AI talent training resource allocation method, system, terminal and storage medium

Abstract

The present application relates to a method, system, terminal and storage medium for allocating AI talent training resources, and relates to the field of artificial intelligence education technology, including obtaining student login trigger information; obtaining student personal files based on login trigger information; judging whether the student personal files meet the requirements of the preset first login files; if not, analyzing the student personal files to determine the student's selected learning courses; if yes, controlling the preset computing power management platform to recommend courses to students to determine the selected learning courses; and controlling the computing power management platform to allocate GPU resources to the corresponding students according to the selected learning courses. The present application has the effect of improving the resource utilization rate of GPU computing power resources.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence education technology, and in particular to an AI talent training resource allocation method, system, terminal and storage medium. Background Art

[0002] Artificial intelligence education refers to the integration of AI-related knowledge, skills and technologies into the teaching and learning process, aiming to help students understand the basic concepts, principles and applications of AI.

[0003] In related technologies, in artificial intelligence education, computer experiments can undoubtedly help students transform the abstract theories they have learned into concrete applications. Therefore, teachers or administrators will physically allocate GPUs on each terminal in the computer room so that each student has a dedicated card, allowing students to use the efficient computing performance of GPUs for model training, reasoning, data preprocessing, and multimodal learning.

[0004] With regard to the above-mentioned technologies, every student can enjoy the same GPU computing power resources. However, each student has a different level, so the students' needs for GPU computing power are also different. Some students' experimental content usually only consumes a small part of the GPU computing power, resulting in a waste of GPU computing power resources and low resource utilization of GPU computing power resources. There is still room for improvement. Summary of the invention

[0005] In order to improve the resource utilization rate of GPU computing resources, the present application provides an AI talent training resource allocation method, system, terminal and storage medium.

[0006] In the first aspect, the present application provides a method for allocating AI talent training resources, which adopts the following technical solutions:

[0007] A method for allocating AI talent training resources, comprising:

[0008] Get the student's login trigger information;

[0009] Obtaining student profiles of students based on login trigger information;

[0010] Determine whether the student's personal profile meets the preset first-time login profile requirements;

[0011] If not, the student profile is analyzed to determine the student’s selected course of study;

[0012] If the conditions are met, the preset computing power management platform is controlled to recommend courses to students to determine the selected learning courses;

[0013] The computing power management platform is controlled to allocate GPU resources to corresponding students according to the selected learning courses.

[0014] By adopting the above technical solution, when the student's login trigger information is detected, the student's personal profile is called. If the student's personal profile does not meet the requirements of the first login profile, it means that the student is not logging in for the first time. Therefore, the student's personal profile is directly analyzed to determine the selected learning course. If it meets the requirements, it means that the student is logging in for the first time. Therefore, the computing power management platform recommends courses for the student to determine the selected learning course. Finally, according to the selected learning course, the computing power management platform is controlled to allocate GPU resources to the corresponding student, so that the GPU resources obtained by the student are suitable for the student's current level, and there will be no situation where only part of the resources of a GPU card are used, thereby improving the resource utilization rate of GPU computing power resources.

[0015] Optionally, the steps of controlling the preset computing power management platform to recommend courses to students to determine the selected learning courses include:

[0016] Control the computing power management platform to push preset personal ability test questions to students to test them;

[0017] Obtain students’ competency assessment results;

[0018] Determine recommended learning courses based on competency assessment results and pre-set competency-course relationships;

[0019] According to the recommended learning courses, the computing power management platform is controlled to push courses to students to determine the selected learning courses.

[0020] By adopting the above technical solution, the computing power management platform is controlled to push personal ability test questions to students to test them, thereby determining the recommended learning courses based on the students' ability assessment results and the ability-course relationship, so that the computing power management platform only recommends courses suitable for students' current ability levels, thereby improving the rationality of students' artificial intelligence education.

[0021] Optionally, the steps of controlling the computing power management platform to push courses to students according to the recommended learning courses to determine the selected learning courses include:

[0022] Determine the GPU resources allocated to the course based on the recommended learning courses and the preset course resource relationship;

[0023] Obtain the remaining GPU resources of the computing power management platform;

[0024] Determine whether the GPU resources allocated to the course meet the requirements of the remaining GPU resources;

[0025] If not, obtain the courses that exceed the GPU resource limit and remove them from the recommended courses.

[0026] If it is in compliance, the GPU resources that match the course are obtained, and the computing power management platform is controlled to push the GPU resources that match the course to the student to determine the selected learning course.

[0027] By adopting the above technical solution, the GPU resources allocated to the course are determined according to the relationship between the recommended learning courses and the course resources. When it is determined that the GPU resources allocated to the course do not meet the requirements of the remaining GPU resources, it means that the course requires too many resources and the remaining GPU resources cannot support the learning of the course. Therefore, the course is eliminated. If it meets the requirements, the course is pushed to the students for them to choose and determine the selected learning course, so that the students can start learning immediately after selecting the course, thereby improving the rationality of determining the selected learning course.

[0028] Optionally, the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the selected learning course include:

[0029] Obtain students’ actual study course based on the selected study course;

[0030] Determine the required GPU resources based on the relationship between the actual learning course and the preset course resources;

[0031] According to the demand, GPU resources control computing power management platform allocates GPU resources to corresponding students;

[0032] Get the GPU resource usage of students;

[0033] Determine whether the GPU resource usage meets the GPU resource requirements;

[0034] If it is in compliance, continue to obtain the student's GPU resource usage for cyclic judgment;

[0035] If it does not meet the requirements, the computing power management platform will be controlled to adjust the student's GPU resources based on the GPU resource usage.

[0036] By adopting the above technical solution, the required GPU resources are determined according to the actual learning courses and the relationship between course resources, and the computing power management platform is controlled to allocate GPU resources to corresponding students according to the required GPU resources, so that the GPU resources obtained by the students meet the current needs, and the GPU resource usage is detected. When the GPU resource usage does not meet the requirements of the required GPU resources, the computing power management platform is controlled to adjust the student's GPU resources to prevent the situation where too many GPU resources are allocated to the student, thereby improving the resource utilization rate of the GPU computing power resources.

[0037] Optionally, the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the demand for GPU resources include:

[0038] Obtain the actual GPU resources of the computing power management platform;

[0039] Determine whether the required GPU resources meet the requirements of the actual GPU resources;

[0040] If not, the computing power management platform is controlled to put the required GPU resources into the resource queue to be allocated to wait for resource allocation, and continue to obtain the actual GPU resources of the computing power management platform for cyclic judgment;

[0041] If it matches, the resource usage of the preset GPU card is obtained;

[0042] Determine whether the resource occupancy meets the preset GPU card full occupancy requirement;

[0043] If it meets the requirements, the computing power management platform is controlled according to the preset direct allocation method to allocate GPU resources to the corresponding students based on the required GPU resources;

[0044] If it does not meet the requirements, the computing power management platform will be controlled according to the preset superposition allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0045] By adopting the above technical solution, when the resource occupancy situation meets the situation that the GPU card is fully occupied, it means that the allocated GPU resources will fully occupy some GPU cards, so the new GPU card is directly enabled to allocate resources to the students. If it does not meet the situation, it means that the GPU card is not fully occupied, so the GPU card resources that are not fully occupied are allocated to the students according to the superposition allocation method, thereby improving the resource utilization rate of the GPU computing resources.

[0046] Optionally, the step of controlling the computing power management platform to allocate GPU resources to corresponding students based on the demand for GPU resources according to a preset superposition allocation method includes:

[0047] Obtain the remaining GPU cards and the corresponding remaining GPU card resources;

[0048] Determine whether the required GPU resources meet the requirements of the remaining GPU card resources;

[0049] If it is compliant, obtain the compliant GPU card and compliant GPU card resources;

[0050] Analyze the eligible GPU cards, eligible GPU card resources and required GPU resources to determine the actual GPU card to be superimposed;

[0051] Control the computing power management platform to allocate GPU resources to corresponding students based on the required GPU resources and the actual stacked GPU cards;

[0052] If it does not meet the requirements, the computing power management platform will be controlled according to the preset adjustment allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0053] By adopting the above technical solution, when it is determined that the required GPU resources are in line with the remaining GPU card resources, it means that the GPU card resources that are not fully occupied can be allocated to students. Therefore, the actual superimposed GPU card is determined after analyzing the GPU card resources that meet the requirements and the GPU resources required, so that the GPU resources of the actual superimposed GPU card are allocated to students according to the required GPU resources, thereby ensuring that the resources of the GPU card are fully occupied, thereby improving the resource utilization rate of the GPU computing resources.

[0054] Optionally, the step of controlling the computing power management platform to allocate GPU resources to corresponding students based on the demand for GPU resources according to a preset adjustment allocation method includes:

[0055] Analyze the remaining GPU card resources and required GPU resources to determine the adjustment of GPU card resources;

[0056] Determine whether the adjustment of GPU card resources meets the requirements of the remaining GPU card resources;

[0057] If it does not meet the requirements, the computing power management platform is controlled to allocate GPU resources to the corresponding students according to the preset direct allocation method based on the demand GPU resources;

[0058] If it meets the requirements, the GPU card and GPU resources are adjusted;

[0059] Control the computing power management platform to reallocate GPU resources and control the computing power management platform to adjust GPU cards and required GPU resources to allocate GPU resources to corresponding students.

[0060] By adopting the above technical solution, the remaining GPU card resources and the required GPU resources are analyzed to determine the adjusted GPU card resources. When the adjusted GPU card resources meet the requirements of the remaining GPU card resources, the computing power management platform is controlled to reallocate the GPU resources to the adjusted GPU resources, and the computing power management platform is controlled to allocate the GPU resources to the corresponding students based on the adjusted GPU card and the required GPU resources, thereby making the resource allocation occupancy rate of the GPU card higher, thereby improving the resource utilization rate of the GPU computing power resources.

[0061] In the second aspect, the present application provides an AI talent training resource allocation system, which adopts the following technical solutions:

[0062] An AI talent training resource allocation system, comprising:

[0063] The acquisition module is used to obtain login trigger information and student personal files;

[0064] A memory, used to store a program of the AI ​​talent training resource allocation method described in any one of the above items;

[0065] The program in the memory can be loaded and executed by the processor and implement any of the above-mentioned methods for allocating AI talent training resources.

[0066] By adopting the above technical solution, the processor loads and executes a program of an AI talent training resource allocation method stored in the memory, so that the acquisition module acquires a series of data related to the AI ​​talent training resource allocation, so that when the student's login trigger information is detected, the student's personal profile is called. If the student's personal profile does not meet the requirements of the first login profile, it means that the student is not logging in for the first time. Therefore, the student's personal profile is directly analyzed to determine the selected learning course. If it meets the requirements, it means that the student is logging in for the first time. Therefore, the computing power management platform recommends courses for the student to determine the selected learning course. Finally, according to the selected learning course, the computing power management platform is controlled to allocate GPU resources to the corresponding student, so that the GPU resources obtained by the student are suitable for the student's current level, and there will be no situation where only part of the resources of a GPU card are used, thereby improving the resource utilization rate of GPU computing power resources.

[0067] In a third aspect, the present application provides a smart terminal, which adopts the following technical solution:

[0068] An intelligent terminal includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute any one of the above-mentioned methods for allocating AI talent training resources.

[0069] By adopting the above technical solution, by operating the intelligent terminal, the processor loads and executes a computer program of an AI talent training resource allocation method stored in the memory, so that when the student's login trigger information is detected, the student's personal file is called. If the student's personal file does not meet the requirements of the first login file, it means that the student is not logging in for the first time. Therefore, the student's personal file is directly analyzed to determine the selected learning course. If it meets the requirements, it means that the student is logging in for the first time. Therefore, the computing power management platform recommends courses for the student to determine the selected learning course. Finally, according to the selected learning course, the computing power management platform is controlled to allocate GPU resources to the corresponding student, so that the GPU resources obtained by the student are suitable for the student's current level, and there will be no situation where only part of the resources of a GPU card are used, thereby improving the resource utilization rate of GPU computing power resources.

[0070] In a fourth aspect, the present application provides a computer storage medium capable of storing corresponding programs, which has the characteristics of facilitating the improvement of resource utilization of GPU computing resources, and adopts the following technical solutions:

[0071] A computer-readable storage medium storing a computer program that can be loaded by a processor and execute any of the above-mentioned AI talent training resource allocation methods.

[0072] By adopting the above technical solution, a computer program of an AI talent training resource allocation method is stored in a computer-readable storage medium, so that the processor loads and executes the computer program in the storage medium, so that when the student's login trigger information is detected, the student's personal profile is called. If the student's personal profile does not meet the requirements of the first login profile, it indicates that the student is not logging in for the first time. Therefore, the student's personal profile is directly analyzed to determine the selected learning course. If it meets the requirements, it indicates that the student is logging in for the first time. Therefore, the computing power management platform recommends courses for the student to determine the selected learning course. Finally, according to the selected learning course, the computing power management platform is controlled to allocate GPU resources to the corresponding student, so that the GPU resources obtained by the student are suitable for the student's current level, and there will be no situation where only part of the resources of a GPU card are used, thereby improving the resource utilization rate of GPU computing power resources.

[0073] In summary, the present application includes at least one of the following beneficial technical effects:

[0074] When the student's login trigger information is detected, the student's personal profile is called. If the student's personal profile does not meet the requirements of the first login profile, it means that the student is not logging in for the first time. Therefore, the student's personal profile is directly analyzed to determine the selected learning course. If it meets the requirements, it means that the student is logging in for the first time. Therefore, the computing power management platform recommends courses for the student to determine the selected learning course. Finally, according to the selected learning course, the computing power management platform is controlled to allocate GPU resources to the corresponding student, so that the GPU resources obtained by the student are suitable for the student's current level, and there will be no situation where only part of the resources of a GPU card are used, thereby improving the resource utilization rate of GPU computing resources;

[0075] By controlling the computing power management platform to push personal ability test questions to students to test them, the recommended learning courses are determined based on the students' ability assessment results and the relationship between ability and courses, so that the computing power management platform only recommends courses suitable for students' current ability levels, thereby improving the rationality of students' artificial intelligence education;

[0076] The required GPU resources are determined based on the actual learning courses and the relationship between course resources, and the computing power management platform is controlled to allocate GPU resources to the corresponding students based on the required GPU resources, so that the GPU resources obtained by the students meet the current needs, and the GPU resource usage is detected. When the GPU resource usage does not meet the requirements of the required GPU resources, the computing power management platform is controlled to adjust the student's GPU resources to prevent the situation where too many GPU resources are allocated to the student, thereby improving the resource utilization of GPU computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a flow chart of a method for allocating AI talent training resources in an embodiment of the present application.

[0078] Figure 2 It is a flowchart of the steps of controlling the preset computing power management platform to recommend courses for students to determine the selected learning courses in an embodiment of the present application.

[0079] Figure 3 This is a flowchart of the steps of controlling the computing power management platform to push courses to students according to the recommended learning courses in an embodiment of the present application to determine the selected learning courses.

[0080] Figure 4 It is a flowchart of the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the selected learning course in an embodiment of the present application.

[0081] Figure 5 This is a flowchart of the steps in which the computing power management platform controls GPU resources according to demand in an embodiment of the present application to allocate GPU resources to corresponding students.

[0082] Figure 6 This is a flowchart of the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the preset superposition allocation method in the embodiment of the present application.

[0083] Figure 7 It is a flowchart of the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the required GPU resources according to the preset adjustment and allocation method in an embodiment of the present application. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-Figure 7 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0085] The embodiment of the present application discloses a method for allocating resources for AI talent training, including a computing power management platform, which includes an education system and a resource allocation system. After a student logs in to the education system, the education system detects the student's login trigger information, and calls the student's personal profile in response to the login trigger information. If it is determined that the student's personal profile does not meet the requirements of the first login profile, it indicates that the student is not logging in for the first time, so the student's selected learning course is directly determined from the student's personal profile. If it does not meet the requirements, it indicates that the student is logging in to the education system for the first time, so the education system recommends courses to the student to determine the selected learning course, and then controls the resource allocation system to allocate GPU resources to the corresponding student according to the selected learning course, so that the GPU resources allocated to the student meet the current learning needs, prevent a student from monopolizing a GPU card and not fully using GPU resources, and thereby improve the resource utilization rate of GPU computing resources.

[0086] Reference Figure 1 , the present application embodiment discloses a method for allocating AI talent training resources, comprising the following steps:

[0087] Step S100: Obtain the student's login trigger information.

[0088] The login trigger information refers to the information of the student logging into the education system. When the student logs into the education system, a login request is sent to the education system, so that the education system receives the student's login trigger information.

[0089] Step S101: Obtain the student's personal profile based on the login trigger information.

[0090] Among them, when the education system receives the login trigger information, it indicates that the student has logged into the education system. At this time, the education system responds to the login trigger information, calls the student's personal file according to the account entered by the student, and provides data support for the subsequent determination of the courses that the student wants to study.

[0091] The student's personal profile refers to the student's course progress and test results of artificial intelligence learning in the education system, which is obtained by the education system summarizing the student's learning situation. When the student logs into the education system, the education system calls the student's personal profile and sends it to the resource allocation system.

[0092] Step S102: Determine whether the student's personal profile meets the preset first-time login profile requirements.

[0093] Among them, the first login file refers to the file when the student logs into the education system for the first time, that is, a blank file. The requirements for the first login file refer to that it must be consistent with the first login file.

[0094] The resource allocation system determines whether the student’s personal file is consistent with the first login file, thereby determining whether the student is logging into the education system for the first time.

[0095] Step S1021: If not, the student's personal profile is analyzed to determine the student's selected study course.

[0096] Among them, if the resource allocation system determines that the student's personal file is inconsistent with the first login file, it means that the student is not logging into the education system for the first time. Therefore, the student's personal file is analyzed to determine the student's selected learning course and provide data support for subsequent resource allocation.

[0097] Selecting learning courses means that students select the courses they want to study. The number is not fixed and is selected by the students. The selected learning courses in this step are obtained by the resource allocation system identifying the courses recorded in the student's personal file.

[0098] Step S1022: If it is in compliance, the preset computing power management platform is controlled to recommend courses to the student to determine the selected learning course.

[0099] If the resource allocation system determines that the student's personal profile is consistent with the first login profile, it means that the student is logging into the education system for the first time. Therefore, there is no course information for the student in the student's personal profile. Therefore, the education system is controlled to recommend courses for the student to determine the selected learning course. For specific methods, refer to Figure 2 steps.

[0100] The computing power management platform refers to a platform used for artificial intelligence education and management of GPU resources. In the embodiments of this application, the OrionX software-defined AI computing power resource pooling and the Gemini AI development and training platform are mainly used to provide IaaS and PaaS services to the outside world. OrionX is used to pool GPU card resources and share them, and GPUs are dynamically mounted and released on demand, breaking the resource islands caused by the monopoly of GPU cards, forming a GPU resource pool, and dynamically adjusting the resource ratio of development and training during the day and night, so as to achieve a reasonable allocation of GPU and CPU resources and enable different AI business models to run in parallel on the same GPU card.

[0101] The selected learning courses in this step are consistent with the definition of the selected learning courses in step S1021, and the courses are recommended to students by the education system and selected by the students.

[0102] Step S103: Control the computing power management platform to allocate GPU resources to corresponding students according to the selected learning course.

[0103] Among them, after the education system determines the selected learning course, the resource allocation system allocates the corresponding GPU resources to the corresponding students according to the selected learning course. The specific method is referred to Figure 4 The steps are carried out so that the GPU resources obtained by students are suitable for their current learning, and there will be no situation where a GPU card is monopolized by one student and not fully occupied, thereby improving the resource utilization of GPU computing resources.

[0104] Reference Figure 2 , the steps of controlling the preset computing power management platform to recommend courses for students to determine the selected learning courses include:

[0105] Step S200: Control the computing power management platform to push preset personal ability test questions to students to test them.

[0106] Among them, the education system in the control computing power management platform pushes personal ability test questions to students, allowing students to take tests and provide data support for the subsequent determination of pushed courses.

[0107] Personal ability test questions refer to test questions that test students' basic knowledge of AI and personal abilities, covering multiple dimensions such as programming skills, mathematical foundations, and logical reasoning. The specific questions are determined by the operator based on actual conditions.

[0108] Step S201: Obtain the student's ability assessment results.

[0109] Among them, the ability assessment results refer to the results of students' tests, which are obtained by the education system collecting students' test results and are specifically expressed in different scores.

[0110] Step S202: Determine recommended learning courses based on the ability assessment results and the preset ability-course relationship.

[0111] Among them, the ability-course relationship refers to the correspondence between different abilities and learning courses. The higher the ability, the higher the difficulty of the required learning course. The operator forms a mapping table by matching different abilities with learning courses one by one.

[0112] Recommended learning courses refer to courses that are suitable for students' current level and are found by the education system in the mapping table corresponding to the ability-course relationship based on the results of ability assessment.

[0113] Step S203: Control the computing power management platform to push courses to students based on the recommended learning courses to determine the selected learning courses.

[0114] Among them, after the education system determines the recommended learning courses, the education system is controlled to push courses to students according to the recommended learning courses, so that students can choose and determine the selected learning courses. For specific methods, refer to Figure 3 steps.

[0115] Reference Figure 3 , the steps of controlling the computing power management platform to push courses to students according to the recommended learning courses to determine the selected learning courses include:

[0116] Step S300: Determine the course allocation GPU resources according to the recommended learning course and the preset course resource relationship.

[0117] The course resource relationship refers to the correspondence between different courses and the required GPU resources. Generally speaking, the higher the difficulty of the course, the more GPU resources are required. The operator forms a mapping table by mapping different courses to the required GPU resources one by one.

[0118] Course allocation GPU resources refer to the GPU resources required for the recommended courses, which are found by the education system in the mapping table corresponding to the course resource relationship based on the recommended learning courses.

[0119] Step S301: Obtain the remaining GPU resources of the computing power management platform.

[0120] Among them, the remaining GPU resources refer to the resources that have not been occupied in the GPU resource pool. The resource allocation system monitors the GPU resource pool and sends it to the education system.

[0121] Step S302: Determine whether the GPU resources allocated to the course meet the requirements of the remaining GPU resources.

[0122] The requirement of remaining GPU resources refers to not exceeding the remaining GPU resources. The education system determines whether the GPU resources allocated for the course do not exceed the remaining GPU resources, thereby determining whether the student can immediately learn the course.

[0123] Step S3021: If not, obtain the courses where GPU resources exceed the limit, and remove the courses where GPU resources exceed the limit from the recommended learning courses.

[0124] Among them, if the education system determines that the GPU resources allocated for a course exceed the remaining GPU resources, it means that the students cannot study the course immediately and need to wait in line. Therefore, the course is defined as a GPU resource excess course and the GPU resource excess course is removed from the recommended learning courses.

[0125] GPU resource excess courses refer to courses whose required GPU resources exceed the remaining GPU resources, which are determined by the education system by comparing the required GPU resources with the remaining GPU resources.

[0126] Step S3022: If it is in compliance, then obtain GPU resources that match the course, and control the computing power management platform to push the GPU resources that match the course to the student to determine the selected learning course.

[0127] Among them, if the education system determines that the GPU resources allocated for the course do not exceed the remaining GPU resources, it means that the students can study the course immediately, so the course is defined as a GPU resource-compliant course, and the education system is controlled to push the GPU resource-compliant course to the students, so that the students can select and determine the selected learning course.

[0128] GPU resource eligible courses refer to courses whose required GPU resources do not exceed the remaining GPU resources, which are determined by the education system by comparing the required GPU resources with the remaining GPU resources.

[0129] Reference Figure 4 , the steps of controlling the computing power management platform to allocate GPU resources to corresponding students according to the selected learning course include:

[0130] Step S400: Acquire the actual learning courses of the students based on the selected learning courses.

[0131] The actual learning course refers to the course that the student currently chooses to study among the selected learning courses, which is obtained by the education system identifying the student's choice.

[0132] Step S401: Determine the required GPU resources according to the relationship between the actual learning course and the preset course resources.

[0133] The course resource relationship in this step is consistent with the course resource relationship in step S300, and will not be described in detail here.

[0134] The required GPU resources refer to the GPU resources required by the student's current learning course, which are found by the education system from the mapping table corresponding to the course resource relationship based on the actual learning course.

[0135] Step S402: The GPU resource control computing power management platform allocates GPU resources to corresponding students according to the demand.

[0136] After the education system determines the required GPU resources, the resource allocation system allocates GPU resources to the corresponding students based on the required GPU resources. For specific methods, refer to Figure 5 steps, so that the GPU resources obtained by students are suitable for their use, and there will be no situation where students monopolize the GPU card and cause insufficient use of resources.

[0137] Step S403: Obtain the student's GPU resource usage.

[0138] Among them, GPU resource usage refers to the students' usage of the allocated GPU resources, which is obtained by checking the resource usage by the resource allocation system.

[0139] Step S404: Determine whether the GPU resource usage meets the GPU resource requirement.

[0140] The requirement of the required GPU resources refers to being within a fluctuation range of the required GPU resources, and the size of the fluctuation range is determined by an operator according to actual conditions.

[0141] The resource allocation system determines whether the GPU resource usage is within the fluctuation range of the required GPU resources, thereby determining whether the students are fully utilizing the allocated GPU resources.

[0142] Step S4041: If it is in compliance, continue to obtain the student's GPU resource usage for cyclic judgment.

[0143] Among them, if the resource allocation system determines that the GPU resource usage is within the fluctuation range of the required GPU resources, it means that the student fully uses the allocated GPU resources. Therefore, the student's GPU resource usage continues to be detected, thereby continuously monitoring changes in the student's GPU resource usage.

[0144] Step S4042: If not, the computing power management platform is controlled to adjust the student's GPU resources according to the GPU resource usage.

[0145] Among them, if the resource allocation system determines that the GPU resource usage is not within the fluctuation range of the required GPU resources, it means that the student has not fully used the allocated GPU resources. Therefore, the resource allocation system is controlled to adjust the student's GPU resources according to the GPU resource usage. The specific adjustment method can be to analyze the maximum value in the GPU resource usage and adjust the GPU resources with the maximum value, or to calculate the average value or the latest value, and then adjust it with the average value or the latest value.

[0146] Reference Figure 5 The steps for the GPU resource control computing power management platform to allocate GPU resources to corresponding students according to the demand include:

[0147] Step S500: Obtain actual GPU resources of the computing power management platform.

[0148] The actual GPU resources refer to the resources currently remaining in the GPU resource pool, which are obtained by monitoring the GPU resource pool by the resource allocation system.

[0149] Step S501: Determine whether the required GPU resources meet the requirements of the actual GPU resources.

[0150] The actual GPU resource requirement means that it does not exceed the actual GPU resources. The resource allocation system determines whether the required GPU resources do not exceed the actual GPU resources, thereby determining whether GPU resources can be allocated to students in time when students actually start learning the course.

[0151] Step S5011: If not, the computing power management platform is controlled to put the required GPU resources into the resource queue to be allocated to wait for resource allocation, and continue to obtain the actual GPU resources of the computing power management platform for cyclic judgment.

[0152] Among them, if the resource allocation system determines that the required GPU resources exceed the actual GPU resources, it means that the remaining GPU resources in the GPU resource pool are currently insufficient to support students' course learning. Therefore, the resource allocation system will put the required GPU resources in the queue of resources to be allocated and continue to detect the actual GPU resources, so as to wait for the excess GPU resources to be released and allocate GPU resources to students in time.

[0153] Step S5012: If it is in compliance, the resource usage of the preset GPU card is obtained.

[0154] Among them, if the resource allocation system determines that the required GPU resources do not exceed the actual GPU resources, it means that the remaining GPU resources in the GPU resource pool are currently sufficient to support students' course learning. Therefore, the resource usage of the GPU card is detected to provide data support for the subsequent specific allocation of GPU resources.

[0155] The GPU card refers to the physical GPU card that forms the GPU resource pool. The resource occupancy refers to the resource occupancy ratio of the enabled GPU card, which is obtained by the resource allocation system by detecting the resource allocation ratio of the enabled GPU card.

[0156] Step S502: Determine whether the resource occupancy meets the preset GPU card full occupancy requirement.

[0157] The GPU card full occupancy condition refers to a condition where all resources of the GPU card are occupied, and the requirement for the GPU card full occupancy condition refers to being consistent with the GPU card full occupancy condition.

[0158] The resource allocation system determines whether the resource occupancy is consistent with the full occupancy of the GPU card, thereby determining whether the currently running task has occupied all the resources of the enabled GPU card.

[0159] Step S5021: If it meets the requirements, the computing power management platform is controlled according to the preset direct allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0160] Among them, if the resource allocation system determines that the resource occupancy is consistent with the full occupancy of the GPU card, it means that the currently running task has occupied all the resources of the enabled GPU card. Therefore, the resource allocation system is controlled according to the direct allocation method to enable a new GPU card, and the GPU card resources are allocated to the corresponding students based on the required GPU resources.

[0161] The direct allocation method refers to a method of directly enabling a new GPU card and allocating the GPU card resources to students, which is stored in the resource allocation system by the operator.

[0162] Step S5022: If not, the computing power management platform is controlled according to the preset superposition allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0163] If the resource allocation system determines that the resource occupancy is inconsistent with the GPU card full occupancy, it means that the currently running task has not occupied all the resources of the enabled GPU card, and the enabled GPU card still has surplus resources. Therefore, the resource allocation system is controlled according to the superposition allocation method to allocate GPU resources to the corresponding students according to the demand GPU resources. For specific methods, refer to Figure 6 steps.

[0164] The superposition allocation method refers to a method of superimposing the student's GPU demand resources on the already enabled GPU card so that multiple tasks can run in parallel on the same GPU card, which is stored in the resource allocation system by the operator.

[0165] Reference Figure 6 The steps of controlling the computing power management platform to allocate GPU resources to corresponding students based on the demand for GPU resources according to the preset superposition allocation method include:

[0166] Step S600: Acquire remaining GPU cards and corresponding remaining GPU card resources.

[0167] The remaining GPU card refers to a GPU card with remaining resources, and the remaining GPU card resources refer to the total occupied resources of the remaining GPU card. The resource allocation system detects the resources of the enabled GPU card. If the resources are not occupied by 100%, the GPU card is a remaining GPU card, and the occupied resources of the GPU card are detected to obtain the remaining GPU card resources.

[0168] Step S601: Determine whether the required GPU resources meet the requirements of the remaining GPU card resources.

[0169] The requirement of remaining GPU card resources means that the sum of the resources with the remaining GPU card resources does not exceed 100%. The resource allocation system determines whether the sum of the required GPU resources and the remaining GPU card resources does not exceed 100%, thereby determining whether the resources required for the student's task can be superimposed on the enabled GPU card.

[0170] Step S6011: If compliant, obtain a compliant GPU card and compliant GPU card resources.

[0171] Among them, if the resource allocation system determines that the sum of the required GPU resources and the remaining GPU card resources does not exceed 100%, it means that the resources required for the student's task can be superimposed on the enabled GPU card. Therefore, the compliant GPU card and compliant GPU card resources are detected to provide data support for the subsequent superposition and allocation of resources.

[0172] A compliant GPU card refers to a GPU card whose total remaining GPU card resources and required GPU resources do not exceed 100%. A compliant GPU card resource refers to the current total resources of a compliant GPU card, which is obtained by detecting the compliant GPU card by the resource allocation system.

[0173] Step S60111: Analyze the eligible GPU cards, eligible GPU card resources, and required GPU resources to determine the actual stacked GPU card.

[0174] The actual stacked GPU card refers to the GPU card that actually performs resource stacking. The resource allocation system calculates the sum of eligible GPU card resources and required GPU resources, and selects the eligible GPU card with the largest sum as the actual stacked GPU card.

[0175] Step S60112: Control the computing power management platform to allocate GPU resources to corresponding students based on the required GPU resources and the actual stacked GPU cards.

[0176] Among them, after determining the actual superimposed GPU card, the resource allocation system is controlled to allocate the GPU resources of the actual superimposed GPU card to the corresponding students according to the required GPU resources, so as to allocate the resources of the enabled GPU card to the students, thereby improving the utilization rate of the GPU resources.

[0177] Step S6012: If not, the computing power management platform is controlled according to the preset adjustment allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0178] If the resource allocation system determines that the sum of the required GPU resources and the remaining GPU card resources exceeds 100%, it means that the resources required for the student's task cannot be directly superimposed on the enabled GPU card. Therefore, the resource allocation system is controlled according to the allocation adjustment method to allocate GPU resources to the corresponding student based on the required GPU resources. For specific methods, refer to Figure 7 steps.

[0179] The adjustment allocation method refers to a method of adjusting existing tasks on a GPU card to other GPU cards, thereby releasing the resources of the GPU card for use by another student, and is stored in the resource allocation system by the operator.

[0180] Reference Figure 7 The steps of controlling the computing power management platform to allocate GPU resources to corresponding students based on the demand for GPU resources according to the preset adjustment allocation method include:

[0181] Step S700: analyzing the remaining GPU card resources and the required GPU resources to determine the adjustment of the GPU card resources.

[0182] Among them, adjusting the GPU card resources refers to the sum of occupied resources obtained after replacing a task resource of the remaining GPU card with the required GPU resource. The resource allocation system simulates the exchange of each task resource in the remaining GPU card resources with the required GPU resource for all the remaining GPU cards one by one, and selects the largest total resource as the adjusted GPU card resources.

[0183] Step S701: Determine whether the adjustment of GPU card resources meets the requirement of remaining GPU card resources.

[0184] The remaining GPU card resources requirement in this step refers to exceeding the remaining GPU card resources. The resource allocation system determines whether the adjusted GPU card resources exceed the remaining GPU card resources, thereby determining whether the adjusted GPU resources have a higher utilization rate than before the adjustment.

[0185] Step S7011: If not, the computing power management platform is controlled according to the preset direct allocation method to allocate GPU resources to the corresponding students based on the required GPU resources.

[0186] Among them, if the resource allocation system determines that the adjustment of GPU card resources does not exceed the remaining GPU card resources, it means that the resource utilization rate after adjustment is even lower. Therefore, no adjustment allocation is made, and the resource allocation system is controlled according to the direct allocation method to start a new GPU card, and the GPU resources of the GPU card are allocated to the corresponding students based on the required GPU resources.

[0187] The direct allocation method in this step is consistent with the direct allocation method in step S5021, and will not be described in detail here.

[0188] Step S7012: If it is in compliance, obtain adjustment of the GPU card and adjustment of GPU resources.

[0189] If the resource allocation system determines that the adjusted GPU card resources exceed the remaining GPU card resources, it indicates that the resource utilization rate after adjustment is higher. Therefore, the adjustment of the GPU card and the adjustment of the GPU resources are detected to provide data support for subsequent GPU resource allocation.

[0190] Adjusting GPU cards refers to GPU cards that need to be adjusted in resources. The resource allocation system compares all adjustment GPU card resources that exceed the remaining GPU card resources, and selects the GPU card corresponding to the largest adjustment GPU card resources as the adjustment GPU card. Adjusting GPU resources refers to task resources on the adjustment GPU card that need to be exchanged with the required GPU resources, which are obtained by the resource allocation system identifying the remaining GPU card resources to be adjusted to the replacement resources of the adjustment GPU card resources.

[0191] Step S702: Control the computing power management platform to reallocate the GPU resources, and control the computing power management platform to allocate GPU resources to corresponding students by adjusting the GPU card and the required GPU resources.

[0192] Among them, after determining to adjust the GPU card and adjust the GPU resources, the adjusted GPU resources are released from the adjusted GPU card and allocated to the remaining GPU cards by the above-mentioned direct allocation or superimposed allocation or adjustment allocation method, and the resource allocation system is controlled to allocate the GPU resources of the adjusted GPU card to the corresponding students according to the demand GPU resources, thereby improving the utilization rate of GPU resources.

[0193] Based on the same inventive concept, the embodiment of the present application provides an AI talent training resource allocation system, including:

[0194] The acquisition module is used to obtain login trigger information, student personal profile, ability assessment results, remaining GPU resources, GPU resources exceeding courses, GPU resources meeting courses, actual learning courses, GPU resource usage, actual GPU resources, resource occupancy, remaining GPU card resources, meeting GPU cards, meeting GPU card resources, adjusting GPU cards, and adjusting GPU resources;

[0195] A memory, used to store a program of a method for allocating AI talent training resources;

[0196] The program in the processor memory can be loaded and executed by the processor to implement an AI talent training resource allocation method.

[0197] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0198] An embodiment of the present application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed by a method for allocating resources for AI talent training.

[0199] Computer storage media include, for example, various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0200] Based on the same inventive concept, an embodiment of the present application provides an intelligent terminal, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes a method for allocating AI talent training resources.

[0201] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0202] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for allocating AI talent training resources, characterized in that: include: Get the student's login trigger information; Obtaining student profiles of students based on login trigger information; Determine whether the student's personal profile meets the preset first-time login profile requirements; If not, the student profile is analyzed to determine the student’s selected course of study; If the conditions are met, the preset computing power management platform is controlled to recommend courses to students to determine the selected learning courses; Control the computing power management platform to allocate GPU resources to corresponding students based on the selected learning courses; The step of controlling the computing power management platform to allocate GPU resources to corresponding students according to the selected learning courses includes: obtaining the actual learning courses of students based on the selected learning courses; determining the required GPU resources according to the relationship between the actual learning courses and the preset course resources; controlling the computing power management platform to allocate GPU resources to corresponding students according to the required GPU resources; obtaining the GPU resource usage of students; judging whether the GPU resource usage meets the requirements of the required GPU resources; if yes, continuing to obtain the GPU resource usage of students for cyclic judgment; if no, controlling the computing power management platform to adjust the GPU resources of students according to the GPU resource usage; The step of controlling the computing power management platform to allocate GPU resources to corresponding students according to the required GPU resources includes: obtaining the actual GPU resources of the computing power management platform; judging whether the required GPU resources meet the requirements of the actual GPU resources; if not, controlling the computing power management platform to put the required GPU resources into the resource queue to be allocated to wait for resource allocation, and continuing to obtain the actual GPU resources of the computing power management platform for cyclic judgment; if yes, obtaining the resource occupancy of the preset GPU card; judging whether the resource occupancy meets the requirements of the preset GPU card full occupancy; if yes, controlling the computing power management platform to allocate GPU resources to corresponding students according to the required GPU resources according to the preset direct allocation method; if not, controlling the computing power management platform to allocate GPU resources to corresponding students according to the required GPU resources according to the preset superposition allocation method; The steps of controlling the computing power management platform to allocate GPU resources to corresponding students based on the required GPU resources according to the preset superposition allocation method include: obtaining the remaining GPU cards and the corresponding remaining GPU card resources; judging whether the required GPU resources meet the requirements of the remaining GPU card resources; if so, obtaining the compliant GPU cards and the compliant GPU card resources; analyzing the compliant GPU cards, the compliant GPU card resources and the required GPU resources to determine the actual superimposed GPU card, the actual superimposed GPU card refers to the GPU card that actually performs resource superposition, and the resource allocation system calculates the sum of the compliant GPU card resources and the required GPU resources, and selects the compliant GPU card with the largest sum as the actual superimposed GPU card; controlling the computing power management platform to allocate GPU resources to corresponding students based on the required GPU resources and the actual superimposed GPU card; if not, controlling the computing power management platform to allocate GPU resources to corresponding students based on the required GPU resources according to the preset adjustment allocation method.

2. The method for allocating AI talent training resources according to claim 1, characterized in that: The steps of controlling the preset computing power management platform to recommend courses to students to determine the selected learning courses include: Control the computing power management platform to push preset personal ability test questions to students to test them; Obtain students’ competency assessment results; Determine recommended learning courses based on competency assessment results and pre-set competency-course relationships; According to the recommended learning courses, the computing power management platform is controlled to push courses to students to determine the selected learning courses.

3. The method for allocating AI talent training resources according to claim 2, characterized in that: The steps of controlling the computing power management platform to push courses to students based on the recommended learning courses to determine the selected learning courses include: Determine the GPU resources allocated to the course based on the recommended learning courses and the preset course resource relationship; Obtain the remaining GPU resources of the computing power management platform; Determine whether the GPU resources allocated to the course meet the requirements of the remaining GPU resources; If not, obtain the courses that exceed the GPU resource limit and remove them from the recommended courses. If it is in compliance, the GPU resources that match the course are obtained, and the computing power management platform is controlled to push the GPU resources that match the course to the student to determine the selected learning course.

4. The method for allocating AI talent training resources according to claim 1, characterized in that: The steps of controlling the computing power management platform to allocate GPU resources to corresponding students based on the demand for GPU resources according to the preset adjustment allocation method include: Analyze the remaining GPU card resources and required GPU resources to determine the adjustment of GPU card resources; Determine whether the adjustment of GPU card resources meets the requirements of the remaining GPU card resources; If it does not meet the requirements, the computing power management platform is controlled to allocate GPU resources to the corresponding students according to the preset direct allocation method based on the demand GPU resources; If it meets the requirements, the GPU card and GPU resources are adjusted; Control the computing power management platform to reallocate GPU resources and control the computing power management platform to adjust GPU cards and required GPU resources to allocate GPU resources to corresponding students.

5. An AI talent training resource allocation system, characterized in that: include: The acquisition module is used to obtain login trigger information and student personal profiles; A memory, used to store a program of an AI talent training resource allocation method according to any one of claims 1 to 4; The program in the memory can be loaded and executed by the processor and implements an AI talent training resource allocation method as described in any one of claims 1 to 4.

6. An intelligent terminal, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes a method for allocating AI talent training resources according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes a method for allocating AI talent training resources according to any one of claims 1 to 4.

Images