A computing task processing method, apparatus, device, and readable storage medium

Abstract

This invention provides a computing task processing method, apparatus, device, and readable storage medium, relating to the field of wireless local area networks (WLANs). The method includes receiving a computing task allocation frame sent by a wireless access point, the computing task allocation frame carrying a field indicating the splitting position of the computing task; splitting the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks; and calculating the calculation results of the sub-computing tasks located before the first splitting position to obtain a first calculation result. This invention determines the splitting position of the computing task for each terminal device through the wireless access point. The terminal device only needs to calculate the computing tasks before the first splitting position; the remaining split computing tasks are calculated with the assistance of the wireless access point, thereby realizing the computing task processing function of the wireless access point in the WLAN and ensuring the energy efficiency performance of the terminal device in processing computing tasks.

Description

Technical Field

[0001] This invention relates to the field of wireless local area networks, and more specifically, to a computing task processing method, apparatus, device, and readable storage medium. Background Technology

[0002] With the deepening application of artificial intelligence technology, future terminal devices and wireless access points in wireless LANs will possess intelligent computing capabilities, enabling them to complete intelligent computational inference tasks. When facing complex intelligent computing tasks, the learning framework can be split, allowing the wireless access point and terminal device to collaborate and jointly process the machine learning inference task. In this scenario, the wireless access point selects the task splitting position for the computational task of the wireless LAN terminal device. The terminal device only executes the sub-computation task before the splitting position and transmits the machine learning model parameters after the split layer to the wireless access point via the wireless LAN communication standard protocol. The wireless access point then assists in completing the subsequent sub-computation tasks.

[0003] Considering the interconnected and coupled relationship between the selection of the splitting location for each terminal device and the allocation of uplink transmission resources in a multi-terminal device scenario, which jointly determines the energy efficiency of the terminal devices in completing machine learning computation tasks, it is necessary to propose a technical solution to address the intelligent computing needs of WLAN terminal devices. This solution should combine the characteristics of the splitting learning framework with the communication transmission characteristics of WLANs to enable multiple terminal devices and wireless access points to jointly perform intelligent computing tasks based on the splitting learning framework within a WLAN, thereby improving the efficiency of machine learning intelligent processing tasks in WLAN devices. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, device, and readable storage medium for processing computational tasks, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0005] In a first aspect, this application provides a computing task processing method applied to a terminal device, the method comprising:

[0006] Receive a computing task allocation frame sent by a wireless access point, wherein the computing task allocation frame carries a field for indicating the splitting position of the computing task;

[0007] The computing task is split into multiple sub-computing tasks based on the computing task allocation frame.

[0008] The calculation results of the sub-computation tasks located before the first split position are calculated to obtain the first calculation result.

[0009] Secondly, this application also provides a computing task processing method applied to a wireless access point, the method comprising:

[0010] A computing task allocation frame is sent to the terminal device, the computing task allocation frame carrying a field for indicating the split position of the computing task;

[0011] The terminal device receives a computing task allocation confirmation frame, which indicates that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

[0012] Thirdly, this application also provides a computing task processing apparatus, comprising:

[0013] The first receiving unit is used to receive a computing task allocation frame sent by a wireless access point, wherein the computing task allocation frame carries a field for indicating the splitting position of the computing task.

[0014] The splitting unit is used to split the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks;

[0015] The first computing unit is used to calculate the computing results of the sub-computation tasks located before the first split position, and obtain the first computing result.

[0016] Fourthly, this application also provides a computing task processing apparatus, comprising:

[0017] The first sending unit is used to send a computing task allocation frame to the terminal device, wherein the computing task allocation frame carries a field for indicating the split position of the computing task;

[0018] The first receiving unit is configured to receive a computing task allocation confirmation frame sent by the terminal device. The computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

[0019] Fifthly, this application also provides a computing task processing device, comprising:

[0020] memory for storing computer programs;

[0021] A processor is used to implement the steps of the computing task processing method when executing the computer program.

[0022] Sixthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described computational task processing method.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention determines the splitting position of the computing task for each terminal device through the wireless access point. The terminal device only needs to calculate the computing task before the first splitting position, and the remaining split computing tasks will be calculated by the wireless access point with the assistance of the wireless access point. This realizes the computing task processing function of the wireless access point in the wireless local area network and ensures the energy efficiency performance of the terminal device in processing computing tasks.

[0025] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of the system structure consisting of the terminal device and the wireless access point described in this embodiment of the invention;

[0028] Figure 2 This is a schematic diagram of the computational task processing method described in an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of the structure of the computation task query frame described in an embodiment of the present invention;

[0030] Figure 4 This is a schematic diagram of the structure of the computing task request frame described in an embodiment of the present invention;

[0031] Figure 5 This is an interactive schematic diagram of the computing task processing method described in an embodiment of the present invention;

[0032] Figure 6 This is a schematic diagram of the structure of the computing task allocation frame described in an embodiment of the present invention;

[0033] Figure 7 This is a schematic diagram of the computing task processing device described in an embodiment of the present invention;

[0034] Figure 8 This is a schematic diagram of another computing task processing device described in an embodiment of the present invention;

[0035] Figure 9 This is a schematic diagram of the computing task processing device described in an embodiment of the present invention.

[0036] Marked in the image:

[0037] 10. First receiving unit; 20. Splitting unit; 30. First computing unit; 40. First transmitting unit; 50. First receiving unit; 800. Computing task processing device; 801. Processor; 802. Memory; 803. Multimedia component; 804. I / O interface; 805. Communication component. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0039] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0040] See Figure 1 This is a schematic diagram of a system architecture in a wireless local area network (WLAN) to which terminal devices (Stations, STAs) and wireless access points (Access Points, APs) collaborate to complete computational tasks based on a split learning framework. For ease of description, STAs will be used to represent terminal devices and APs will be used to represent wireless access points in the following text.

[0041] The system comprises a single AP and multiple associated STAs. STA1, STA2, and STA3 transmit information with the AP via Radio Resource Units RU1, RU2, and RU3, respectively. Split learning refers to decomposing a complete computational task into multiple sub-computational tasks, which are deployed in different locations, such as STAs and APs in a wireless LAN. Each STA performs computation only on the received sub-computational task and then transmits the result to the next STA. The next STA receives the result and continues computation based on its locally deployed sub-computational tasks, then transmits the result to the next STA again, until the entire computational task is completed.

[0042] Considering that computational tasks include various types of tasks, this application takes machine learning tasks as an example. The computational tasks may be different in different scenarios, but the steps involved are basically the same, and no special restrictions are made here.

[0043] Example 1:

[0044] The technical solutions provided by the embodiments of the present invention will be described below with reference to the accompanying drawings. Please refer to... Figure 2 The present invention provides a computing task processing method applied to a STA (Terminal Device), which includes steps S10, S20 and S30.

[0045] S10. Receive a computing task allocation frame sent by the wireless access point, the computing task allocation frame carrying a field for indicating the splitting position of the computing task;

[0046] S20. The computing task is split based on the computing task allocation frame to obtain multiple sub-computing tasks;

[0047] S30. Calculate the results of the sub-computation tasks located before the first split position to obtain the first calculation result;

[0048] Specifically, in this embodiment of the application, after each STA receives the computing task allocation frame sent by the AP, it splits the local machine learning model according to the machine learning splitting position information indicated in the computing task allocation frame, and calculates the first part of the local machine learning model to obtain the first calculation result.

[0049] Specifically, before the STA receives the computing task allocation frame sent by the AP, steps S01 and S02 are also included:

[0050] Step S01. Receive a computing task query frame sent by the wireless access point. The computing task query frame is used to instruct the wireless access point to query the terminal device for a field indicating whether there is a computing task that needs assistance.

[0051] Specifically, the AP sends a computation task query frame to each STA to determine if any STA has a computation task requiring assistance. The computation task query frame is a type of trigger frame. For example... Figure 3 The diagram shows the structure of a computation task query frame. In this embodiment, the trigger type field in the common information field of the computation task query frame is used to achieve the above function.

[0052] Each computation task query frame consists of Frame Control, Duration, Access Address (RA), Transmit Address (TA), Common Info, User Information List, Padding, and Frame Check Sequence (FCS). The Common Info field contains at least one of the following bit fields: trigger type, uplink length, reserved bits, and trigger dependent common info. The trigger type field has a preset value. Table 1 illustrates the possible values ​​for the trigger type field in the trigger frame and their corresponding meanings. In Table 1, the preset value for the trigger type field in the computation task query frame is 8, indicating that the AP wants to know if the STA has any computation tasks that require assistance.

[0053] Table 1

[0054]

[0055] Step S02. When the target field in the computation task query frame is a target value, after a short frame interval, a computation task request frame is sent to the wireless access point. The computation task request frame carries fields indicating the type and number of computation tasks for which the wireless access point is requested to assist in completing the computation task.

[0056] Specifically, taking STA processing machine learning computation tasks as an example, when the STA receives a computation task query frame and the Trigger Type field in the Common Info field is set to 8, after waiting for the Short Interframe Space (SIFS), the STA transmits information such as the number of machine learning computation tasks to be processed and the type of machine learning model to the AP via a computation task request frame. As a response to the computation task query frame, the information in the computation task request frame can be carried in any MAC frame sent by the STA to the AP. For example... Figure 4 The diagram shows the structure of a computation task request frame. In this embodiment, the Control ID, Model Type, Scaling Factor, and QueueSize fields in the computation task request frame are used to achieve the above functions.

[0057] Each computation task request frame consists of fields such as Frame Control, Duration, Address, High Throughput Control (HT control), Frame Body, and FCS (Frame Checksum). The High Throughput Control field comprises a High Efficiency field and an Advanced Control (Acontrol) field. The Acontrol field includes a Control ID and Control Information field, which consists of the Model Type, Scaling Factor, and Task Size field.

[0058] The Control ID field indicates the type of the current MAC frame. For example, Table 2 shows the possible values ​​for the Control ID field of a MAC frame and the meaning of each value. In Table 2, the default value for the Control ID field of a computation task request frame is 7, indicating that this MAC frame type is a computation task request frame, carrying relevant information describing the STA machine learning computation task.

[0059] Table 2

[0060] Control ID field value MAC frame meaning 0 Triggered response scheduling …… 7 Compute Task Request Frame 8-14 Reserved fields 15 ONES frames

[0061] In the control information field, the model type field indicates the type of machine learning model that the STA needs to process. Specifically, a non-zero value for the ML_IM (Machine Learning_Image) field indicates an image processing model; a non-zero value for the ML_TXT (Machine Learning_Text) field indicates a file processing model; a non-zero value for the ML_VI (Machine Learning_Video) field indicates a video processing model; and a non-zero value for the ML_VO (Machine Learning_Voice) field indicates a speech processing model. Each model type is represented by 2 bits, therefore, besides the case of a zero value, each model type can be represented by three non-zero values, meaning there are three selectable machine learning model structures for each model type.

[0062] The scaling factor field and the task size field together represent the number of machine learning computation tasks that the STA needs to process. The values ​​for the scaling factor field are shown in Table 3. For example, when the scaling factor is 0 and the task size field is 10, it means that the STA currently has 16 * 10 = 160 machine learning computation tasks that need to be processed.

[0063] Table 3

[0064] scaling factor field value scaling factor 0 16 1 256 2 2048 3 32768

[0065] Specifically, after the STA receives the computing task allocation frame sent by the AP, the process also includes steps S03, S04, S05, S06, and S07:

[0066] Step S03. After a short interframe interval, a computing task allocation confirmation frame is sent to the wireless access point. The computing task allocation confirmation frame is used to indicate that the terminal device has confirmed receipt of the computing task allocation frame. The specific implementation of the computing task allocation confirmation frame can refer to the CTS (Clear to Send) frame in the prior art, and will not be described in detail here.

[0067] Step S04. Receive the calculation result transmission trigger frame sent by the wireless access point. The calculation result transmission trigger frame carries a field for instructing the STA on the radio resource units required for uplink transmission of the calculation result. The specific implementation of the calculation result transmission trigger frame can refer to the trigger frame in the prior art. The embodiments of this application will not be described in detail here.

[0068] Step S05. After the short interframe interval, transmit the result of the first calculation task to the wireless access point through the allocated uplink radio resource unit;

[0069] Specifically, after the STA receives the computation result transmission trigger frame sent by the AP, it does not need to provide feedback. After waiting for SIFS, it transmits the split model parameter data (i.e., the first computation task result) to the AP uplink according to the allocated Resource Unit (RU).

[0070] Step S06. Receive the calculation result confirmation frame sent by the wireless access point. The calculation result confirmation frame is used to indicate that the wireless access point confirms that it has received the first calculation result. The specific implementation of the calculation result confirmation frame can refer to the multi-STA block acknowledgment (Multi-STA BlockAck) frame in the prior art, and will not be described in detail here.

[0071] Step S07. Receive the second calculation result sent by the wireless access point, and obtain the target calculation result of the calculation task based on the first calculation result and the second calculation result. The second calculation result is the result calculated by the sub-computation task on the wireless access point.

[0072] Specifically, the AP continues to calculate the second part of the machine learning model deployed on the AP based on the received model parameter data, until the machine learning model calculation task is completed and the second calculation result is obtained. The AP transmits the second calculation result downlink to each STA through transmission interaction with the STA. Each STA can obtain the final inference result based on the calculation results of the two parts, thus efficiently completing the entire calculation task.

[0073] like Figure 5 The diagram illustrates the entire computational task processing method described above. First, when a STA has a machine learning task to process, the AP sends a computational task query frame to each STA to obtain information on whether each STA has any machine learning computational tasks that require assistance. After receiving the computational task query frame, the STA transmits a computational task request frame to the AP. This frame contains the type of machine learning task that each STA needs to assist in processing and the number of tasks to be processed.

[0074] After receiving a computation task request frame, the AP obtains information such as the machine learning type and number of tasks for each STA. With the optimization objective of maximizing the energy efficiency required for each STA to complete its machine learning task, the AP calculates the optimal splitting position and RU allocation for each STA's machine learning model. The AP then sequentially sends the calculated optimal model splitting position and RU allocation for each STA to each STA via a computation task allocation frame and a computation result transmission trigger frame.

[0075] When a STA receives a computation task allocation frame, it calculates a local machine learning model based on the machine learning model splitting location information in the computation task allocation frame. After waiting for SIFS, it sends a computation task allocation confirmation frame back to the AP to indicate that the STA has received the machine learning model splitting location information sent by the AP. After receiving the computation task allocation confirmation frame, the AP sends a computation result transmission trigger frame. This computation result transmission trigger frame contains Radio Resource Unit (RU) information required by each STA for uplink transmission of splitting layer machine learning parameters.

[0076] After receiving the calculation result transmission trigger frame, the STA transmits the split-layer calculated model parameter data to the AP according to the allocated RU uplink transmission. Upon receiving the model parameter data, the AP sends a calculation result confirmation frame to each STA, indicating that it has correctly received the uplink model parameter data transmitted by each STA. Simultaneously, the AP continues to calculate the next part of the machine learning model deployed on the AP based on the received model parameter data, until the machine learning model calculation task is completed and the machine learning calculation result is obtained. In subsequent transmission interactions with each STA, the AP will downlink transmit the calculated result to each STA, enabling each STA to combine the calculation results from both parts to obtain its own final calculation result.

[0077] from Figure 5 As can be seen, the transmission interval between frames is usually set to SIFS, which can satisfy the frame propagation delay and the processing delay of the node from receiving the transmission status.

[0078] Example 2:

[0079] Another computing task processing method provided by this embodiment of the invention is applied to an AP, i.e., a wireless access point. The method includes steps S40 and S50.

[0080] Step S40. Send a computing task allocation frame to the terminal device. The computing task allocation frame carries a field for indicating the split position of the computing task.

[0081] Step S50. Receive a computing task allocation confirmation frame sent by the terminal device. The computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

[0082] Specifically, after the AP sends the computation task allocation frame to each STA, it receives the computation task allocation confirmation frame from each STA after a certain period of time, indicating that each STA has split the learning model according to the split position information carried in the computation task allocation frame.

[0083] Specifically, before the AP sends the allocation frame to the STA, the process also includes steps S08, S09, S010, S011, and S012:

[0084] Step S08. Send a computing task query frame to the terminal device. The computing task query frame carries a field indicating whether the terminal device needs assistance with a computing task.

[0085] Step S09. When the target field in the computing task query frame is a target value, after a short frame interval, a computing task request frame sent by the terminal device is received. The computing task request frame carries information related to the type and number of computing tasks for which the wireless access point is requested to assist in completing the computing task.

[0086] Specifically, the AP sends a computation task query frame to each STA to obtain the number of machine learning tasks, the type of machine learning model, and the energy consumption of each floating-point operation of the STA. After a certain time interval, the AP receives a computation task request frame sent by the STA. The computation task request frame carries the number of machine learning tasks and the type of machine learning model required by the AP to determine the split position.

[0087] The computation task queries the frame structure as follows: Figure 3 As shown, obtaining the number of machine learning tasks and the type of machine learning model for a STA requires using the Reserved field in the User Info sequence of the computation task request frame. See Example 1 for specific usage; details will not be elaborated here. The structure of the computation task request frame is as follows: Figure 4 As shown, the fields used in this application are described in Example 1, and will not be repeated here.

[0088] Step S010. Based on the information in the computing task request frame, construct an optimization model, wherein the optimization model takes the energy efficiency performance of the computing task as the objective function;

[0089] Step S011. Solve the optimization model to obtain the computation task splitting position corresponding to the maximum objective function and the wireless resource unit allocation required for the transmission of the first computation result;

[0090] Specifically, after receiving the computation task request frame transmitted by the STA, the AP analyzes the computational load and the amount of intermediate layer model parameter data based on relevant information such as the machine learning model type and the number of machine learning tasks. With the optimization objective of maximizing the energy efficiency required for the STA to complete the machine learning task, a network optimization algorithm is used to determine the optimal model splitting position for each STA when splitting the learning process and the uplink radio resource units (RUs) required to transmit the corresponding model parameters.

[0091] Energy efficiency performance is defined as the ratio of the total number of bits of data for the machine learning task to the total energy consumption of each STA. It indicates the average amount of data that each STA can process per unit of energy consumption for the machine learning task. A higher value indicates that the system can process machine learning tasks with less energy consumption, i.e., it processes machine learning tasks more effectively. The formula for calculating energy efficiency performance is:

[0092]

[0093] Where, η EE For energy efficiency performance; D n F represents the number of machine learning tasks involved in the nth STA; n The size of the unit task data in bits for the machine learning task involved in the nth STA; Let n be the transmission power of the nth STA; r is the amount of model parameter data output by the split layer at the m-th split position; n E represents the transmission rate of the nth STA. FLOPs,n Calculate the energy consumption of a floating-point operation for the nth STA; The computational cost of the nth STA; N is the number of STAs; k m This represents the m-th splitting position.

[0094] The number of machine learning tasks and the type of machine learning model for each STA are obtained through computation task request frames sent by the STA to the AP. The size of the model parameter data and the computational load of the split layer are obtained by the AP through analysis of the machine learning model structure. The transmission rate between the STA and the AP is determined through negotiation between them. After obtaining these parameters, the AP executes a network optimization algorithm to select the optimal machine learning model splitting position for each STA and allocates the RU communication resource units used by the STA for uplink transmission of model parameter data, thereby maximizing energy efficiency.

[0095] Step S012. Construct a computing task allocation frame based on the computing task splitting location, and construct a computing result transmission trigger frame based on the wireless resource unit allocation;

[0096] Specifically, the optimized split positions are transmitted to each STA via a computation task allocation frame. The computation task allocation frame is a type of trigger frame. For example... Figure 7 The diagram shown is a structural schematic of a computation task allocation frame. In this embodiment, the above function needs to be implemented using the Common Info field of the computation task allocation frame.

[0097] The computation task allocation frame consists of Frame Control, Duration, Access Address (RA), Transmit Address (TA), Common Info, User Information List, Padding, and Frame Check Sequence (FCS). The Common Info field includes a trigger type field. The trigger type field has preset values, as shown in Table 1 of Example 1, and will not be repeated here. In Table 1, the preset value of the trigger type field for the computation task allocation frame is 9, indicating that this frame carries the splitting location information of the machine learning computation task model.

[0098] The user information sequence of the computation task allocation frame contains machine learning computation task model splitting information for multiple STAs, such as... Figure 6As shown, each STA can be numbered and named STA1, STA2, ..., STAn, indicating a total of n STAs. Each STA's information includes STA ID (STA identifier), layer allocation, forward error correction coding, modulation and coding scheme, dual-carrier modulation, spatial stream allocation, AP expected received power, and reserved fields. The layer allocation field indicates the selection of the splitting position. This field is 8 bits long and can represent 256 possibilities. For smaller machine learning models (e.g., less than 256 layers), the layer allocation field can directly indicate the specific layer splitting position. For example, a layer allocation value of 0 indicates the 0th splitting point, meaning the model does not split and the complete computation task is processed on the STA; a layer allocation value of 8 indicates the 8th layer of the model is the splitting position. For larger machine learning models (e.g., more than 256 layers), the layer allocation field can indicate the position of the typical splitting layer. For example, a value of 1 in the Layer Allocation field indicates that the model's first typical split layer is split. Here, the typical split layer is a commonly used model split layer position determined in advance based on the model structure analysis used in the specific machine learning task, and both AP and STA know the typical split positions of the machine learning model.

[0099] Specifically, after the AP sends the computing task allocation frame to the STA, the process also includes steps S013, S014, and S015:

[0100] Step S013. Send a calculation result transmission trigger frame to the terminal device, wherein the calculation result transmission trigger frame carries a field for instructing the terminal device on the radio resource units required for uplink transmission of the first calculation result;

[0101] Specifically, the trigger frame sent by the AP contains RU allocation information for each STA, which is used to instruct the STA to use the allocated RU to transmit data uplink and to notify the STA to start transmitting data uplink.

[0102] Step S014. After receiving the first calculation result transmitted by the terminal device through the allocated wireless resource unit, perform calculations on subsequent sub-computation tasks to obtain the second calculation result;

[0103] Step S015. Send the second calculation result to the terminal device;

[0104] Specifically, after receiving the machine learning split-layer model parameters transmitted uplink from each STA, the AP performs the remaining model calculations based on the corresponding machine learning model to obtain the calculation results. Simultaneously, the AP sends a calculation result acknowledgment frame, indicating that it has received the split-layer model parameters from each STA. In subsequent transmissions, the AP transmits the machine learning model calculation results to the corresponding STA.

[0105] Example 3:

[0106] like Figure 7 As shown, this embodiment provides a computing task processing device, the device including:

[0107] The first receiving unit 10 is used to receive a computing task allocation frame sent by a wireless access point, wherein the computing task allocation frame carries a field for indicating the splitting position of the computing task.

[0108] The splitting unit 20 is used to split the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks;

[0109] The first calculation unit 30 is used to calculate the calculation results of the sub-computation tasks located before the first split position, and obtain the first calculation result.

[0110] In one specific embodiment disclosed in this application, the apparatus further includes:

[0111] The second receiving unit is used to receive a computing task query frame sent by the wireless access point. The computing task query frame carries a field that instructs the wireless access point to query the terminal device whether there is a computing task that needs assistance.

[0112] The first sending unit is configured to send a computing task request frame to the wireless access point after a short frame interval when the target field in the computing task query frame is a target value. The computing task request frame carries fields indicating the type and number of computing tasks for which the wireless access point is requested to assist in completing the task.

[0113] In one specific embodiment disclosed in this application, the apparatus further includes:

[0114] The second sending unit is used to send a computing task allocation confirmation frame to the wireless access point after the short frame interval, wherein the computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame.

[0115] The third receiving unit is used to receive a calculation result transmission trigger frame sent by the wireless access point. The calculation result transmission trigger frame carries a field for indicating the wireless resource units required for uplink transmission of the first calculation result.

[0116] The third transmitting unit is used to transmit the first calculation result to the wireless access point via the allocated radio resource unit after the short frame interval.

[0117] The fourth receiving unit is used to receive a calculation result confirmation frame sent by the wireless access point. The calculation result confirmation frame is used to indicate that the wireless access point confirms that it has received the first calculation result and starts to execute subsequent sub-calculation tasks accordingly to obtain the second calculation result.

[0118] The second computing unit is used to receive the second computing result sent by the wireless access point, and to obtain the target computing result of the computing task based on the first computing result and the second computing result, wherein the second computing result is the result calculated by the sub-computing task on the wireless access point.

[0119] Example 4:

[0120] like Figure 8 As shown, this embodiment provides a computing task processing device, the device including:

[0121] The first sending unit 40 is used to send a computing task allocation frame to the terminal device, wherein the computing task allocation frame carries a field for indicating the split position of the computing task.

[0122] The first receiving unit 50 is configured to receive a computing task allocation confirmation frame sent by the terminal device. The computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

[0123] In one specific embodiment disclosed in this application, the apparatus further includes:

[0124] The second sending unit is used to send a computing task query frame to the terminal device. The computing task query frame carries a field indicating whether the terminal device has any computing tasks that need assistance.

[0125] The second receiving unit is configured to receive a computing task request frame sent by the terminal device after a short frame interval when the target field in the computing task query frame is a target value. The computing task request frame carries information related to the type and number of computing tasks for which the wireless access point is requested to assist in completing the computing task.

[0126] The first construction unit is used to construct an optimization model based on the information in the computing task request frame, wherein the optimization model takes the energy efficiency performance of the computing task as the objective function.

[0127] The solving unit is used to solve the optimization model to obtain the computation task splitting position corresponding to the maximum objective function and the wireless resource unit allocation required for the transmission of the first computation result.

[0128] The second construction unit is used to construct a computing task allocation frame based on the computing task splitting location, and to construct a computing result transmission trigger frame based on the wireless resource unit allocation.

[0129] In one specific embodiment disclosed in this application, the apparatus further includes:

[0130] The third sending unit is used to send a calculation result transmission trigger frame to the terminal device. The calculation result transmission trigger frame carries a field for instructing the terminal device on the radio resource units required for uplink transmission of the first calculation result.

[0131] The third receiving unit is used to receive the first calculation result transmitted by the terminal device through the allocated wireless resource unit, and then perform calculations on subsequent sub-computation tasks to obtain the second calculation result.

[0132] The fourth sending unit is used to send the second calculation result to the terminal device.

[0133] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0134] Example 5:

[0135] Corresponding to the above method embodiments, this embodiment also provides a computing task processing device. The computing task processing device described below and the computing task processing method described above can be referred to each other.

[0136] Figure 9 This is a block diagram illustrating a computing task processing device 800 according to an exemplary embodiment. For example... Figure 9 As shown, the computing task processing device 800 may include a processor 801 and a memory 802. The computing task processing device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0137] The processor 801 controls the overall operation of the computing task processing device 800 to complete all or part of the steps in the aforementioned computing task processing method. The memory 802 stores various types of data to support the operation of the computing task processing device 800. This data may include, for example, instructions for any application or method operating on the computing task processing device 800, and application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 802 or transmitted via communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 805 is used for wired or wireless communication between the computing task processing device 800 and other devices. Wireless communication may include Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of these. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0138] In an exemplary embodiment, the computing task processing device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the computing task processing method described above.

[0139] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the above-described computing task processing method. For example, the computer-readable storage medium may be the memory 802 including program instructions, which may be executed by the processor 801 of the computing task processing device 800 to complete the above-described computing task processing method.

[0140] Example 6:

[0141] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the computing task processing method described above.

[0142] A readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the computational task processing method described in the above method embodiments.

[0143] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0144] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0145] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for processing computational tasks, characterized in that, Applied to a terminal device, the method includes: Receive a computing task allocation frame sent by a wireless access point, wherein the computing task allocation frame carries a field for indicating the splitting position of the computing task; The computing task is split into multiple sub-computing tasks based on the computing task allocation frame. The calculation results of the sub-computation tasks located before the first split position are calculated to obtain the first calculation result.

2. The computational task processing method according to claim 1, characterized in that... Before receiving the computing task allocation frame sent by the wireless access point, the process includes: The device receives a computing task query frame sent by the wireless access point, the computing task query frame carrying a field for instructing the wireless access point to query the terminal device whether there is a computing task that needs assistance. When the target field in the computation task query frame is set to a target value, a computation task request frame is sent to the wireless access point after a short frame interval. The computation task request frame carries fields indicating the type and number of computation tasks for which the wireless access point is requested to assist in completing the computation task. The target field is a trigger type field, and the target value is 8.

3. The computational task processing method according to claim 2, characterized in that... After splitting the computation task based on the computation task allocation frame, calculating the sub-computation tasks located before the first split position, and obtaining the first calculation result, the process includes: After the short inter-frame interval, a computing task allocation confirmation frame is sent to the wireless access point. The computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame. The system receives a calculation result transmission trigger frame sent by the wireless access point, the calculation result transmission trigger frame carrying a field for indicating the radio resource units required for uplink transmission of the first calculation result; After the short inter-frame interval, the first calculation result is transmitted uplink to the wireless access point through the allocated radio resource unit; The wireless access point receives a calculation result confirmation frame, which indicates that the wireless access point confirms that it has received the first calculation result and starts to execute subsequent sub-computation tasks accordingly to obtain the second calculation result. The system receives a second calculation result sent by the wireless access point, and obtains the target calculation result of the calculation task based on the first calculation result and the second calculation result. The second calculation result is the result calculated by the sub-computation task on the wireless access point.

4. A method for processing computational tasks, characterized in that, Applied to wireless access points, the method includes: A computing task allocation frame is sent to the terminal device, the computing task allocation frame carrying a field for indicating the split position of the computing task; The terminal device receives a computing task allocation confirmation frame, which indicates that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

5. The computational task processing method according to claim 4, characterized in that... Before sending the computing task allocation frame to the terminal device, the process includes: A computing task query frame is sent to the terminal device, the computing task query frame carrying a field indicating whether the terminal device has any computing tasks that need assistance. When the target field in the computation task query frame is a target value, after a short frame interval, the terminal device sends a computation task request frame. The computation task request frame carries information related to the type and number of computation tasks for which the wireless access point is requested to assist in completing the computation task. The target field is a trigger type field, and the target value is 8. Based on the information in the computing task request frame, an optimization model is constructed, with the energy efficiency performance of the computing task as the objective function. The optimization model is solved to obtain the computation task splitting position corresponding to the maximum objective function and the wireless resource unit allocation required for the transmission of the first computation result. A computing task allocation frame is constructed based on the computing task splitting location, and a computing result transmission trigger frame is constructed based on the wireless resource unit allocation.

6. The computational task processing method according to claim 4, characterized in that... Upon receiving a computing task allocation confirmation frame from the terminal device, the computing task allocation confirmation frame indicates that the terminal device has confirmed receipt of the computing task allocation frame, including: A calculation result transmission trigger frame is sent to the terminal device, the calculation result transmission trigger frame carrying a field for instructing the terminal device on the radio resource units required for uplink transmission of the first calculation result; After receiving the first calculation result transmitted by the terminal device through the allocated wireless resource unit, the system performs calculations on subsequent sub-computation tasks to obtain the second calculation result. The second calculation result is sent to the terminal device.

7. A computing task processing device, characterized in that, include: The first receiving unit is used to receive a computing task allocation frame sent by a wireless access point, wherein the computing task allocation frame carries a field for indicating the splitting position of the computing task. The splitting unit is used to split the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks; The first computing unit is used to calculate the computing results of the sub-computation tasks located before the first split position, and obtain the first computing result.

8. A computing task processing device, characterized in that, include: The first sending unit is used to send a computing task allocation frame to the terminal device, wherein the computing task allocation frame carries a field for indicating the split position of the computing task; The first receiving unit is configured to receive a computing task allocation confirmation frame sent by the terminal device. The computing task allocation confirmation frame is used to indicate that the terminal device confirms that it has received the computing task allocation frame. The terminal device splits the computing task based on the computing task allocation frame to obtain multiple sub-computing tasks. The terminal device calculates the calculation results of the sub-computing tasks located before the first split position to obtain a first calculation result.

9. A computing task processing device, characterized in that, include: memory for storing computer programs; A processor, configured to implement the steps of the computational task processing method as described in any one of claims 1 to 3 or 4 to 6 when executing the computer program.

10. A readable storage medium, characterized in that, The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the computational task processing method as described in any one of claims 1 to 3 or 4 to 6.

Images