Task processing method and device, computer equipment and storage medium

Abstract

The invention discloses a task processing method and device, computer equipment and a storage medium. The method belongs to the technical field of end-cloud collaboration, and specifically comprises the following steps: in response to a task execution request carrying a target task, determining task information of the target task and current terminal capability information of a terminal; and determining a task processing end according to the task information and the current terminal capability information. And processing the target task based on the task processing end. Wherein the task processing end is a terminal and / or a cloud end; the task information comprises at least one of task complexity and task privacy. Compared with a traditional'one-cut 'type task allocation mechanism, when the task processing end is determined, not only is the current terminal capability information of the terminal considered, but also the task complexity and the task privacy of the target task are considered, namely, the terminal resources are fully utilized, the personalized requirements of the user are considered, and the task processing end is determined. And the task reasoning burden of the cloud is reduced, and the task execution efficiency and the user satisfaction are improved.

Description

Technical Field

[0001] This application relates to the field of edge-cloud collaboration technology, specifically to a task processing method, apparatus, computer equipment, and storage medium. Background Technology

[0002] The collaborative deployment of large and small models on the cloud involves deploying large AI (Artificial Intelligence) models in the cloud, leveraging their powerful computing capabilities and rich knowledge representation capabilities to perform complex task processing and deep reasoning; at the same time, deploying small AI models on the edge to handle rapid local data collection, preliminary processing, and simple reasoning tasks.

[0003] In current research on collaborative deployment of large and small models on both the edge and cloud, the deployer often pre-configures the edge device to perform simple processing on data such as images, sounds, and environmental data using the edge model, while the cloud device performs deep learning, analysis, and computational inference on the data uploaded from the edge device and returns the results to the edge device. However, this "one-size-fits-all" task processing mechanism lacks rationality, resulting in slightly more difficult tasks being assigned to the cloud. This not only increases the task processing pressure on the cloud but also reduces task processing efficiency and affects user experience. Summary of the Invention

[0004] Therefore, it is necessary to provide a task processing method, apparatus, computer equipment, and storage medium that can effectively improve task processing efficiency in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a task processing method applied to a terminal, the method comprising:

[0006] In response to a task execution request carrying a target task, the system determines the task information of the target task and the current terminal capability information of the terminal; wherein the task information includes at least one of task complexity and task privacy.

[0007] Based on the task information and the current terminal capability information, determine the task processing terminal; wherein, the task processing terminal is the terminal and / or the cloud.

[0008] The target task is processed based on the task processing terminal.

[0009] In one embodiment, determining the task processing terminal based on task information and current terminal capability information includes:

[0010] Based on the current terminal capability information, determine the first evaluation value;

[0011] Determine the second evaluation value based on the task complexity;

[0012] A third evaluation value is determined based on task privacy.

[0013] The task processing end is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0014] In one embodiment, determining the task processing end based on a first evaluation value, a second evaluation value, and a third evaluation value includes:

[0015] The first, second, and third assessment values ​​are weighted and summed to obtain the total assessment score.

[0016] The task processing end is determined based on the relationship between the total evaluation score and the judgment threshold.

[0017] In one embodiment, the current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information.

[0018] In one embodiment, determining a first evaluation value based on current terminal capability information includes:

[0019] The first score is determined based on the current terminal hardware resource information;

[0020] The second score is determined based on the current computing resource information;

[0021] The third score is determined based on the current communication resource information;

[0022] The first evaluation value is determined based on the first score, the second score, and the third score.

[0023] In one embodiment, determining a first evaluation value based on a first score, a second score, and a third score includes:

[0024] Determine the target endpoint model associated with the target task;

[0025] Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score;

[0026] Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score;

[0027] The first evaluation value is determined based on the terminal capability score.

[0028] In one embodiment, determining a first evaluation value based on the terminal capability score includes:

[0029] Determine the accuracy of the target end-side model;

[0030] The product of accuracy and terminal capability score is used as the first evaluation value.

[0031] In one embodiment, determining the task complexity of the target task includes:

[0032] Based on the task content of the target task, select related tasks from the historical tasks;

[0033] Obtain processing information for related tasks processed by the terminal;

[0034] Based on the processed information, determine the task complexity of the target task.

[0035] In one embodiment, task privacy includes at least one of privacy requirements and user privacy terms information.

[0036] In one embodiment, the current terminal hardware resource information includes at least one of the following: device model, current battery level, and current temperature;

[0037] Current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency;

[0038] Current communication resource information includes at least one of the following: current edge-side model capabilities, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capabilities, and currently cloud-stored models that can be deployed on the edge.

[0039] In one embodiment, in response to a task execution request carrying a target task, the task information of the target task and the current terminal capability information of the terminal are determined, including:

[0040] In response to a task execution request carrying a target task, determine the association model of the target task;

[0041] When the association model includes the edge model, the task information of the target task and the current terminal capability information of the terminal are determined.

[0042] In one embodiment, after processing the target task at the task processing end, the method further includes:

[0043] Obtain the processing effect of the target task from the task processing terminal;

[0044] If the processing effect does not meet the preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized.

[0045] Secondly, this application provides a task processing apparatus configured in a terminal, the apparatus comprising:

[0046] The first determining module is used to determine the task information of the target task and the current terminal capability information of the terminal in response to a task execution request carrying a target task; wherein the task information includes at least one of task complexity and task privacy.

[0047] The second determining module is used to determine the task processing end based on the task information and the current terminal capability information; wherein, the task processing end is the terminal and / or the cloud.

[0048] The processing module is used to process the target task based on the task processing terminal.

[0049] Thirdly, this application also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0050] In response to a task execution request carrying a target task, obtain the current terminal performance information;

[0051] Determine the edge-cloud collaborative decision value based on the current terminal performance information;

[0052] Based on edge-cloud collaborative decision values, the execution end for executing the target task is determined; where the execution end is the terminal and / or the cloud.

[0053] Fourthly, this application also provides a computer-readable storage medium on which a computer program is stored, and which, when executed by a processor, performs the following steps:

[0054] In response to a task execution request carrying a target task, obtain the current terminal performance information;

[0055] Determine the edge-cloud collaborative decision value based on the current terminal performance information;

[0056] Based on edge-cloud collaborative decision values, the execution end for executing the target task is determined; where the execution end is the terminal and / or the cloud.

[0057] Fifthly, this application also provides a computer program product comprising a computer program that, when executed by a processor, performs the following steps:

[0058] In response to a task execution request carrying a target task, obtain the current terminal performance information;

[0059] Determine the edge-cloud collaborative decision value based on the current terminal performance information;

[0060] Based on edge-cloud collaborative decision values, the execution end for executing the target task is determined; where the execution end is the terminal and / or the cloud.

[0061] The aforementioned task processing method, apparatus, computer equipment, and storage medium, in response to a task execution request carrying a target task, determine the task information of the target task and the current terminal capability information of the terminal. Based on the task information and the current terminal capability information, a task processing terminal is determined. Based on the task processing terminal, the target task is processed. The task processing terminal is a terminal and / or the cloud; the task information includes at least one of task complexity and task privacy. Compared to the traditional "one-size-fits-all" task allocation mechanism, this application, when determining the task processing terminal, considers not only the current terminal capability information of the terminal but also the task complexity and task privacy of the target task. This fully utilizes terminal resources, considers the personalized needs of users, reduces the task inference burden on the cloud, and improves task execution efficiency and user satisfaction. Attached Figure Description

[0062] Figure 1 This is an application environment diagram of a task processing method provided in this embodiment;

[0063] Figure 2 This is a flowchart illustrating the first task processing method provided in this embodiment;

[0064] Figure 3 This is a schematic diagram illustrating the principle of terminal-cloud collaborative decision-making provided in this embodiment;

[0065] Figure 4 This is a flowchart illustrating the first method of determining the task processing terminal provided in this embodiment;

[0066] Figure 5 This is a flowchart illustrating the second method for determining the task processing terminal provided in this embodiment;

[0067] Figure 6 This is a flowchart illustrating the second task processing method provided in this embodiment;

[0068] Figure 7 This is a structural block diagram of a task processing device provided in this embodiment;

[0069] Figure 8 This is an internal structural diagram of the computer device provided in this embodiment. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0071] This application primarily focuses on edge-cloud collaboration technology. Edge-cloud collaboration leverages the local computing power and short-range privacy data processing advantages of edge devices, along with the powerful computing resources, massive amounts of data, and advanced model capabilities of the cloud, to achieve collaborative work between the two sides via network connection, thereby providing more efficient, intelligent, and personalized AI services and application experiences. Currently, edge-cloud collaboration solutions mainly focus on the collaborative deployment of model segmentation and the collaborative deployment of large and small models on both the edge and cloud.

[0072] Model segmentation deployment refers to dividing a complete AI model into multiple parts according to certain rules and strategies, and then deploying different models to the edge and cloud sides respectively according to the characteristics and advantages of edge devices and cloud sides, so that they can play the function of the whole model when working together, in order to optimize model performance, reduce resource consumption, and improve response speed.

[0073] The collaborative deployment of large and small models on the cloud involves deploying large AI models in the cloud, leveraging their powerful computing capabilities and rich knowledge representation to perform complex task processing and deep inference. Meanwhile, traditional solutions deploy small AI models on the edge to handle rapid local data collection, preliminary processing, and simple inference tasks. Based on this application, the current terminal capabilities, as well as the complexity and privacy of the target task, are fully considered to accurately determine the task processing end for handling the target task. This effectively utilizes terminal resources, reduces the burden on the cloud, and improves task processing efficiency.

[0074] The task processing method provided in this application embodiment can be applied to, for example, Figure 1 In the illustrated application environment, for example, in response to a task execution request carrying a target task, the terminal determines the task information of the target task and the terminal's current terminal capability information. Based on the task information and the current terminal capability information, the terminal determines the task processing end. Based on the task processing end, the target task is processed. The task information includes at least one of task complexity and task privacy; the task processing end is the terminal and / or the cloud.

[0075] Among them, the cloud refers to the cloud server, which is mainly used to deploy large cloud models, as well as store and deploy models that can be deployed on the edge; and the cloud also has a communication module for deployable models on the edge and for distributing task results.

[0076] A terminal refers to a user device that has communication functions. For example, it can be a mobile phone, a computer, or other terminal device; it can also be a smart watch, a smart bracelet, or other smart wearable device; or it can be an Internet of Things (IoT) terminal device.

[0077] In one embodiment, Figure 2 This is a flowchart illustrating a task processing method provided in an embodiment of this application, applied to... Figure 1 Taking the terminal in the example, the method includes the following steps:

[0078] S201, in response to a task execution request carrying a target task, determines the task information of the target task and the current terminal capability information of the terminal.

[0079] In this application, the target task refers to the task that needs to be performed, and the target task in this application mainly refers to AI (Artificial Intelligence) tasks. The task execution request refers to a message requesting the execution of the target task. The terminal refers to a user device with communication capabilities, such as a mobile phone, computer, smartwatch, smart bracelet, or IoT terminal device. Current terminal capability information refers to the terminal's capability information at the current moment, mainly used to measure the terminal's performance in processing the target task based on the edge-side model. Task information includes at least one of task complexity and task privacy. Task complexity refers to the complexity of the target task, which can also be understood as task difficulty. Task privacy refers to the user's privacy requirements regarding the target task.

[0080] Optionally, in this embodiment, the current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information. The current terminal hardware resource information primarily measures the terminal's basic hardware resources at the current moment. The current storage and computing resource information primarily measures the hardware resources, including storage and computing power, that can only be used at the current moment. The current communication resource information primarily measures the communication conditions for end-to-cloud collaboration at the current moment, as well as the cloud capabilities.

[0081] Optionally, in some embodiments, the current terminal hardware resource information includes at least one of the following: device model, current battery level, and current temperature.

[0082] Optionally, in some embodiments, the current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency. Optionally, in this embodiment, the terminal computing resources include at least one of the following: CPU, NPU (Neural Network Processing Unit), GPU (Graphics Processing Unit), and APU (Accelerated Processing Unit); the current resource utilization rate is obtained by weighted summation of the resource utilization rates of each terminal computing resource at the current moment.

[0083] Optionally, in some embodiments, the current communication resource information includes at least one of the following: current edge-side model capability, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capability, and currently cloud-stored models deployable on the edge. Optionally, in this embodiment, the current network quality is determined based on one or more of the following indicators: RSRP (Reference Signal Receiving Power / Reference Signal Received Power), SINR (Signal Interference Noise Ratio), RSRQ (Reference Signal Received Quality), and RSSI (Received Signal Strength Indication).

[0084] Optionally, in some embodiments, cloud computing power consumption information includes at least one of cloud computing power resource consumption and cloud computing power resource remaining amount.

[0085] Optionally, in some embodiments, task privacy includes at least one of privacy requirements and user privacy policy information. Privacy requirements include, but are not limited to, whether processing is supported only on the device side. User privacy policy information includes whether call data, location data, in-app usage data, cookies, logs, and other information are allowed to be uploaded to the cloud.

[0086] Optionally, one possible implementation of determining the task complexity of the target task in this embodiment is to select related tasks from historical tasks based on the task content of the target task. Processing information of the related tasks processed by the terminal is obtained. The task complexity of the target task is determined based on the processing information. The processing information includes, but is not limited to, at least one of the following indicators: processing time, computing resource consumption, and accuracy. Another possible implementation of determining the task complexity of the target task based on processing information in this embodiment is to determine a processing difficulty score based on the processing information. The processing difficulty score is used as the task complexity of the target task. Yet another possible implementation of selecting related tasks from historical tasks based on the task content of the target task in this embodiment is to match the task content of the target task with the task content of historical tasks, and select related tasks from historical tasks based on the matching results. The matching algorithm can be at least one of Euclidean distance, cosine similarity, Manhattan distance, Jaccard distance, word2Vec, Naive Bayes, etc.

[0087] Optionally, another possible implementation for determining the task complexity of the target task in this embodiment is to deploy an intelligent analysis model on the terminal. The target task is input into the intelligent analysis model, which then determines the task complexity. The intelligent analysis model can be a pre-trained neural network model.

[0088] Optionally, in this embodiment, one possible implementation for determining the task information of the target task and the current terminal capability information of the terminal in response to a task execution request carrying a target task is to determine the associated model of the target task in response to the task execution request carrying a target task. If the associated model includes a terminal-side model, the task information of the target task and the current terminal capability information of the terminal are determined. Here, the associated model refers to the AI ​​model used to process the target task. If the associated model does not include a terminal-side model, the target task is assigned to the cloud, and the cloud model processes the target task. Only if the associated model includes a terminal-side model is the task information of the target task and the current terminal capability information of the terminal further determined. In this embodiment, one possible implementation for determining the associated model of the target task in response to a task execution request carrying a target task is to obtain the task ID (Identification) of the target task in response to the task execution request carrying a target task. Based on a lookup table method, the associated model of the target task is determined using the task ID.

[0089] S202, determine the task processing terminal based on task information and current terminal capability information.

[0090] The task processing end refers to a terminal and / or the cloud. In this application, processing the target task specifically refers to the AI ​​model on the task processing end.

[0091] Optionally, in this embodiment, a capability analysis model is deployed on the terminal. The task information and the current terminal capability information are input into the capability analysis model, which then determines the task processing terminal. The capability analysis model can be a neural network model.

[0092] Optionally, in this embodiment, the task processing end can be a terminal, the cloud, or both.

[0093] S203, based on the task processing end, processes the target task.

[0094] As an optional implementation of this application, when the task processing end is a terminal, the target terminal-side model for processing the target task is determined based on the task content or task ID of the target task. The target task is then processed by the target terminal-side model. If the target terminal-side model is not deployed locally on the terminal, such as... Figure 3 As shown, it is necessary to download the target edge model from the cloud and deploy the target edge model locally.

[0095] Another optional implementation method of this application is as follows: Figure 3 As shown, the terminal's functional modules include sensors, a computing module, an edge-cloud decision-maker, an edge-side model deployment module, and a communication module. It can assign tasks based on collected data and the edge-cloud decision-maker, and can also complete simple tasks using edge-side models deployed on the edge. It can interact with the cloud server regarding allocation strategies, related data, and model downloads. The cloud server's functional modules include a large model deployment module (for deploying cloud models), an edge-adapted small model storage module (for storing edge-side models), and a communication module. It uses the large model to perform inference for the corresponding task based on data uploaded from the edge and can return the task processing results to the corresponding terminal. It can also distribute corresponding edge-side models according to the terminal's needs. When the task processing end consists of the terminal and the cloud, the terminal and the cloud process the target task based on the allocation strategy. The allocation strategy is determined by the terminal's edge-cloud decision-maker based on the target task. Optionally, the terminal performs preliminary processing on the target task based on the target edge-side model to obtain preliminary processing results. The terminal sends the preliminary processing results, the allocation strategy, and the model information (e.g., model ID) of the target edge-side model to the cloud. The cloud selects the target cloud model based on the model information of the target edge-side model. Based on the target cloud model, preliminary processing results, and allocation strategy, the target task is further processed to obtain the task processing result. One possible implementation for selecting the target cloud model based on the model information of the target terminal model is to determine candidate cloud models associated with (used in conjunction with) the target terminal model based on the model information. Based on the accuracy of each candidate cloud model, the candidate cloud model with the highest accuracy is selected as the target cloud model. For example, taking a mobile phone user activating a reservation function (target task) during a call, and the terminal needing to optimize the decision-making process and push the processing result to the user as an example, this embodiment is further explained: The target terminal model for processing the reservation function on the mobile phone is an ASR model and a preliminary intent recognition model. The target cloud model for processing the reservation function on the cloud is a deep intent recognition model. When processing the reservation function task, the mobile phone calls the terminal-deployed ASR model to perform real-time text transcription of the call content and inputs the transcription result into the preliminary intent recognition model to obtain the preliminary intent result, for example, "Reserving dinner for two people at 6 pm tomorrow; a certain restaurant." The mobile phone sends the preliminary intent recognition results to the cloud. Based on these results, the deep intent recognition model deployed on the cloud obtains the user's true intent: to reserve dinner for two people at a restaurant in Guangzhou, China, on [Date] at 18:00. The mobile phone then sends the reservation page to the call screen, and the user confirms to complete the reservation process.

[0096] The aforementioned task processing method, in response to a task execution request carrying a target task, determines the task information of the target task and the current terminal capability information of the terminal. Based on the task information and the current terminal capability information, a task processing terminal is determined. The target task is processed based on the task processing terminal. The task processing terminal is the terminal and / or the cloud; the task information includes at least one of task complexity and task privacy. Compared to the traditional "one-size-fits-all" task allocation mechanism, this application, when determining the task processing terminal, considers not only the current terminal capability information of the terminal but also the task complexity and task privacy of the target task. This fully utilizes terminal resources, considers user privacy needs and security, reduces the task inference burden on the cloud, and improves task execution efficiency and user satisfaction.

[0097] In one embodiment, to more accurately determine the task processing end, such as Figure 4 As shown, an optional implementation of S202 includes:

[0098] S401, determine the first evaluation value based on the current terminal capability information.

[0099] The first evaluation value refers to the evaluation score determined based on the current terminal capability information, which is used to comprehensively evaluate the current terminal capability.

[0100] As an optional implementation of this application, the current terminal capability information is input into the evaluation model, and the evaluation model outputs a first evaluation value.

[0101] Another optional implementation of this application embodiment is as follows: A first score is determined based on the current terminal hardware resource information. A second score is determined based on the current computing resource information. A third score is determined based on the current communication resource information. A first evaluation value is determined based on the first, second, and third scores. In this embodiment, an optional implementation of determining the first evaluation value based on the first, second, and third scores is as follows: A target end-side model associated with the target task is determined. A first weighting coefficient for the first score, a second weighting coefficient for the second score, and a third weighting coefficient for the third score are determined based on the target end-side model. A weighted sum is performed on the first, second, and third scores based on the first, second, and third weighting coefficients to obtain the terminal capability score. The first evaluation value is determined based on the terminal capability score. In this embodiment, an optional implementation of determining the first evaluation value based on the terminal capability score is as follows: The accuracy of the target end-side model is determined. The product of the accuracy and the terminal capability score is used as the first evaluation value. Optionally, in this embodiment, the first evaluation value can be determined according to the following formula: P j =M j ·B j Among them, P jIndicates the first evaluation value; M J This indicates the accuracy of the target end-side model. B j This represents the terminal capability score. In this embodiment, the terminal capability score can be determined according to the following formula: in, F represents the first weighting coefficient; ue ω represents the first score; ω represents the second weighting coefficient; F re σ represents the second score; F represents the third weighting coefficient. net This represents the third score. For example, it obtains the terminal's current hardware resource information (current phone model XX, 16GB+512GB memory, current battery 40%, current temperature 42°C), current computing resource information (current CPU utilization 35%, current NPU utilization 15%, current GPU utilization 10%, and current RAM utilization 40%), and current communication resource information (current uplink peak rate 5Gbit / s, current downlink peak rate 10Gbit / s, current network transmission latency 0.5s); the accuracy of the target edge model deployed on the terminal is 80%; task privacy information includes user-allowed call data, location, in-application usage data, cookies, logs, etc. The data processed by the edge model is related to natural language, and the second score corresponds to the current computing resource information. And the current terminal hardware resource information F ue Corresponding sub-feature (current battery level F) pow Current temperature F Tem Current communication resource information F net Corresponding sub-features (current uplink and downlink peak rates F) DR Current computational delay F delay The calculation is performed using the average aggregation method; where (N, M) takes the following values: Tem(40, 70); CPU / GPU / NPU(40, 75); RAM(50, 70); current terminal battery level. Peak rate Current transmission delay S set User allows cloud-based access, score -1;

[0102] Based on the terminal's first, second, and third scores, the calculation method for the first evaluation value is as follows:

[0103] calculate The first, second, and third weighting coefficients are determined based on the target-side model. Specifically, a lookup table method can be used to retrieve the first, second, and third weighting coefficients based on the target-side model.

[0104] S402, Determine the second evaluation value based on the task complexity.

[0105] The second evaluation value refers to the evaluation score determined based on task complexity, used to assess the complexity of the target task.

[0106] As an optional implementation of this application, the task complexity is input into the evaluation model, and the evaluation model outputs a second evaluation value.

[0107] Another optional implementation of this application involves determining a second evaluation value based on the task complexity and the complexity interval to which the task complexity belongs. Each complexity interval has a corresponding evaluation value. The evaluation value corresponding to the complexity interval to which the task complexity belongs is used as the second evaluation value.

[0108] S403, determine the third evaluation value based on mission privacy.

[0109] The third evaluation value refers to the score determined based on task privacy and used to assess user privacy needs.

[0110] As an optional implementation of this application, task privacy is input into the evaluation model, and the evaluation model outputs a third evaluation value.

[0111] Another optional implementation method of this application is to determine the privacy level based on the privacy requirements in the task privacy and the user privacy terms. A third evaluation value is then determined based on the privacy level. If the privacy level is Level 1, it indicates no privacy requirements, and the third evaluation value is -1. If the privacy level is Level 2, it indicates a certain need for privacy protection, and the third evaluation value is 0. As the privacy level increases, the corresponding third evaluation value will gradually increase.

[0112] S404, determine the task processing end based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0113] Optionally, in this embodiment, the first evaluation value, the second evaluation value, and the third evaluation value are weighted and summed, and the task processing end is determined based on the summation result.

[0114] In this embodiment, a first evaluation value is determined based on the current terminal capability information. A second evaluation value is determined based on task complexity. A third evaluation value is determined based on task privacy. The task processing terminal is determined based on the first, second, and third evaluation values. This embodiment quantifies the terminal capability information, task complexity, and task privacy, and determines the task processing terminal based on the quantized results, effectively increasing the accuracy of the determined task processing terminal.

[0115] In one embodiment, to further improve the accuracy of the confirmed task processing endpoint, such as... Figure 5 As shown, one optional implementation of S404 includes:

[0116] S501, the first evaluation value, the second evaluation value, and the third evaluation value are weighted and summed to obtain the total evaluation score.

[0117] Optionally, in this embodiment, based on the target-end cloud collaborative decision-making model, the first evaluation value, the second evaluation value, and the third evaluation value are weighted and summed to obtain the total evaluation score. Specifically, the target-end cloud collaborative decision-making model includes a first aggregation weight corresponding to the first evaluation value, a second aggregation weight corresponding to the second evaluation value, and a third aggregation weight corresponding to the third evaluation value. Based on the first aggregation weight, the second aggregation weight, and the third aggregation weight, the first evaluation value, the second evaluation value, and the third evaluation value are weighted and summed to obtain the total evaluation score. Optionally, the total evaluation score can be determined using the following formula: Where P represents the total assessment value; P i Let represent the i-th evaluation value; α represents the aggregate weight corresponding to the i-th evaluation value. Specifically, the total evaluation score can be obtained using the following formula: P = X · M j ·B j +Y·P k +Z·(S k ·S set ); where X represents the first aggregation weight; Y represents the second aggregation weight; Z represents the third aggregation weight; P k Indicates the second evaluation value; S k ·S set Indicates the third evaluation value; S k Indicates privacy requirements; S set This indicates information regarding privacy terms.

[0118] Optionally, in this embodiment, after processing the target task on the task processing end, one possible implementation of the task processing method is to obtain the processing effect of the target task processed by the task processing end. If the processing effect does not meet preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized. The processing effect can be determined based on at least one of the following indicators: accuracy, inference speed, recall, and power consumption. For example, taking the accuracy of processing the target task as the processing effect, if the accuracy is lower than a preset threshold, the target-end cloud collaborative decision-making model is optimized. Specifically, the first aggregate weight, the second aggregate weight, and the third aggregate weight of the target-end cloud collaborative decision-making model are optimized to improve the accuracy of subsequent decisions made by the target-end cloud collaborative decision-making model.

[0119] Optionally, one possible implementation of the task processing method in this embodiment is to train an initial end-to-cloud collaborative decision-making model to obtain a target end-to-cloud collaborative decision-making model based on the terminal's historical terminal capability information (e.g., including at least one of historical terminal hardware resource information, historical terminal storage and computing resource information, historical communication resource information, and historical cloud computing power consumption information), historical task information, and offline datasets. The purpose of training is to continuously optimize the aggregation weights of the initial end-to-cloud collaborative decision-making model to obtain a first aggregation weight, a second aggregation weight, and a third aggregation weight. The offline dataset includes the terminal model and corresponding terminal battery level, memory usage, computing resource operation frequency and occupancy thresholds, historical tasks and their complexity, task privacy requirements, privacy terms, etc. It should be noted that the training process of the initial end-to-cloud collaborative decision-making model can be executed on the terminal or in the cloud. The target end-to-cloud collaborative decision-making model needs to be deployed on the terminal. In this embodiment, one optional implementation for training an initial end-to-cloud collaborative decision-making model to obtain a target end-to-cloud collaborative decision-making model based on the terminal's historical terminal capability information, historical task information, and offline dataset is as follows: Another optional implementation involves training the initial end-to-cloud collaborative decision-making model to obtain candidate end-to-cloud collaborative decision-making models based on the terminal's historical terminal capability information, historical task information, and offline dataset. These candidate models are then sent to the cloud. The cloud, based on the candidate models sent from multiple managed terminals, aggregates the weights of the multiple candidate end-to-cloud collaborative decision-making models to obtain the target end-to-cloud collaborative decision-making model. Specifically, the average of the aggregate weights corresponding to each evaluation value of the multiple candidate end-to-cloud collaborative decision-making models is calculated, and the average of the aggregate weights corresponding to each evaluation value is used as the aggregate weight corresponding to that evaluation value in the target end-to-cloud collaborative decision-making model. The cloud then sends the target end-to-cloud collaborative decision-making model back to the terminal. For example, multiple terminals receive an initial end-to-cloud collaborative decision-making model from the server, along with their historical terminal capability information, historical task information, and offline datasets. These terminals then train the initial model to obtain candidate end-to-cloud collaborative decision-making models. The multiple terminals upload the candidate models or the aggregated weights corresponding to their evaluation values ​​to the cloud. The cloud performs aggregation to obtain the target end-to-cloud collaborative decision-making model and determines whether the training termination condition is met. If the condition is met, the target end-to-cloud collaborative decision-making model is distributed to the corresponding multiple terminals.

[0120] S502, determine the task processing end based on the relationship between the total evaluation score and the judgment threshold.

[0121] Optionally, this embodiment determines the associated model for processing the target task. If the associated model only includes a terminal-side model and the total evaluation score is greater than the judgment threshold, the terminal is used as the task processing terminal, and the associated model is used as the target terminal-side model. If the associated model only includes a terminal-side model and the total evaluation score is not greater than the judgment threshold, the cloud is used as the task processing terminal, and a target cloud model for processing the target task is determined. The cloud then processes the target task based on the target cloud model. If the associated model includes both a terminal-side model and a cloud model, and the total evaluation score is greater than the judgment threshold, both the terminal and the cloud are used as task processing terminals, with the terminal-side model used as the target terminal-side model and the cloud model used as the target cloud model. The target task is processed based on the target terminal-side model of the terminal and the target cloud model of the cloud. If the associated model includes both a terminal-side model and a cloud model, and the total evaluation score is not greater than the judgment threshold, the cloud is used as the task processing terminal, and a target cloud model for processing the target task is determined. The cloud then processes the target task based on the target cloud model.

[0122] In this embodiment, the first, second, and third evaluation values ​​are weighted and summed to obtain a total evaluation score. The task processing terminal is determined based on the relationship between the total evaluation score and a judgment threshold. This embodiment further improves the accuracy of the determined task processing terminal.

[0123] In one embodiment, Figure 6 As shown, an optional implementation of a task processing method includes:

[0124] S601, based on the terminal's historical terminal capability information, historical task information, and offline dataset, trains the initial terminal-cloud collaborative decision-making model to obtain the target terminal-cloud collaborative decision-making model. The offline dataset includes the terminal model and corresponding calculation thresholds for terminal battery power, memory usage, computing resource operation frequency and occupancy, historical tasks and their complexity, task privacy requirements, and privacy terms.

[0125] S602, in response to a task execution request carrying a target task, determines the association model of the target task.

[0126] S603, when the associated model includes the end-side model, select associated tasks from historical tasks based on the task content of the target task.

[0127] S604, Obtain processing information for related tasks processed by the terminal.

[0128] S605, Based on the processed information, determine the task complexity of the target task.

[0129] S606, Obtain the task privacy information of the target task and the current terminal capability information of the terminal. Task privacy includes at least one of privacy requirements and user privacy terms. Current terminal capability information includes at least one of current terminal hardware resource information, current in-memory computing resource information, and current communication resource information. Current terminal hardware resource information includes at least the device model, current battery level, and current temperature. Current in-memory computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization, current terminal operating frequency, current memory size, current memory usage, and current thread switching frequency. Current communication resource information includes at least one of the following: current edge-side model capability, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud model capability, and currently cloud-stored models deployable on the edge.

[0130] S607 determines the first score based on the current terminal hardware resource information.

[0131] S608, determine the second score based on the current computing resource information.

[0132] S609, determine the third score based on the current communication resource information.

[0133] S610, Determine the target end-side model associated with the target task.

[0134] S611, based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score.

[0135] S612, based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, performs a weighted summation of the first score, the second score, and the third score to obtain the terminal capability score.

[0136] S613, determine the accuracy of the target end-side model.

[0137] S614 uses the product of accuracy and terminal capability score as the first evaluation value.

[0138] S615, determine the second evaluation value based on the task complexity.

[0139] S616, determine the third evaluation value based on mission privacy.

[0140] S617, based on the target cloud collaborative decision-making model, calculates the weighted sum of the first evaluation value, the second evaluation value, and the third evaluation value to obtain the total evaluation score.

[0141] S618, determine the task processing end based on the relationship between the total evaluation score and the judgment threshold.

[0142] S619, obtain the processing effect of the target task by the task processing end.

[0143] S620 optimizes the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value when the processing effect does not meet the preset conditions.

[0144] The task processing method of this embodiment, in response to a task execution request carrying a target task, determines the task information of the target task and the current terminal capability information of the terminal. Based on the task information and the current terminal capability information, a task processing terminal is determined. The target task is processed based on the task processing terminal. The task processing terminal is the terminal and / or the cloud; the task information includes at least one of task complexity and task privacy. Compared to the traditional "one-size-fits-all" task allocation mechanism, this application, when determining the task processing terminal, considers not only the current terminal capability information of the terminal but also the task complexity and task privacy of the target task. This fully utilizes terminal resources, considers the personalized needs of users, reduces the task inference burden on the cloud, and improves task execution efficiency and user satisfaction.

[0145] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0146] Based on the same inventive concept, this application also provides a task processing apparatus for implementing the task processing method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more task processing apparatus embodiments provided below can be found in the limitations of the task processing method described above, and will not be repeated here.

[0147] In one embodiment, by Figure 7 A structural block diagram of a task processing apparatus in one embodiment is shown. Figure 7 As shown, a task processing device 1 is provided, which includes: a first determining module 10, a second determining module 20, and a processing module 30, wherein:

[0148] The first determining module 10 is used to determine the task information of the target task and the current terminal capability information of the terminal in response to a task execution request carrying a target task; wherein the task information includes at least one of task complexity and task privacy.

[0149] The second determining module 20 is used to determine the task processing end based on the task information and the current terminal capability information; wherein, the task processing end is the terminal and / or the cloud.

[0150] The processing module 30 is used to process the target task based on the task processing terminal.

[0151] In one embodiment, the upper Figure 7 The second determining module 20 is also specifically used for:

[0152] Based on the current terminal capability information, determine the first evaluation value;

[0153] Determine the second evaluation value based on the task complexity;

[0154] A third evaluation value is determined based on task privacy.

[0155] The task processing end is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0156] In one embodiment, the upper Figure 7 The second determining module 20 is also specifically used for:

[0157] The first, second, and third assessment values ​​are weighted and summed to obtain the total assessment score.

[0158] The task processing end is determined based on the relationship between the total evaluation score and the judgment threshold.

[0159] In one embodiment, the current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information.

[0160] In one embodiment, the upper Figure 7 The second determining module 20 is also specifically used for:

[0161] The first score is determined based on the current terminal hardware resource information;

[0162] The second score is determined based on the current computing resource information;

[0163] The third score is determined based on the current communication resource information;

[0164] The first evaluation value is determined based on the first score, the second score, and the third score.

[0165] In one embodiment, the upper Figure 7 The second determining module 20 is also specifically used for:

[0166] Determine the target endpoint model associated with the target task;

[0167] Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score;

[0168] Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score;

[0169] The first evaluation value is determined based on the terminal capability score.

[0170] In one embodiment, the upper Figure 7 The second determining module 20 is also specifically used for:

[0171] Determine the accuracy of the target end-side model;

[0172] The product of accuracy and terminal capability score is used as the first evaluation value.

[0173] In one embodiment, the upper Figure 7 The first determining module 10 in the module is also specifically used for:

[0174] Based on the task content of the target task, select related tasks from the historical tasks;

[0175] Obtain processing information for related tasks processed by the terminal;

[0176] Based on the processed information, determine the task complexity of the target task.

[0177] In one embodiment, task privacy includes at least one of privacy requirements and user privacy terms information.

[0178] In one embodiment, the current terminal hardware resource information includes at least one of the following: device model, current battery level, and current temperature;

[0179] Current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency;

[0180] Current communication resource information includes at least one of the following: current edge-side model capabilities, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capabilities, and currently cloud-stored models that can be deployed on the edge.

[0181] In one embodiment, the upper Figure 7 The first determining module 10 in the module is also specifically used for:

[0182] In response to a task execution request carrying a target task, determine the association model of the target task;

[0183] When the association model includes the edge model, the task information of the target task and the current terminal capability information of the terminal are determined.

[0184] In one embodiment, a task processing device 1 further includes:

[0185] The acquisition module is used to acquire the processing effect of the target task by the task processing terminal;

[0186] The optimization module is used to optimize the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value when the processing effect does not meet the preset conditions.

[0187] Each module in the aforementioned task processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.

[0188] In one embodiment, a computer device is provided, which may be a platform-side device, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores task processing information. The network interface communicates with an external user via a network connection. When the computer program is executed by the processor, it implements a task processing method.

[0189] Those skilled in the art will understand that Figure 8The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specifically, the computer device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0190] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0191] In response to a task execution request carrying a target task, the system determines the task information of the target task and the current terminal capability information of the terminal; wherein the task information includes at least one of task complexity and task privacy.

[0192] Based on the task information and the current terminal capability information, determine the task processing terminal; wherein, the task processing terminal is the terminal and / or the cloud.

[0193] The target task is processed based on the task processing terminal.

[0194] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the task processing terminal based on task information and current terminal capability information, including:

[0195] Based on the current terminal capability information, determine the first evaluation value;

[0196] Determine the second evaluation value based on the task complexity;

[0197] A third evaluation value is determined based on task privacy.

[0198] The task processing end is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0199] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a task processing end based on a first evaluation value, a second evaluation value, and a third evaluation value, including:

[0200] The first, second, and third assessment values ​​are weighted and summed to obtain the total assessment score.

[0201] The task processing end is determined based on the relationship between the total evaluation score and the judgment threshold.

[0202] In one embodiment, when the processor executes the computer program, it further implements the following steps: the current terminal capability information includes at least one of the following: current terminal hardware resource information, current in-memory computing resource information, current communication resource information, and cloud computing power consumption information.

[0203] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a first evaluation value based on current terminal capability information, including:

[0204] The first score is determined based on the current terminal hardware resource information;

[0205] The second score is determined based on the current computing resource information;

[0206] The third score is determined based on the current communication resource information;

[0207] The first evaluation value is determined based on the first score, the second score, and the third score.

[0208] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a first evaluation value based on a first score, a second score, and a third score, including:

[0209] Determine the target endpoint model associated with the target task;

[0210] Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score;

[0211] Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score;

[0212] The first evaluation value is determined based on the terminal capability score.

[0213] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a first evaluation value based on the terminal capability score, including:

[0214] Determine the accuracy of the target end-side model;

[0215] The product of accuracy and terminal capability score is used as the first evaluation value.

[0216] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the task complexity of the target task, including:

[0217] Based on the task content of the target task, select related tasks from the historical tasks;

[0218] Obtain processing information for related tasks processed by the terminal;

[0219] Based on the processed information, determine the task complexity of the target task.

[0220] In one embodiment, when the processor executes a computer program, it further performs the following steps: task privacy includes at least one of privacy requirements and user privacy terms information.

[0221] In one embodiment, when the processor executes the computer program, it further performs the following steps: the current terminal hardware resource information includes at least one of the following: model, current battery level, and current temperature;

[0222] Current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency;

[0223] Current communication resource information includes at least one of the following: current edge-side model capabilities, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capabilities, and currently cloud-stored models that can be deployed on the edge.

[0224] In one embodiment, when the processor executes the computer program, it further performs the following steps: in response to a task execution request carrying a target task, determining task information of the target task and current terminal capability information of the terminal, including:

[0225] In response to a task execution request carrying a target task, determine the association model of the target task;

[0226] When the association model includes the edge model, the task information of the target task and the current terminal capability information of the terminal are determined.

[0227] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0228] Obtain the processing effect of the target task from the task processing terminal;

[0229] If the processing effect does not meet the preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized.

[0230] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0231] In response to a task execution request carrying a target task, the system determines the task information of the target task and the current terminal capability information of the terminal; wherein the task information includes at least one of task complexity and task privacy.

[0232] Based on the task information and the current terminal capability information, determine the task processing terminal; wherein, the task processing terminal is the terminal and / or the cloud.

[0233] The target task is processed based on the task processing terminal.

[0234] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the task processing terminal based on task information and current terminal capability information, including:

[0235] Based on the current terminal capability information, determine the first evaluation value;

[0236] Determine the second evaluation value based on the task complexity;

[0237] A third evaluation value is determined based on task privacy.

[0238] The task processing end is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0239] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a task processing end based on a first evaluation value, a second evaluation value, and a third evaluation value, including:

[0240] The first, second, and third assessment values ​​are weighted and summed to obtain the total assessment score.

[0241] The task processing end is determined based on the relationship between the total evaluation score and the judgment threshold.

[0242] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: the current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information.

[0243] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining a first evaluation value based on current terminal capability information, including:

[0244] The first score is determined based on the current terminal hardware resource information;

[0245] The second score is determined based on the current computing resource information;

[0246] The third score is determined based on the current communication resource information;

[0247] The first evaluation value is determined based on the first score, the second score, and the third score.

[0248] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a first evaluation value based on a first score, a second score, and a third score, including:

[0249] Determine the target endpoint model associated with the target task;

[0250] Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score;

[0251] Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score;

[0252] The first evaluation value is determined based on the terminal capability score.

[0253] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a first evaluation value based on the terminal capability score, including:

[0254] Determine the accuracy of the target end-side model;

[0255] The product of accuracy and terminal capability score is used as the first evaluation value.

[0256] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining the task complexity of the target task, including:

[0257] Based on the task content of the target task, select related tasks from the historical tasks;

[0258] Obtain processing information for related tasks processed by the terminal;

[0259] Based on the processed information, determine the task complexity of the target task.

[0260] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: task privacy includes at least one of privacy requirements and user privacy terms information.

[0261] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: the current terminal hardware resource information includes at least one of the following: model, current battery level, and current temperature;

[0262] Current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency;

[0263] Current communication resource information includes at least one of the following: current edge-side model capabilities, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capabilities, and currently cloud-stored models that can be deployed on the edge.

[0264] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: in response to a task execution request carrying a target task, determining task information of the target task and current terminal capability information of the terminal, including:

[0265] In response to a task execution request carrying a target task, determine the association model of the target task;

[0266] When the association model includes the edge model, the task information of the target task and the current terminal capability information of the terminal are determined.

[0267] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0268] Obtain the processing effect of the target task from the task processing terminal;

[0269] If the processing effect does not meet the preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized.

[0270] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0271] In response to a task execution request carrying a target task, the system determines the task information of the target task and the current terminal capability information of the terminal; wherein the task information includes at least one of task complexity and task privacy.

[0272] Based on the task information and the current terminal capability information, determine the task processing terminal; wherein, the task processing terminal is the terminal and / or the cloud.

[0273] The target task is processed based on the task processing terminal.

[0274] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the task processing terminal based on task information and current terminal capability information, including:

[0275] Based on the current terminal capability information, determine the first evaluation value;

[0276] Determine the second evaluation value based on the task complexity;

[0277] A third evaluation value is determined based on task privacy.

[0278] The task processing end is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

[0279] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a task processing end based on a first evaluation value, a second evaluation value, and a third evaluation value, including:

[0280] The first, second, and third assessment values ​​are weighted and summed to obtain the total assessment score.

[0281] The task processing end is determined based on the relationship between the total evaluation score and the judgment threshold.

[0282] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: the current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information.

[0283] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining a first evaluation value based on current terminal capability information, including:

[0284] The first score is determined based on the current terminal hardware resource information;

[0285] The second score is determined based on the current computing resource information;

[0286] The third score is determined based on the current communication resource information;

[0287] The first evaluation value is determined based on the first score, the second score, and the third score.

[0288] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a first evaluation value based on a first score, a second score, and a third score, including:

[0289] Determine the target endpoint model associated with the target task;

[0290] Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score;

[0291] Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score;

[0292] The first evaluation value is determined based on the terminal capability score.

[0293] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining a first evaluation value based on the terminal capability score, including:

[0294] Determine the accuracy of the target end-side model;

[0295] The product of accuracy and terminal capability score is used as the first evaluation value.

[0296] In one embodiment, when the computer program is executed by a processor, it further performs the following steps: determining the task complexity of the target task, including:

[0297] Based on the task content of the target task, select related tasks from the historical tasks;

[0298] Obtain processing information for related tasks processed by the terminal;

[0299] Based on the processed information, determine the task complexity of the target task.

[0300] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: task privacy includes at least one of privacy requirements and user privacy terms information.

[0301] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: the current terminal hardware resource information includes at least one of the following: model, current battery level, and current temperature;

[0302] Current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency;

[0303] Current communication resource information includes at least one of the following: current edge-side model capabilities, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capabilities, and currently cloud-stored models that can be deployed on the edge.

[0304] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: in response to a task execution request carrying a target task, determining task information of the target task and current terminal capability information of the terminal, including:

[0305] In response to a task execution request carrying a target task, determine the association model of the target task;

[0306] When the association model includes the edge model, the task information of the target task and the current terminal capability information of the terminal are determined.

[0307] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0308] Obtain the processing effect of the target task from the task processing terminal;

[0309] If the processing effect does not meet the preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized.

[0310] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these. The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0311] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A task processing method, characterized in that, Applied to a terminal, the method includes: In response to a task execution request carrying a target task, the task information of the target task and the current terminal capability information of the terminal are determined; wherein, the task information includes at least one of task complexity and task privacy; Based on the task information and the current terminal capability information, a task processing terminal is determined; wherein, the task processing terminal is the terminal and / or the cloud. The target task is processed based on the task processing terminal.

2. The method according to claim 1, characterized in that, Based on the task information and the current terminal capability information, the task processing terminal is determined, including: Based on the current terminal capability information, a first evaluation value is determined; Determine the second evaluation value based on the task complexity; Based on the stated task privacy, a third evaluation value is determined; The task processing terminal is determined based on the first evaluation value, the second evaluation value, and the third evaluation value.

3. The method according to claim 2, characterized in that, The step of determining the task processing terminal based on the first evaluation value, the second evaluation value, and the third evaluation value includes: The first evaluation value, the second evaluation value, and the third evaluation value are weighted and summed to obtain the total evaluation score. The task processing terminal is determined based on the relationship between the total evaluation score and the judgment threshold.

4. The method according to claim 2, characterized in that, The current terminal capability information includes at least one of the following: current terminal hardware resource information, current storage and computing resource information, current communication resource information, and cloud computing power consumption information.

5. The method according to claim 4, characterized in that, Determining the first evaluation value based on the current terminal capability information includes: Based on the current terminal hardware resource information, a first score is determined; The second score is determined based on the current storage and computing resource information. The third score is determined based on the current communication resource information. The first evaluation value is determined based on the first score, the second score, and the third score.

6. The method according to claim 5, characterized in that, The determination of the first evaluation value based on the first score, the second score, and the third score includes: Determine the target edge model associated with the target task; Based on the target end-side model, determine the first weight coefficient of the first score, the second weight coefficient of the second score, and the third weight coefficient of the third score; Based on the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, the first score, the second score, and the third score are weighted and summed to obtain the terminal capability score. The first evaluation value is determined based on the terminal capability score.

7. The method according to claim 6, characterized in that, Determining the first evaluation value based on the terminal capability score includes: Determine the accuracy of the target end-side model; The product of the accuracy rate and the terminal capability score is used as the first evaluation value.

8. The method according to claim 1, characterized in that, Determining the task complexity of the target task includes: Based on the task content of the target task, select related tasks from the historical tasks; Obtain processing information of the terminal for the associated task; Based on the processing information, the task complexity of the target task is determined.

9. The method according to claim 2, characterized in that, The privacy of the task includes at least one of the following: privacy requirements and user privacy terms information.

10. The method according to claim 4, characterized in that, The current terminal hardware resource information includes at least one of the following: device model, current battery level, and current temperature; The current computing resource information includes at least one of the following: current terminal computing resource size, current resource utilization rate, current terminal operating frequency, current memory size, current memory usage rate, and current thread switching frequency; The current communication resource information includes at least one of the following: current edge-side model capability, current network quality, current network bandwidth, current communication latency, current data transmission speed, current network computing power consumption experience value, current computing latency, current cloud-based model capability, and currently cloud-stored models that can be deployed on the edge.

11. The method according to claim 1, characterized in that, The step of responding to a task execution request carrying a target task and determining the task information of the target task and the current terminal capability information of the terminal includes: In response to a task execution request carrying a target task, determine the association model of the target task; When the association model includes an end-side model, the task information of the target task and the current terminal capability information of the terminal are determined.

12. The method according to claim 2, characterized in that, After processing the target task based on the task processing terminal, the method further includes: Obtain the processing effect of the task processing terminal on the target task; If the processing effect does not meet the preset conditions, the aggregate weights of the first evaluation value, the second evaluation value, and the third evaluation value are optimized.

13. A task processing device, characterized in that, Configured in a terminal, the device includes: The first determining module is configured to, in response to a task execution request carrying a target task, determine the task information of the target task and the current terminal capability information of the terminal; wherein the task information includes at least one of task complexity and task privacy. The second determining module is used to determine the task processing terminal based on the task information and the current terminal capability information; wherein the task processing terminal is the terminal and / or the cloud. The processing module is used to process the target task based on the task processing terminal.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 12.

16. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 12.

Images