Task allocation method and device, equipment and storage medium
By generating and training the task allocation model, the problem of inaccurate allocation of pending items in the service industry is solved, and more efficient and accurate task allocation is achieved, improving user experience.
Patent Information
- Application Number
- CN202411551033.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-31
- Publication Date
- 2025-03-14
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the service industry, especially in the fields of medical care, insurance, e-commerce and professional consulting, it is difficult for the existing technology to accurately and promptly allocate pending matters to matched departments for processing.
By obtaining the sample data set, including the pending items and identified task allocation results, a preset MOE network model and LoRA adapter generate a task allocation model, and by adjusting the model parameters until convergence, the target task allocation model is obtained for task allocation.
Improve the efficiency and accuracy of task processing, ensuring that pending matters can be accurately allocated to the appropriate department, thereby improving the user experience.
Smart Images
Figure CN119623922A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to a task allocation method, device, equipment and storage medium. Background Art
[0002] With the rapid development of big models, AI interactive robots based on big models are gradually replacing human interaction with users in the service industry, especially in the fields of medical care, insurance, e-commerce, and professional consulting. Big models for question and answer have been widely used. For example, the application of big models in business scenarios in the field of insurance claims includes multiple processing links, such as reporting, information recording, investigation and loss assessment, document review, settlement review, approval, and payment. Each processing link will be handled by different departments, but it is difficult to accurately and timely assign each task to the matching department for processing in the claims event.
[0003] Therefore, how to allocate pending matters to appropriate departments for processing is an urgent problem to be solved. Summary of the invention
[0004] The main purpose of this application is to provide a task allocation method, device, equipment and storage medium, aiming to improve the efficiency and accuracy of task allocation of pending matters.
[0005] In a first aspect, the present application provides a task allocation method, the task allocation method comprising the following steps:
[0006] Acquire a sample data set, the sample data set comprising a plurality of sample data, the sample data comprising pending items and identified task allocation results;
[0007] Obtain a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model;
[0008] Selecting a sample data from the sample data set as target sample data, and inputting the target to-be-processed items in the target sample data into the preset task allocation model for task allocation, to obtain a predicted task allocation result;
[0009] Determining whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result;
[0010] If the preset task allocation model has not converged, adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters, and return to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model;
[0011] The target pending items are obtained, and the target pending items are input into the target task allocation model to obtain a task allocation result.
[0012] In a second aspect, the present application further provides a task allocation device, the task allocation device comprising an acquisition module, a selection module, a generation module, a determination module and an adjustment module, wherein:
[0013] The acquisition module is used to acquire a sample data set, wherein the sample data set includes a plurality of sample data, and the sample data includes pending items and identified task allocation results;
[0014] The acquisition module is further used to acquire a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model;
[0015] The selection module is used to select a sample data from the sample data set as the target sample data;
[0016] The generating module is used to input the target pending items in the target sample data into the preset task allocation model for task allocation, and obtain a predicted task allocation result;
[0017] The determination module is used to determine whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result;
[0018] The adjustment module is used to adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters if the preset task allocation model has not converged;
[0019] The generating module is used to return to the step of looping and selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model;
[0020] The acquisition module is also used to acquire target pending items;
[0021] The generation module is used to input the target pending items into the target task allocation model to obtain a task allocation result.
[0022] In a third aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the task allocation method as described above are implemented.
[0023] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the task allocation method as described above are implemented.
[0024] The present application provides a task allocation method, device, equipment and storage medium. The present application obtains a sample data set, which includes multiple sample data, and the sample data includes pending items and identified task allocation results; obtains a preset MOE network model and multiple LoRA adapters, and inserts the multiple LoRA adapters into a gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model; selects a sample data from the sample data set as a target sample data, and inputs the target pending items in the target sample data into the preset task allocation model for task allocation to obtain a predicted task allocation result; determines whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result; if the preset task allocation model does not converge, adjusts the model parameters of at least one of the gated network, multiple expert networks and multiple LoRA adapters, and returns to the step of looping and executing the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain a target task allocation model; obtains the target pending items, and inputs the target pending items into the target task allocation model to obtain the task allocation result. According to the preset MOE network model and multiple LoRA adapters, the present invention can accurately construct a preset task allocation model, and then train the preset task allocation model based on a sample data set until convergence, so as to accurately obtain a target task allocation model, and then perform task allocation for the target pending items based on the target task allocation model, so as to accurately obtain the task allocation result, thereby greatly improving the efficiency and accuracy of task processing, so as to improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0026] Figure 1 A flowchart of a task allocation method provided in an embodiment of the present application;
[0027] Figure 2 A structural diagram of a task allocation model provided in an embodiment of the present application;
[0028] Figure 3 A schematic diagram of the structure of another task allocation model provided in an embodiment of the present application;
[0029] Figure 4 A schematic block diagram of a task allocation device provided in an embodiment of the present application;
[0030] Figure 5 A schematic block diagram of the structure of a computer device provided in an embodiment of the present application.
[0031] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0032] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0033] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.
[0034] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0035] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0036] With the rapid development of big models, AI interactive robots based on big models are gradually replacing human interaction with users in the service industry, especially in the fields of medical care, insurance, e-commerce, and professional consulting. Big models for question and answer have been widely used. For example, the application of big models in business scenarios in the field of insurance claims includes multiple processing links, such as reporting, information recording, investigation and loss assessment, document review, settlement review, approval, and payment. Each processing link will be handled by different departments, but it is difficult to accurately and timely assign each task to the matching department for processing in the claims event.
[0037] To solve the above problems, embodiments of the present application provide a task allocation method, apparatus, device and storage medium. The task allocation method includes: obtaining a sample data set, wherein the sample data set includes multiple sample data, and the sample data includes pending items and identified task allocation results; obtaining a preset MOE network model and multiple LoRA adapters, and inserting the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model; selecting a sample data from the sample data set as the target sample data, and inputting the target pending items in the target sample data into the preset task allocation model for task allocation to obtain a predicted task allocation result; determining whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result; if the preset task allocation model does not converge, adjusting the model parameters of at least one of the gated network, the multiple expert networks and the multiple LoRA adapters, and returning to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain a target task allocation model; obtaining a target pending item, and inputting the target pending item into the target task allocation model to obtain a task allocation result.
[0038] The task allocation method can be applied to computer devices, which can be electronic devices such as mobile phones, tablet computers, laptop computers, desktop computers, personal digital assistants and wearable devices.
[0039] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.
[0040] Please refer to Figure 1 , Figure 1 A flowchart of a task allocation method provided in an embodiment of the present application.
[0041] like Figure 1As shown, the task allocation method includes steps S101 to S106.
[0042] Step S101: Acquire a sample data set, where the sample data set includes a plurality of sample data, and the sample data includes pending items and identified task allocation results.
[0043] The sample data set includes multiple sample data, each of which includes pending items and identified task allocation results. The pending items are issues to be processed by users, for example, pending items may be insurance claims, medical consultations, and e-commerce product sales, etc. The task allocation results are the links and processing nodes that should be processed at the current time point or the selected time point of the pending items.
[0044] In some embodiments, the pending items and the task assignment results corresponding to the pending items are obtained, and the pending items and the task assignment results corresponding to the pending items are used as a set of sample data, and the steps of obtaining the pending items and the task assignment results corresponding to the pending items and the task assignment results corresponding to the pending items are performed repeatedly to obtain a sample data set. By obtaining multiple sample data, the sample data set can be accurately obtained.
[0045] Step S102: Obtain a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model.
[0046] The preset task allocation model includes a plurality of preset MOE network models, and the preset MOE network model includes a gated network and a plurality of expert networks.
[0047] It should be noted that the MOE network model can be selected according to actual conditions, and the embodiment of the present invention does not specifically limit this. For example, the MOE network model can be a Qwen1.5-MoE-A2.7B-Chat-GPTQ-Int4 model.
[0048] In some embodiments, the expert network includes an encoder, an attention layer, and a decoder; the task assignment result corresponding to the pending item and the pending item is obtained, and the task assignment result corresponding to the pending item and the pending item is used as a set of sample data, and the steps of obtaining the task assignment result corresponding to the pending item and the pending item and the pending item are used as a set of sample data are executed cyclically to obtain a sample data set. A sample data is selected from the sample data set as the target sample data, and the target pending item is input into the encoder of the preset expert network for feature extraction to obtain a fourth feature vector; the fourth feature vector is input into the attention layer for feature processing to obtain a fifth feature vector, and the fifth feature vector is input into the decoder to obtain a predicted task assignment result; according to the predicted task assignment result and the identified task assignment result in the target sample data, it is determined whether the preset expert network converges, if the preset expert network does not converge, the model parameters of the preset expert network are adjusted, and the step of selecting a sample data from the sample data set as the target sample data is returned to be executed until the preset expert network converges to obtain the expert network. Through the above steps, an expert network can be accurately trained, and the above steps can be repeatedly performed to obtain multiple expert networks.
[0049] In some embodiments, based on the predicted task assignment results and the identified task assignment results in the target sample data, the method for determining whether the preset expert network has converged can be: based on the predicted task assignment results and the identified task assignment results, determine the loss value of the preset expert network, if the loss value is less than or equal to the preset loss value, determine that the preset expert network has converged; if the loss value of the preset expert network is greater than the preset loss value, determine that the preset expert network has not converged. Among them, the preset loss value can be set according to actual conditions, and the implementation examples of the present invention do not make specific limitations on this. By calculating the loss value of the preset expert network, it is possible to accurately know whether the preset expert network has converged, which greatly improves the efficiency and accuracy of the preset expert network training.
[0050] In some embodiments, the gated network and each expert network include an attention layer; each LoRA adapter is inserted into the attention layer of the gated network and each expert network to generate a preset task allocation model. By inserting the LoRA adapter in each attention layer, the efficiency and accuracy of the task allocation model for task allocation can be effectively improved.
[0051] For example, Figure 2As shown, the preset task allocation model 10 includes a gated network 11, an expert network a21, an expert network b22 and an expert network c23. LoRA adapters are inserted into the gated network 11, the expert network a21, the expert network b22 and the expert network c23, respectively, to obtain the following: Figure 3 The preset task allocation model 10 is shown.
[0052] In some embodiments, each LoRA adapter is inserted into the attention layer of the gated network and / or part of the expert network to generate a preset task allocation model. By adding the LoRA adapter to the partial attention layer, the efficiency and accuracy of the task allocation model for task allocation can be improved, and the training time of the task allocation model can be reduced.
[0053] For example, Figure 3 As shown, the preset task allocation model 10 includes a gated network 11, an expert network a21, an expert network b22 and an expert network c23. LoRA adapters are inserted into the gated network 11, the expert network a21 and the expert network c23 respectively, and the following is obtained: Figure 3 The preset task allocation model 10 is shown.
[0054] Step S103: select a sample data from the sample data set as target sample data, and input the target to-be-processed items in the target sample data into the preset task allocation model for task allocation to obtain a predicted task allocation result.
[0055] In some embodiments, a sample data is selected from the sample data set as the target sample data, wherein the selection method can be selected according to the actual situation, and the embodiments of the present invention do not specifically limit this. For example, the selection method can be random selection, or selection can be performed according to the time sequence in which the sample data is established.
[0056] In some embodiments, the target pending items are input into the gating network in the MOE network model for expert network selection to obtain the target expert network; the target pending items are input into the encoder of the target expert network for feature extraction to obtain the first feature vector; the first feature vector is input into the attention layer for feature processing to obtain the second feature vector, and the second feature vector is input into the LoRA adapter connected to the attention layer for feature adjustment to obtain the third feature vector; the third feature vector is input into the decoder to obtain the predicted task allocation result. The gating network can accurately allocate the target pending items to the appropriate target expert network, and the target expert network processes the target pending items to accurately obtain the task allocation result.
[0057] Step S104: determining whether the preset task allocation model has converged according to the predicted task allocation result and the identified task allocation result.
[0058] In some embodiments, the loss value of the preset task allocation model is determined based on the predicted task allocation result and the identified task allocation result; if the loss value is less than or equal to the preset loss value, it is determined that the preset task allocation model has converged; if the loss value is greater than the preset loss value, it is determined that the preset task allocation model has not converged. Among them, the preset loss value can be set according to actual conditions, and the embodiment of the present invention does not make specific limitations on this. For example, the preset loss value can be set to 0.002. According to the loss value calculated based on the predicted task allocation result and the identified task allocation result, it is possible to accurately know whether the preset task allocation model has converged, which greatly improves the efficiency and accuracy of training the preset task allocation model.
[0059] In some embodiments, the method for determining the loss value of a preset task allocation model based on the predicted task allocation results and the identified task allocation results can be: performing a cross entropy calculation based on the predicted task allocation results and the identified task allocation results to obtain a first cross entropy result, and using the first cross entropy result as the loss value of the preset task allocation model.
[0060] It should be noted that the method for determining the loss value of the preset task allocation model based on the predicted task allocation results and the identified task allocation results can be: calculating the cosine similarity between the predicted task allocation results and the identified task allocation results, and subtracting the value of the cosine similarity from the unit 1 to obtain a numerical value as the loss value.
[0061] In some embodiments, the method of determining the loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result can be: calculating the cross entropy of the predicted task allocation result and the identified task allocation result to obtain a first loss value; obtaining a second loss value corresponding to the gated network, and superimposing the first loss value and the second loss value to obtain the loss value of the preset task allocation model. The second loss value can be a pre-set loss value, and the loss value of the model is determined based on the first loss value and the second loss value, which can improve the accuracy of the loss value calculation of the task allocation result.
[0062] Step S105: If the preset task allocation model has not converged, adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters, and return to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model.
[0063] If the loss value of the preset task allocation model is greater than the preset loss value, it is determined that the preset task allocation model has not converged; the model parameters of at least one of the gated network, the multiple expert networks, and the multiple LoRA adapters are adjusted, and the step of selecting a sample data from the sample data set as the target sample data is returned to the loop execution until the preset task allocation model converges to obtain the target task allocation model. By adjusting the model parameters in the preset task allocation model and training it cyclically, a converged target task allocation model can be accurately obtained.
[0064] In some embodiments, the method of adjusting the model parameters of at least one of the gated network, multiple expert networks, and multiple LoRA adapters can be: obtaining a preset first model adjustment strategy, the preset first model adjustment strategy is used to adjust the parameters of the LoRA adapter corresponding to each expert network; adjusting the parameters of the LoRA adapter corresponding to at least one expert network according to the preset first model adjustment strategy. The parameters of the LoRA adapter corresponding to the expert network can be effectively adjusted according to the preset first model adjustment strategy.
[0065] It should be noted that the preset first model adjustment strategy can be set according to actual conditions, and the embodiment of the present invention does not specifically limit this. For example, the parameters of the LoRA adapters corresponding to all expert networks are adjusted. Or, the parameters of the LoRA adapters corresponding to the expert networks are randomly adjusted.
[0066] In some embodiments, the method of adjusting the model parameters of at least one of the gated network, multiple expert networks, and multiple LoRA adapters can be: obtaining a preset second model adjustment strategy, the preset second model adjustment strategy is used to adjust the parameters of the gated network, multiple expert networks, and each LoRA adapter; adjusting the parameters of the gated network, multiple expert networks, and multiple LoRA adapters according to the preset second model adjustment strategy. By adjusting the parameters of the gated network, each expert network, and each LoRA adapter through the preset second model adjustment strategy, the model parameters of the task allocation model can be comprehensively adjusted.
[0067] It should be noted that the preset second model adjustment strategy can be set according to actual conditions, and the embodiment of the present invention does not specifically limit this. For example, the preset second model adjustment strategy can be to adjust the parameters of the expert network; or the preset second model adjustment strategy can be to adjust the parameters of the gated network and each expert network; or the preset second model adjustment strategy can be to adjust the parameters of the gated network, each expert network and each LoRA adapter.
[0068] In some embodiments, if the preset task allocation model has not converged, the model parameters of the gated network, the multiple expert networks and at least one of the multiple LoRA adapters are adjusted, and the loop execution is returned to select a sample data from the sample data set as the target sample data, and the target pending items in the target sample data are input into the preset task allocation model for task allocation to obtain a predicted task allocation result; based on the predicted task allocation result and the identified task allocation result, determine whether the preset task allocation model has converged, until the preset task allocation model converges to obtain the target task allocation model.
[0069] Step S106: Obtain target pending items, and input the target pending items into the target task allocation model to obtain a task allocation result.
[0070] Obtain the target pending items, input the target pending items into the target task allocation model, and obtain the task allocation result. By identifying the target pending items through the target task allocation model, the task allocation result can be accurately obtained, thereby greatly improving the efficiency and accuracy of task processing, so as to improve the user experience.
[0071] Exemplarily, the target pending matter is an insurance claim, and the insurance claim is input into the target task allocation model for identification, and the pending task at the current moment is obtained as an investigation and loss assessment, and the insurance claim is allocated to the investigation and loss assessment department to complete the task allocation.
[0072] Exemplarily, the target pending matter is a medical consultation event. The medical consultation event is input into the target task allocation model for identification, and the current pending task is obtained as an expert doctor's compound. The medical consultation event is assigned to the corresponding expert doctor to complete the task allocation, which greatly improves the efficiency of medical consultation.
[0073] The task allocation method provided in the above embodiment obtains a sample data set, which includes multiple sample data, and the sample data includes pending items and identified task allocation results; obtains a preset MOE network model and multiple LoRA adapters, and inserts the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model; selects a sample data from the sample data set as the target sample data, and inputs the target pending items in the target sample data into the preset task allocation model for task allocation to obtain a predicted task allocation result; determines whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result; if the preset task allocation model does not converge, adjusts the model parameters of at least one of the gated network, multiple expert networks and multiple LoRA adapters, and returns to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model; obtains the target pending items, and inputs the target pending items into the target task allocation model to obtain the task allocation result. According to the preset MOE network model and multiple LoRA adapters, the present invention can accurately construct a preset task allocation model, and then train the preset task allocation model based on a sample data set until convergence, so as to accurately obtain a target task allocation model, and then perform task allocation for the target pending items based on the target task allocation model, so as to accurately obtain the task allocation result, thereby greatly improving the efficiency and accuracy of task processing, so as to improve the user experience.
[0074] See also Figure 4 , Figure 4 A schematic block diagram of a task allocation device provided in an embodiment of the present application.
[0075] like Figure 4 As shown, the task allocation device 200 includes an acquisition module 210, a selection module 220, a generation module 230, a determination module 240 and an adjustment module 250, wherein:
[0076] The acquisition module 210 is used to acquire a sample data set, wherein the sample data set includes a plurality of sample data, and the sample data includes pending items and identified task allocation results;
[0077] The acquisition module 210 is also used to acquire a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model;
[0078] The selection module 220 is used to select a sample data from the sample data set as the target sample data;
[0079] The generating module 230 is used to input the target pending items in the target sample data into the preset task allocation model for task allocation to obtain a predicted task allocation result;
[0080] The determination module 240 is used to determine whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result;
[0081] The adjustment module 250 is used to adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters if the preset task allocation model has not converged;
[0082] The generating module 230 is used to return to the step of looping and selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model;
[0083] The acquisition module 210 is also used to acquire target pending items;
[0084] The generating module 230 is further used to input the target pending items into the target task allocation model to obtain a task allocation result.
[0085] In some embodiments, the generating module 230 is further configured to:
[0086] Each of the LoRA adapters is respectively inserted into the gating network and the attention layer of each of the expert networks to generate a preset task allocation model.
[0087] In some embodiments, the determining module 240 is further configured to:
[0088] Determining a loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result;
[0089] If the loss value is less than or equal to a preset loss value, it is determined that the preset task allocation model has converged;
[0090] If the loss value is greater than a preset loss value, it is determined that the preset task allocation model has not converged.
[0091] In some embodiments, the determining module 240 is further configured to:
[0092] Calculating the cross entropy of the predicted task allocation result and the identified task allocation result to obtain a first loss value;
[0093] A second loss value corresponding to the gating network is obtained, and the first loss value and the second loss value are superimposed to obtain a loss value of the preset task allocation model.
[0094] In some embodiments, the adjustment module 250 is further configured to:
[0095] Obtaining a preset first model adjustment strategy, where the preset first model adjustment strategy is used to adjust the parameters of the LoRA adapter corresponding to each of the expert networks;
[0096] The parameters of the LoRA adapter corresponding to at least one of the expert networks are adjusted according to the preset first model strategy.
[0097] In some embodiments, the adjustment module 250 is further configured to:
[0098] Obtaining a preset second model adjustment strategy, wherein the preset second model adjustment strategy is used to adjust parameters of the gated network, the multiple expert networks, and each of the LoRA adapters;
[0099] Parameters of the gated network, the multiple expert networks, and the multiple LoRA adapters are adjusted according to the preset second model adjustment strategy.
[0100] In some embodiments, the generating module 230 is further configured to:
[0101] Inputting the target pending items into the gated network in the MOE network model to select an expert network, thereby obtaining a target expert network;
[0102] Inputting the target pending items into the encoder of the target expert network for feature extraction to obtain a first feature vector;
[0103] Inputting the first feature vector into the attention layer for feature processing to obtain a second feature vector, and inputting the second feature vector into a LoRA adapter connected to the attention layer for feature adjustment to obtain a third feature vector;
[0104] The third feature vector is input into the decoder to obtain the predicted task allocation result.
[0105] It should be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-mentioned task allocation device can refer to the corresponding process in the aforementioned task allocation method embodiment, and will not be repeated here.
[0106] See also Figure 5 , Figure 5A schematic block diagram of the structure of a computer device provided in an embodiment of the present application.
[0107] like Figure 5 As shown, the computer device 300 includes a processor 302 and a memory 303 connected via a system bus 301 , wherein the memory 303 may include a storage medium and an internal memory.
[0108] The storage medium can store a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any task allocation method.
[0109] The processor 302 is used to provide computing and control capabilities and support the operation of the entire computer device.
[0110] The internal memory provides an environment for the operation of the computer program in the storage medium. When the computer program is executed by the processor, the processor can execute any task allocation method.
[0111] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0112] It should be understood that the processor 302 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0113] In one embodiment, the processor 302 is used to run a computer program stored in the memory to implement the following steps:
[0114] Acquire a sample data set, the sample data set comprising a plurality of sample data, the sample data comprising pending items and identified task allocation results;
[0115] Obtain a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model;
[0116] Selecting a sample data from the sample data set as target sample data, and inputting the target to-be-processed items in the target sample data into the preset task allocation model for task allocation, to obtain a predicted task allocation result;
[0117] Determining whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result;
[0118] If the preset task allocation model has not converged, adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters, and return to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model;
[0119] The target pending items are obtained, and the target pending items are input into the target task allocation model to obtain a task allocation result.
[0120] In one embodiment, the gated network and each of the expert networks include an attention layer, and the processor 302 is used to implement the insertion of the multiple LoRA adapters into the gated network and the multiple expert networks in the preset MOE network model, respectively, to generate a preset task allocation model, and to implement:
[0121] Each of the LoRA adapters is respectively inserted into the gating network and the attention layer of each of the expert networks to generate a preset task allocation model.
[0122] In one embodiment, when the processor 302 determines whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result, it is configured to implement:
[0123] Determining a loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result;
[0124] If the loss value is less than or equal to a preset loss value, it is determined that the preset task allocation model has converged;
[0125] If the loss value is greater than a preset loss value, it is determined that the preset task allocation model has not converged.
[0126] In one embodiment, when the processor 302 determines the loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result, it is configured to implement:
[0127] Calculating the cross entropy of the predicted task allocation result and the identified task allocation result to obtain a first loss value;
[0128] A second loss value corresponding to the gating network is obtained, and the first loss value and the second loss value are superimposed to obtain a loss value of the preset task allocation model.
[0129] In one embodiment, the processor 302, when implementing the adjustment of the model parameters of at least one of the gating network, the plurality of expert networks, and the plurality of LoRA adapters, is used to implement:
[0130] Obtaining a preset first model adjustment strategy, where the preset first model adjustment strategy is used to adjust the parameters of the LoRA adapter corresponding to each of the expert networks;
[0131] The parameters of the LoRA adapter corresponding to at least one of the expert networks are adjusted according to the preset first model strategy.
[0132] In one embodiment, the processor 302, when implementing the adjustment of the model parameters of at least one of the gating network, the multiple expert networks, and the multiple LoRA adapters, is used to implement:
[0133] Obtaining a preset second model adjustment strategy, wherein the preset second model adjustment strategy is used to adjust parameters of the gated network, the multiple expert networks, and each of the LoRA adapters;
[0134] Parameters of the gated network, the multiple expert networks, and the multiple LoRA adapters are adjusted according to the preset second model adjustment strategy.
[0135] In one embodiment, the expert network includes an encoder, an attention layer, and a decoder. When the processor 302 inputs the target items to be processed in the target sample data into the preset task allocation model for task allocation and obtains the predicted task allocation result, it is used to implement:
[0136] Inputting the target pending items into the gated network in the MOE network model to select an expert network, thereby obtaining a target expert network;
[0137] Inputting the target pending items into the encoder of the target expert network for feature extraction to obtain a first feature vector;
[0138] Inputting the first feature vector into the attention layer for feature processing to obtain a second feature vector, and inputting the second feature vector into a LoRA adapter connected to the attention layer for feature adjustment to obtain a third feature vector;
[0139] The third feature vector is input into the decoder to obtain the predicted task allocation result.
[0140] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the computer device described above can refer to the corresponding process in the aforementioned task allocation method embodiment, which will not be repeated here.
[0141] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the task allocation method of the present application.
[0142] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may be non-volatile or volatile. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.
[0143] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.
[0144] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.
[0145] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0146] It should also be understood that the term "and / or" used in the present specification refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.
[0147] The serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application.
Claims
1. A task allocation method, characterized in that: include: Acquire a sample data set, the sample data set comprising a plurality of sample data, the sample data comprising pending items and identified task allocation results; Obtain a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model; Selecting a sample data from the sample data set as target sample data, and inputting the target to-be-processed items in the target sample data into the preset task allocation model for task allocation, to obtain a predicted task allocation result; Determining whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result; If the preset task allocation model has not converged, adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters, and return to the loop to execute the step of selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model; The target pending items are obtained, and the target pending items are input into the target task allocation model to obtain a task allocation result.
2. The task allocation method according to claim 1, characterized in that: The gated network and each of the expert networks include an attention layer, and the plurality of LoRA adapters are respectively inserted into the gated network and the plurality of expert networks in the preset MOE network model to generate a preset task allocation model, including: Each of the LoRA adapters is respectively inserted into the gating network and the attention layer of each of the expert networks to generate a preset task allocation model.
3. The task allocation method according to claim 1, characterized in that: The determining, according to the predicted task allocation result and the identified task allocation result, whether the preset task allocation model converges includes: Determining a loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result; If the loss value is less than or equal to a preset loss value, it is determined that the preset task allocation model has converged; If the loss value is greater than a preset loss value, it is determined that the preset task allocation model has not converged.
4. The task allocation method according to claim 3, characterized in that: The determining the loss value of the preset task allocation model according to the predicted task allocation result and the identified task allocation result includes: Calculating the cross entropy of the predicted task allocation result and the identified task allocation result to obtain a first loss value; A second loss value corresponding to the gating network is obtained, and the first loss value and the second loss value are superimposed to obtain a loss value of the preset task allocation model.
5. The task allocation method according to claim 1, characterized in that: Adjusting a model parameter of at least one of the gating network, the plurality of expert networks, and the plurality of LoRA adapters comprises: Obtaining a preset first model adjustment strategy, wherein the preset first model adjustment strategy is used to adjust the parameters of the LoRA adapter corresponding to each of the expert networks; The parameters of the LoRA adapter corresponding to at least one of the expert networks are adjusted according to the preset first model strategy.
6. The task allocation method according to claim 1, characterized in that: The adjusting the model parameters of at least one of the gating network, the plurality of expert networks and the plurality of LoRA adapters comprises: Obtaining a preset second model adjustment strategy, wherein the preset second model adjustment strategy is used to adjust parameters of the gated network, the multiple expert networks, and each of the LoRA adapters; Parameters of the gated network, the multiple expert networks, and the multiple LoRA adapters are adjusted according to the preset second model adjustment strategy.
7. The task allocation method according to any one of claims 1 to 6, characterized in that: The expert network includes an encoder, an attention layer and a decoder; the target items to be processed in the target sample data are input into the preset task allocation model for task allocation to obtain a predicted task allocation result, including: Inputting the target pending items into the gated network in the MOE network model to select an expert network, thereby obtaining a target expert network; Inputting the target pending items into the encoder of the target expert network for feature extraction to obtain a first feature vector; Inputting the first feature vector into the attention layer for feature processing to obtain a second feature vector, and inputting the second feature vector into a LoRA adapter connected to the attention layer for feature adjustment to obtain a third feature vector; The third feature vector is input into the decoder to obtain the predicted task allocation result.
8. A task allocation device, characterized in that: The task allocation device includes an acquisition module, a selection module, a generation module, a determination module and an adjustment module, wherein: The acquisition module is used to acquire a sample data set, wherein the sample data set includes a plurality of sample data, and the sample data includes pending items and identified task allocation results; The acquisition module is further used to acquire a preset MOE network model and multiple LoRA adapters, and insert the multiple LoRA adapters into the gated network and multiple expert networks in the preset MOE network model respectively to generate a preset task allocation model; The selection module is used to select a sample data from the sample data set as the target sample data; The generating module is used to input the target pending items in the target sample data into the preset task allocation model for task allocation, and obtain a predicted task allocation result; The determination module is used to determine whether the preset task allocation model converges according to the predicted task allocation result and the identified task allocation result; The adjustment module is used to adjust the model parameters of at least one of the gating network, the multiple expert networks and the multiple LoRA adapters if the preset task allocation model has not converged; The generating module is used to return to the step of looping and selecting a sample data from the sample data set as the target sample data until the preset task allocation model converges to obtain the target task allocation model; The acquisition module is also used to acquire target pending items; The generation module is used to input the target pending items into the target task allocation model to obtain a task allocation result.
9. A computer device, characterized in that: The computer device comprises a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the task allocation method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the task allocation method according to any one of claims 1 to 7 are implemented.