reasoning architecture deployment method based on cloud-edge-end collaboration, based on cloud servers, edge servers, and physical terminals. The edge servers are equipped with training modules, reasoning modules, and situational awareness centers; the situational awareness center includes an edge server computing power perception module for perceiving the computing resources of the edge server, the bandwidth occupancy between the edge server and the physical distance calculation, and a cloud server computing power perception module for perceiving the bandwidth occupancy between the edge server and the cloud server, the physical distance calculation, and perceiving the computing resources of the cloud server. The training module and the reasoning module perform resource cost calculation to determine the architecture deployment.

[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] Although the related technology can train the model through the edge server and the physical terminal, it does not adjust the large model. When the model has a large amount of data, the edge server and the physical terminal may have insufficient computing power.

[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 method and device for calculating a scene model, and an electronic device, which can compress a large model and distribute it to household appliances for calculation.

[0009] In some embodiments, the scene model calculation method includes: compressing the full-size model of the scene model to be calculated to obtain a compression model calculation task corresponding to the scene model; obtaining the performance score of the household appliance and the complexity score of the compression model calculation task, so as to match the household appliance and the compression model calculation task according to the performance score of the household appliance and the complexity score of the compression model calculation task; and determining a match with the target compression model calculation task so that the target household appliance executes the target compression model calculation task.

[0010] Optionally, the full-size model of the scene model to be calculated is compressed to obtain a compressed model calculation task corresponding to the scene model, including: performing model cropping on the full-size model to obtain a first compressed model; wherein the model cropping is used to achieve sparseness of the weight matrix of the full-size model; performing model quantization on the first compressed model to obtain a second compressed model; wherein the second compressed model is used to achieve bit reduction of the first compressed model; and using the second compressed model as the compressed model calculation task corresponding to the scene model.

[0011] Optionally, obtaining a performance score of the household appliance includes: obtaining a sub-performance score of the household appliance; wherein the sub-performance score includes a computing power score, a network connectivity score and a battery status score; determining a sub-performance weight corresponding to the sub-performance score; wherein the sub-performance weight includes a computing power weight, a network connectivity weight and a battery status weight, the computing power weight corresponds to the computing power score, the network connectivity weight corresponds to the network connectivity score and the battery status weight corresponds to the battery status weight score; and calculating the performance score of the household appliance based on the sub-performance score and the sub-performance weight.

[0012] Optionally, the household appliances and the compression model calculation tasks are matched according to the performance scores of the household appliances and the complexity scores of the compression model calculation tasks, including: matching the performance scores of the household appliances and the complexity scores of the compression model calculation tasks to determine multiple household appliances corresponding to the target compression model calculation tasks; obtaining the priority level of the target compression model calculation tasks and the data processing stability levels of the multiple household appliances; and determining the target household appliances that match the target compression model calculation tasks from the multiple household appliances according to the priority level of the target compression model calculation tasks and the data processing stability levels of the multiple household appliances.

[0013] Optionally, based on the priority level of the target compression model calculation task and the data processing stability level of multiple home appliances, a target home appliance that matches the target compression model calculation task is determined from multiple home appliances, including: sorting the multiple home appliances according to the data processing stability level of the home appliances; determining the target home appliance that matches the target compression model calculation task based on the priority level of the target compression model calculation task; wherein the priority level of the target compression model corresponds to the sorting position of the target home appliance among the sorted multiple home appliances.

[0014] Optionally, the data processing stability level of the household appliance is determined in the following manner: determining the number of acceleration units and / or resource utilization of different household appliances; determining the data processing stability level of the household appliance based on the number of acceleration units and / or resource utilization; wherein the data processing stability level of the household appliance is positively correlated with the number of acceleration units, and the resource utilization of the household appliance is negatively correlated with the number of acceleration units.

[0015] Optionally, a target home appliance that matches the target compression model calculation task is determined so that the target home appliance executes the target compression model calculation task, including: when there are multiple target home appliances, splitting the target compression model calculation task into multiple compression model calculation subtasks; and matching the multiple compression model calculation subtasks with the multiple target home appliances respectively so that the target home appliances execute the compression model calculation subtasks.

[0016] Optionally, the calculation method of the scenario model also includes: after multiple target home appliances execute the compression model calculation subtask, obtaining the distribution calculation results of the multiple target home appliances; merging the distribution calculation results of the multiple target home appliances to obtain the calculation result of the target compression model calculation task.

[0017] In some embodiments, the scene model calculation device includes: a processor and a memory storing program instructions, and the processor is configured to execute the scene model calculation method as described above when running the program instructions.

[0018] In some embodiments, the electronic device includes: an electronic device body; and a computing device for the scene model as described above, installed on the electronic device body.

[0019] The scene model calculation method, device, and electronic device provided by the embodiments of the present disclosure can achieve the following technical effects:

[0020] In the disclosed embodiment, before distributing the computing tasks to the home appliances, the full-size model of the scene model is compressed, and the data processing capability of the server can be used for large-scale parallel processing and big data analysis to reduce the magnitude of the full-size model, thereby enabling the home appliances to effectively run the scene model. When distributing the computing tasks to the home appliances, the target home appliances are determined based on the performance scores of the home appliances and the complexity scores of the compressed model computing tasks, so that the computing tasks of the compressed model can match the computing capabilities of the target home appliances, ensuring that the target home appliances can efficiently run the computing tasks, and realizing the distributed computing of the full-size model of the scene model in the home appliances.

[0021] 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

[0022] 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:

[0023] Figure 1 It is a hardware environment diagram of a method for calculating a scene model provided by an embodiment of the present disclosure;

[0024] Figure 2 is a schematic diagram of a method for calculating a scene model provided by an embodiment of the present disclosure;

[0025] Figure 3 is a schematic diagram of another method for calculating a scene model provided by an embodiment of the present disclosure;

[0026] Figure 4 is a schematic diagram of another method for calculating a scene model provided by an embodiment of the present disclosure;

[0027] Figure 5 is a schematic diagram of another method for calculating a scene model provided by an embodiment of the present disclosure;

[0028] Figure 6 It is a schematic diagram of a computing device for a scene model provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] 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.

[0030] The terms "first", "second", etc. in the technical solutions described in the embodiments of the present disclosure are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present disclosure described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0031] Unless otherwise stated, the term "plurality" means two or more.

[0032] 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.

[0033] 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.

[0034] 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.

[0035] At present, existing technologies mainly optimize and deploy models for a single or several types of devices, and do not fully consider the rich variety of devices in the smart home environment, large differences in hardware performance, and complex network connections. Therefore, current technologies are insufficient in dealing with issues such as coordination and scheduling of heterogeneous devices, network latency, and data security.

[0036] According to one aspect of an embodiment of the present application, a method for calculating a scene model is provided. The method for calculating the scene model is widely used in smart home, smart home, smart home device ecology, smart house ecology and other whole-house intelligent digital control application scenarios. Optionally, in this embodiment, the method for calculating the scene model can be applied to Figure 1 In the hardware environment composed of the home appliance 102 and the server 104 shown in FIG. Figure 1As shown, the server 104 is connected to the home appliance 102 via a network, and can be used to provide services (such as application services, etc.) for the terminal or a client installed on the terminal. A database can be set on the server or independently of the server to provide data storage services for the server 104. Cloud computing and / or edge computing services can be configured on the server or independently of the server to provide data computing services for the server 104.

[0037] The network may include but is not limited to at least one of the following: wired network, wireless network. The wired network may include but is not limited to at least one of the following: wide area network, metropolitan area network, local area network, and the wireless network may include but is not limited to at least one of the following: WIFI (Wireless Fidelity), Bluetooth. The home appliance 102 may be but is not limited to a PC, a mobile phone, a tablet computer, a smart air conditioner, a smart range hood, a smart refrigerator, a smart oven, a smart stove, a smart washing machine, a smart water heater, a smart washing equipment, a smart dishwasher, a smart projection equipment, a smart TV, a smart clothes drying rack, a smart curtain, a smart audio and video, a smart socket, a smart speaker, a smart speaker, a smart fresh air equipment, a smart kitchen and bathroom equipment, a smart bathroom equipment, a smart sweeping robot, a smart window cleaning robot, a smart mopping robot, a smart air purification equipment, a smart steamer, a smart microwave oven, a smart kitchen treasure, a smart purifier, a smart water dispenser, a smart door lock, etc.

[0038] Combination Figure 2 As shown, the present disclosure provides a method for calculating a scene model, including:

[0039] S201, the server compresses the full-size model of the scene model to be calculated, and obtains a compressed model calculation task corresponding to the scene model.

[0040] S202, the server obtains the performance score of the home appliance and the complexity score of the compression model calculation task, so as to match the home appliance with the compression model calculation task according to the performance score of the home appliance and the complexity score of the compression model calculation task.

[0041] S203, the server determines a target home appliance that matches the target compression model calculation task, so that the target home appliance executes the target compression model calculation task.

[0042] By using the scene model calculation method provided by the embodiment of the present disclosure, the full-size model of the scene model is compressed before the calculation task is distributed to the home appliance, and the data processing capability of the server can be used for large-scale parallel processing and big data analysis to reduce the magnitude of the full-size model, thereby enabling the home appliance to effectively run the scene model model. When the calculation task is distributed to the home appliance, the target home appliance is determined based on the performance score of the home appliance and the complexity score of the compression model calculation task, so that the calculation task of the compression model can match the calculation capability of the target home appliance, ensuring that the target home appliance can efficiently run the calculation task, and realizing the distributed calculation of the full-size model of the scene model in the home appliance.

[0043] Optionally, the full-size model of the scene model to be calculated is compressed to obtain a compressed model calculation task corresponding to the scene model, including: performing model cropping on the full-size model to obtain a first compressed model; wherein the model cropping is used to achieve sparseness of the weight matrix of the full-size model; performing model quantization on the first compressed model to obtain a second compressed model; wherein the second compressed model is used to achieve bit reduction of the first compressed model; and using the second compressed model as the compressed model calculation task corresponding to the scene model.

[0044] Combination Figure 3 As shown, the embodiment of the present disclosure provides another method for calculating a scene model, including:

[0045] S301, the server performs model pruning on the full-size model to obtain a first compressed model; wherein the model pruning is used to achieve sparseness of the weight matrix of the full-size model.

[0046] S302, the server performs model quantization on the first compression model to obtain a second compression model; wherein the second compression model is used to achieve bit reduction of the first compression model.

[0047] S303: The server uses the second compression model as the compression model calculation task corresponding to the scene model.

[0048] S304, the server obtains the performance score of the home appliance and the complexity score of the compression model calculation task, so as to match the home appliance with the compression model calculation task according to the performance score of the home appliance and the complexity score of the compression model calculation task.

[0049] S305, the server determines a target home appliance that matches the target compression model calculation task, so that the target home appliance executes the target compression model calculation task.

[0050] In the disclosed embodiment, model compression can be completed on the server. The powerful processing power and data storage capacity provided by the server can be used for large-scale parallel processing and big data analysis. The main technologies for model compression include pruning, network search, quantization, distillation, etc.

[0051] Model pruning reduces the model size and computational complexity by removing some of the weights of the neural network. You can add a sparsity regularization term to make the weight matrix sparse during neural network training or prune by starting a fine-tuning process combined with a thresholding or Top-K strategy.

[0052] Quantization is a method of converting 32-bit floating point numbers into smaller bit rates to reduce model size and inference time. If applied in actual scenarios, common algorithms that can be selected include DoReFaNet, BinaryNet, etc.

[0053] Optionally, obtaining a performance score of the household appliance includes: obtaining a sub-performance score of the household appliance; wherein the sub-performance score includes a computing power score, a network connectivity score and a battery status score; determining a sub-performance weight corresponding to the sub-performance score; wherein the sub-performance weight includes a computing power weight, a network connectivity weight and a battery status weight, the computing power weight corresponds to the computing power score, the network connectivity weight corresponds to the network connectivity score and the battery status weight corresponds to the battery status weight score; and calculating the performance score of the household appliance based on the sub-performance score and the sub-performance weight.

[0054] In the disclosed embodiments, it is necessary to detect and evaluate the performance of household appliances. First, it is necessary to know the computing power and power consumption of each household appliance. Detecting computing power and power consumption includes but is not limited to detecting CPU (Central Processing Unit), GPU (Graphics Processing Unit), memory and battery status, etc. The server can obtain information through polling or event triggering, and upload the computing power and power consumption of household appliances to the management server. When evaluating performance, a multi-factor comprehensive scoring system can be established, including factors such as the computing power of the device, network connectivity, and battery status, to obtain computing power scores, network connectivity scores, and battery status scores.

[0055] For example, the performance score of the household appliance is calculated according to the following formula:

[0056] D=w1×d1+w2×d2+w3×d2

[0057] Among them, D is the performance score of the home appliance, d1 is the computing power score, d2 is the network connectivity score, d3 is the battery status score, w1 is the computing power weight, w2 is the network connectivity weight, and w3 is the battery status weight.

[0058] Specifically, 0 <w1<1,0<w2<1,0<w3<1。

[0059] Assigning a weight to each sub-performance score enables the server to calculate a comprehensive score based on the importance of different scoring factors.

[0060] Optionally, the household appliances and the compression model calculation tasks are matched according to the performance scores of the household appliances and the complexity scores of the compression model calculation tasks, including: matching the performance scores of the household appliances and the complexity scores of the compression model calculation tasks to determine multiple household appliances corresponding to the target compression model calculation tasks; obtaining the priority level of the target compression model calculation tasks and the data processing stability levels of the multiple household appliances; and determining the target household appliances that match the target compression model calculation tasks from the multiple household appliances according to the priority level of the target compression model calculation tasks and the data processing stability levels of the multiple household appliances.

[0061] Combination Figure 4 As shown, the embodiment of the present disclosure provides another method for calculating a scene model, including:

[0062] S401, the server compresses the full-size model of the scene model to be calculated, and obtains a compressed model calculation task corresponding to the scene model.

[0063] S402, the server obtains the performance score of the home appliance and the complexity score of the compression model calculation task.

[0064] S403, the server matches the performance score of the household appliance and the complexity score of the compression model calculation task, and determines a plurality of household appliances corresponding to the target compression model calculation task.

[0065] S404, the server obtains the priority level of the target compression model calculation task and the data processing stability level of multiple home appliances.

[0066] S405, the server determines a target home appliance that matches the target compression model calculation task from multiple home appliances according to the priority level of the target compression model calculation task and the data processing stability level of the multiple home appliances.

[0067] S406, the server determines a target home appliance that matches the target compression model calculation task, so that the target home appliance executes the target compression model calculation task.

[0068] In the disclosed embodiment, in order to meet the requirements of home appliances for ultra-low computing power and ultra-low power consumption, effective resource management is required. The server can automatically detect and evaluate the performance of each device, and perform appropriate management and scheduling according to the computing tasks. Based on the performance score of the home appliance and the complexity score of the compression model computing task, the server can decide to assign the task to a matching device for execution. For example, resource-intensive tasks should be assigned to devices with strong computing power and sufficient power, while lightweight tasks can be undertaken by devices with relatively weak computing power or low power.

[0069] Optionally, the server has dynamic scheduling and load balancing capabilities. If a device fails or resource utilization is too high, the task being processed by the device needs to be quickly migrated to other devices. If new devices are added to the smart home system, resource evaluation and allocation need to be re-performed.

[0070] Optionally, for scenarios with ultra-low power requirements, some energy-saving strategies can be introduced to reduce power consumption, such as running low-priority or deferred tasks during off-peak hours, or designing special modes for devices that are idle for a long time but do not use the shutdown function so that they can be called at any time.

[0071] Optionally, based on the priority level of the target compression model calculation task and the data processing stability level of multiple home appliances, a target home appliance that matches the target compression model calculation task is determined from multiple home appliances, including: sorting the multiple home appliances according to the data processing stability level of the home appliances; determining the target home appliance that matches the target compression model calculation task based on the priority level of the target compression model calculation task; wherein the priority level of the target compression model corresponds to the sorting position of the target home appliance among the sorted multiple home appliances.

[0072] In the disclosed embodiment, multiple home appliances are arranged according to their data processing stability levels. When the devices are idle, the server takes out appropriate tasks from the queue according to the device performance and task type to run. At the same time, special situations also need to be considered, such as the sudden offline of the device or the interruption of the task, which will cause the task to be returned to the queue for reallocation.

[0073] In a smart home system, different home appliances may have different computing resources and capabilities. The server can assign tasks to home appliances for processing according to task requirements. For example, some tasks with high complexity and requiring a lot of computing resources can be handed over to devices with better performance, while devices with poor performance or low battery power are only responsible for processing some simple or low-priority tasks.

[0074] Optionally, the data processing stability level of the household appliance is determined in the following manner: determining the number of acceleration units and / or resource utilization of different household appliances; determining the data processing stability level of the household appliance based on the number of acceleration units and / or resource utilization; wherein the data processing stability level of the household appliance is positively correlated with the number of acceleration units, and the resource utilization of the household appliance is negatively correlated with the number of acceleration units.

[0075] In the disclosed embodiment, the server needs to identify and manage various heterogeneous devices. For example, some home appliances may have built-in processors with specific acceleration units. The server can identify and make full use of these dedicated hardware to accelerate computing. First, the server needs to identify various different hardware devices, understand the computing power of each device and the acceleration units possessed by the device. Including but not limited to CPU, GPU, FPGA (Field-Programmable Gate Array), TPU (Tensor Processing Unit), etc. In addition, establish a connection with the added hardware and obtain relevant information about the added hardware.

[0076] For home appliances with built-in processors with acceleration units (such as GPUs, TPUs, etc.), it is necessary to identify and fully utilize the characteristics of the acceleration units to perform computing tasks. More specifically, it is necessary to match the corresponding processor according to the task type and utilize the processor's acceleration unit. In addition, the operation mode should also be optimized to minimize power consumption while maintaining high performance. In this way, by identifying the acceleration unit of the device, the data processing stability level of the home appliance can be determined based on the number of acceleration units and / or resource utilization.

[0077] Optionally, a target home appliance that matches the target compression model calculation task is determined so that the target home appliance executes the target compression model calculation task, including: when there are multiple target home appliances, splitting the target compression model calculation task into multiple compression model calculation subtasks; and matching the multiple compression model calculation subtasks with the multiple target home appliances respectively so that the target home appliances execute the compression model calculation subtasks.

[0078] Combination Figure 5 As shown, the embodiment of the present disclosure provides another method for calculating a scene model, including:

[0079] S501, the server compresses the full-size model of the scene model to be calculated, and obtains a compressed model calculation task corresponding to the scene model.

[0080] S502, the server obtains the performance score of the home appliance and the complexity score of the compression model calculation task, so as to match the home appliance with the compression model calculation task according to the performance score of the home appliance and the complexity score of the compression model calculation task.

[0081] S503: When there are multiple target home appliances, the server splits the target compression model calculation task into multiple compression model calculation subtasks.

[0082] S504, the server matches the multiple compression model calculation subtasks with the multiple target home appliances respectively, so that the target home appliances execute the compression model calculation subtasks.

[0083] In the disclosed embodiment, by splitting the compression model, each device only needs to process a portion of the data or perform a portion of the calculation.

[0084] The splitting method includes: when the model is too large to be stored or calculated on a single device, the model is split onto multiple devices for parallel calculation. Each device stores a part of the model and calculates it locally. Then the parts are merged through a network connection to produce the output result.

[0085] Another way to split is data parallelism. Each device gets a complete model, but only processes a portion of the data. Then the gradients or other update values ​​calculated by each device are synchronized to all devices to complete a global update. This method is very suitable when the data set is very large and the model is small.

[0086] Optionally, the calculation method of the scenario model also includes: after multiple target home appliances execute the compression model calculation subtask, obtaining the distribution calculation results of the multiple target home appliances; merging the distribution calculation results of the multiple target home appliances to obtain the calculation result of the target compression model calculation task.

[0087] In the disclosed embodiment, after each household appliance processes the data or task, it will send the result back to the central node for merging and integration. By splitting the model for calculation, it will be easier and more efficient to deploy deep learning models in large-scale networks.

[0088] In addition, during the implementation process, it is necessary to select a reasonable communication framework and protocol to improve communication efficiency and data throughput to ensure the overall performance of the smart home system.

[0089] Combination Figure 6As shown, the embodiment of the present disclosure provides a computing device 60 for a scene model, including a processor 600 and a memory 601. Optionally, the device 60 may also include a communication interface 602 and a bus 603. The processor 600, the communication interface 602, and the memory 601 may communicate with each other through the bus 603. The communication interface 602 may be used for information transmission. The processor 600 may call the logic instructions in the memory 601 to execute the scene model computing method of the above embodiment.

[0090] In addition, the logic instructions in the memory 601 described above can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0091] The memory 601 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 600 executes the functional application and data processing by running the program instructions / modules stored in the memory 601, that is, the calculation method of the scene model in the above embodiment is implemented.

[0092] The memory 601 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 601 may include a high-speed random access memory and may also include a non-volatile memory.

[0093] The embodiments of the present disclosure provide an electronic device, including: an electronic device body, and the above-mentioned scene model computing device 60. The scene model computing device 60 is installed in the electronic device body. The installation relationship described here is not limited to placement inside the electronic device body, but also includes installation connections with other components of the electronic device, including but not limited to physical connections, electrical connections or signal transmission connections, etc. Those skilled in the art can understand that the scene model computing device 60 can be adapted to a feasible electronic device body, thereby realizing other feasible embodiments.

[0094] 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.

[0095] 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.

[0096] 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.

[0097] 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.

[0098] 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 method for calculating a scene model, characterized in that: include: Compressing the full-size model of the scene model to be calculated to obtain a compressed model calculation task corresponding to the scene model; Acquire the performance score of the household appliance and the complexity score of the compression model calculation task, so as to match the household appliance with the compression model calculation task according to the performance score of the household appliance and the complexity score of the compression model calculation task; A target home appliance device that matches the target compression model calculation task is determined so that the target home appliance device performs the target compression model calculation task.

2. The method for calculating the scene model according to claim 1, characterized in that: Compress the full-size model of the scene model to be calculated to obtain the compressed model calculation task corresponding to the scene model, including: Performing model pruning on the full-size model to obtain a first compressed model; wherein the model pruning is used to achieve sparseness of the weight matrix of the full-size model; Performing model quantization on the first compression model to obtain a second compression model; wherein the second compression model is used to achieve bit number reduction of the first compression model; The second compression model is used as the compression model calculation task corresponding to the scene model.

3. The method for calculating the scene model according to claim 1, characterized in that: Get performance scores for home appliances, including: Obtaining sub-performance scores of the home appliance; wherein the sub-performance scores include computing power score, network connectivity score and battery status score; Determine the sub-performance weight corresponding to the sub-performance score; wherein the sub-performance weight includes a computing power weight, a network connectivity weight and a battery status weight, the computing power weight corresponds to the computing power score, the network connectivity weight corresponds to the network connectivity score and the battery status weight corresponds to the battery status weight score; The performance score of the household appliance is calculated based on the sub-performance scores and sub-performance weights.

4. The method for calculating the scene model according to claim 1, characterized in that: The home appliance and the compression model computing task are matched according to the performance score of the home appliance and the complexity score of the compression model computing task, including: Matching the performance score of the household appliance and the complexity score of the compression model calculation task, and determining a plurality of household appliances corresponding to the target compression model calculation task; Obtaining the priority level of target compression model computing tasks and the data processing stability level of multiple home appliances; According to the priority level of the target compression model calculation task and the data processing stability level of multiple home appliances, a target home appliance matching the target compression model calculation task is determined from multiple home appliances.

5. The method for calculating the scene model according to claim 4, characterized in that: According to the priority level of the target compression model calculation task and the data processing stability level of the plurality of home appliances, a target home appliance matching the target compression model calculation task is determined from the plurality of home appliances, including: sorting a plurality of home appliances according to the data processing stability levels of the home appliances; According to the priority level of the target compression model calculation task, the target home appliance matching the target compression model calculation task is determined; wherein the priority level of the target compression model corresponds to the sorting position of the target home appliance among the multiple home appliances after sorting.

6. The method for calculating the scene model according to claim 4, characterized in that: The data processing stability level of the home appliance is determined as follows: Determine the number of acceleration units and / or resource utilization of different home appliances; Determining a data processing stability level of the home appliance according to the number of acceleration units and / or resource utilization; Among them, the data processing stability level of home appliances is positively correlated with the number of acceleration units, and the resource utilization rate of home appliances is negatively correlated with the number of acceleration units.

7. The method for calculating a scene model according to any one of claims 1 to 6, characterized in that: Determining a target home appliance device that matches the target compression model calculation task so that the target home appliance device performs the target compression model calculation task, including: In the case where there are multiple target home appliances, the target compression model calculation task is split into multiple compression model calculation subtasks; The plurality of compression model calculation subtasks are matched with the plurality of target home appliances respectively, so that the target home appliances execute the compression model calculation subtasks.

8. The method for calculating the scene model according to claim 7, characterized in that: Also includes: After the multiple target home appliances execute the compression model calculation subtask, the distribution calculation results of the multiple target home appliances are obtained; The distributed calculation results of multiple target home appliances are merged to obtain the calculation results of the target compression model calculation task.

9. A computing device for a scene model, comprising a processor and a memory storing program instructions, characterized in that: The processor is configured to execute the method for calculating the scene model according to any one of claims 1 to 8 when running the program instructions.

10. An electronic device, characterized in that: include: Electronic device body; The scene model computing device as described in claim 9 is installed on the electronic device body.

Citation Information

Patent Citations

  • Resource scheduling method and device of Hadoop cluster

    CN106933664A

  • Method and device for detecting home scene, electronic equipment and medium

    CN114519818A

  • Deep network model lightweight method and system applied to image processing

    CN115186805A

  • Method and device for determining optimal scheduling strategy and electronic equipment

    CN115348324A

  • Task allocation method and device, storage medium and electronic device

    CN118349331A