A computing power scheduling and processing method and system for cloud resources

Abstract

The present invention provides a computing power scheduling and processing method and system for cloud resources, belonging to the technical field of cloud scheduling. Specifically, it includes: determining the matching training requirements of the available working space based on the available training resources in the available working space, determining the training frequency coefficients and frequent training requirements of different matching training requirements according to the historical training data corresponding to different matching training requirements, determining the number of available working spaces corresponding to different frequent training requirements, and combining the frequent training coefficients of different matching training requirements to determine the training requirement weight coefficients of different frequent training requirements. Based on the training requirement weight coefficients of different available working spaces in different frequent training requirements, the matching working spaces in the available working spaces are determined, and the matching working spaces are pushed to the customer, improving the reliability of training processing.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud scheduling, and particularly relates to a computing power scheduling processing method and system for cloud resources. Background Art

[0002] In order to implement the computing power scheduling processing of cloud resources and meet the personalized model training requirements of customers, in the invention patent application CN202010715540.3 "A Resource Scheduling Method for Hybrid Cloud", by calculating the delay time of different scheduling schemes and comparing it with the longest processing time submitted by the user, after multiple rounds of optimization, continuously select a cloud resource scheduling scheme that meets the time constraint but is cheaper in price, thus greatly ensuring the maximization of the utilization rate of private cloud resources. However, there are the following technical problems:

[0003] When performing the computing power scheduling processing of cloud resources, the existing technical solutions ignore the determination of personalized cloud resource scheduling strategies according to the needs of customers. Due to the large differences in the needs of customers, if the personalized cloud resource scheduling strategies cannot be determined according to the needs of customers, the reliability of the cloud resource scheduling processing cannot be improved.

[0004] In view of the above technical problems, specifically, the present application provides a computing power scheduling processing method and system for cloud resources. Summary of the Invention

[0005] To achieve the object of the present invention, the present invention adopts the following technical solutions:

[0006] According to one aspect of the present invention, there is provided a computing power scheduling processing method for cloud resources.

[0007] A computing power scheduling processing method for cloud resources specifically includes:

[0008] S1 Based on the training requirements of the customer, determine the training requirement feature quantity of the customer. Based on the training requirement feature quantity of the customer and the existing training data of the existing idle workspace, determine the similar training states in the historical training state of the existing idle workspace. According to the training problem data of different idle workspaces in the similar training state, determine the available workspace in the idle workspace;

[0009] S2 Based on the available training resources of the available workspace, determine the matching training requirements of the available workspace. According to the historical training data corresponding to different matching training requirements, determine the training frequency coefficient and frequent training requirements of different matching training requirements;

[0010] S3 determines the number of available workspaces corresponding to different frequent training requirements, and combines the frequent training coefficients that match different training requirements to determine the training requirement weight coefficients for different frequent training requirements;

[0011] S4 determines the matching workspaces in the available workspaces based on the training requirement weight coefficients of different available workspaces for different frequent training requirements, and pushes the matching workspaces to the customer.

[0012] The beneficial effects of the present invention are as follows:

[0013] Based on the training problem data of different idle workspaces in a similar training state, the available workspaces in the idle workspaces are determined, thereby realizing the determination of the available workspaces in the idle workspaces from the problem situations in a similar training state under different idle workspaces, avoiding the occurrence of problems in training caused by the mismatch of the hardware devices of the idle workspaces, and improving the reliability of the training use processing of the customer.

[0014] Based on the training requirement weight coefficients of different available workspaces for different frequent training requirements, the matching workspaces in the available workspaces are determined, thereby avoiding the influence of the use of the available workspaces due to the large number of frequent training requirements with matching frequent training requirements or the large number of frequent training requirements with a high degree of training requirements on other frequent training requirements, meeting the use requirements of other frequent training requirements while also meeting the use requirements of the customer.

[0015] A further technical solution lies in that the training requirement feature quantities of the customer include a training model and a training data volume.

[0016] A further technical solution lies in that the existing training data of the idle workspace includes the number of GPUs used, the existing training models, and the training data volumes of different existing training models.

[0017] A further technical solution lies in that the method for determining similar training states in the historical training state is as follows:

[0018] Based on the existing training data of the existing idle workspaces, the existing training models of the idle workspaces and the training data volumes of different existing training models are determined;

[0019] Based on the training model of the customer and the existing training models, the matching training models are determined, and based on the deviation rate of the training data volumes of different matching training models in the historical training state, the training deviation coefficients for different matching training models are determined;

[0020] Determine a comprehensive deviation coefficient based on the average value of the training deviation coefficients of different matching training models, and use the comprehensive deviation coefficient to determine the similar training states in the historical training state.

[0021] A further technical solution is that the similar training states in the historical training state are the historical training states in which the comprehensive deviation coefficient is within a preset deviation coefficient range.

[0022] A further technical solution is that the method for determining the matching workspaces in the available workspaces is as follows:

[0023] Based on the training requirement weight coefficients of the available workspaces for different frequent training requirements, determine the screening training requirements for the frequent training requirements corresponding to the available workspaces;

[0024] Determine whether the available workspace is a matching workspace based on the number of the screening training requirements.

[0025] A further technical solution is that the screening training requirements are the frequent training requirements with training requirement weight coefficients greater than a preset weight coefficient.

[0026] A further technical solution is that when the number of the screening training requirements is greater than a preset training requirement quantity threshold, it is determined that the available workspace is a matching workspace.

[0027] In a second aspect, the present invention provides a computer system, including: a memory and a processor connected by communication, and a computer program stored on the memory and capable of running on the processor, where when the processor runs the computer program, it executes the above-mentioned method for processing the computing power scheduling of cloud resources.

[0028] Other features and advantages will be described in the subsequent specification, and the objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the specification and the drawings.

[0029] To make the above-mentioned objectives, features, and advantages of the present invention more obvious and understandable, the following specifically gives preferred embodiments and, in conjunction with the accompanying drawings, makes detailed descriptions as follows. Description of the Drawings

[0030] By referring to the accompanying drawings and describing its exemplary embodiments in detail, the above and other features and advantages of the present invention will become more obvious.

[0031] Figure 1 is a flowchart of a method for processing the computing power scheduling of cloud resources;

[0032] Figure 2 is a flowchart of a method for determining the similar training states in the historical training state;

[0033] Figure 3 A flowchart of a method for determining available workspaces in an idle workspace;

[0034] Figure 4 A flowchart of a method for determining the training frequency coefficient that matches the training requirements;

[0035] Figure 5 A framework diagram of a computer system. Detailed implementation manners

[0036] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0037] When performing the computing power scheduling process of cloud resources, due to the large differences in the needs of customers, it is necessary to determine personalized cloud resource scheduling strategies according to the needs of customers, so as to improve the reliability of the cloud resource scheduling process.

[0038] Similar training status: Based on the customer's training model and the existing training model, determine the matching training model. Based on the deviation rate of the training data volume in different matching training models in the historical training status, determine the training deviation coefficient for different matching training models. Based on the average value of the training deviation coefficients for different matching training models, determine the comprehensive deviation coefficient. Consider the historical training status with a comprehensive deviation coefficient less than 0.2 as the similar training status.

[0039] Available workspaces in the idle workspace: Based on the training problem data of the idle workspace in the similar training status, determine the similar training status with training problems and use it as the problem training status. When the proportion of the problem training status in the similar training status is less than 0.1, determine that the idle workspace is an available workspace.

[0040] Training frequency coefficient that matches the training requirements: Based on the historical training data corresponding to the matching training requirements, determine the historical training times corresponding to the matching training requirements. Based on the preset frequency coefficient corresponding to the historical training times, determine the training frequency coefficient for the matching training requirements. Consider the matching training requirements with a training frequency coefficient greater than 0.6 as frequent training requirements.

[0041] Training requirement weight coefficient: Based on the number of available workspaces corresponding to frequent training requirements, determine the number of available spaces for the frequent training requirements, and use the number of available spaces to determine the preset training availability coefficient for the frequent training requirements. Based on the frequent training coefficient that matches the training requirements, determine the training requirement weight coefficient for the frequent training requirements based on the ratio of the frequent training coefficient to the preset training availability coefficient.

[0042] Matched workspace: Based on the sum of the training requirement weight coefficients of the available workspaces for different frequent training requirements, determine the sum of the weight coefficients of the available workspaces. When the sum of the weight coefficients of the available workspaces is greater than the preset coefficient threshold, it is determined that the available workspace does not belong to the matched workspace.

[0043] Embodiment 1 To solve the above problems, according to one aspect of the present invention, as Figure 1 shown, a first aspect is provided. The present invention proposes a computing power scheduling and processing method for cloud resources, specifically including:

[0044] S1 Based on the training requirements of the customer, determine the training requirement feature amount of the customer. Using the training requirement feature amount of the customer and the existing training data of the existing idle workspaces as a basis, determine the similar training states in the historical training states of the existing idle workspaces. Based on the training problem data of different idle workspaces in the similar training states, determine the available workspaces in the idle workspaces;

[0045] Further, the training requirement feature amount of the customer includes the training model and the amount of training data.

[0046] Specifically, the existing training data of the idle workspaces includes the number of GPUs used, the existing training models, and the amount of training data for different existing training models.

[0047] Specifically, as Figure 2 shown, the method for determining the similar training states in the historical training states is:

[0048] Based on the existing training data of the existing idle workspaces, determine the existing training models of the idle workspaces and the amount of training data for different existing training models;

[0049] Based on the training model of the customer and the existing training models, determine the matching training models. Based on the deviation rate of the amount of training data for different matching training models in the historical training states, determine the training deviation coefficients for different matching training models;

[0050] Determine a comprehensive deviation coefficient based on the average value of the training deviation coefficients of different matching training models, and use the comprehensive deviation coefficient to determine the similar training states in the historical training state.

[0051] Further, the similar training states in the historical training state are the historical training states where the comprehensive deviation coefficient is within a preset deviation coefficient range.

[0052] In another embodiment, the method for determining the similar training states in the historical training state is as follows:

[0053] Based on the existing training data of the existing idle working space, determine the existing training models of the idle working space and the training data volumes of different existing training models;

[0054] Based on the customer's training model and the existing training models, determine the matching training models;

[0055] Determine the comprehensive deviation coefficient based on the proportion of the deviation quantity of the matching training model, and use the comprehensive deviation coefficient to determine the similar training states in the historical training state.

[0056] Further, the comprehensive deviation coefficient is determined according to the ratio of the deviation quantity of the matching training model to the number of matching training models.

[0057] In another embodiment, the method for determining the similar training states in the historical training state is as follows:

[0058] S11 Based on the existing training data of the existing idle working space, determine the existing training models of the idle working space and the training data volumes of different existing training models, based on the customer's training model and the existing training models, determine the matching training models, and use the proportion of the deviation quantity of the matching training model with the historical training state to determine the proportion of the model quantity deviation;

[0059] Optionally, the above step S11 includes the following content:

[0060] S111 Based on the existing training data of the existing idle working space, determine the existing training models of the idle working space and the training data volumes of different existing training models, based on the customer's training model and the existing training models, determine the matching training models. When there is no matching training model in the historical training state, it is determined that the historical training state does not belong to the similar training state. When there is a matching training model in the historical training state, proceed to step S112;

[0061] S112 uses the matching training model with a deviation from the historical training state as the deviation training model. When the number of the deviation training models does not meet the requirements, it is determined that the historical training state does not belong to the similar training state. When the number of the deviation training models meets the requirements, it proceeds to step S113;

[0062] S113 determines the model quantity deviation ratio based on the proportion of the number of the deviation training models in the number of the matching training models. When the model quantity deviation ratio is greater than the preset ratio, it proceeds to step S114. When the model quantity deviation ratio is not greater than the preset ratio, it proceeds to step S12;

[0063] S114 When the number of the deviation training models is greater than the preset number of deviation models, it is determined that the historical training state does not belong to the similar training state. When the number of the deviation training models is not greater than the preset number of deviation models, it proceeds to step S12.

[0064] S12 determines the data quantity deviation coefficients under different matching training models according to the deviation quantity of the training data volume under different matching training models;

[0065] Optionally, the above step S12 includes the following content:

[0066] S121 determines the total deviation of the training data volume according to the deviation quantity of the training data volume under different matching training models. When the total deviation of the training data volume does not meet the requirements, it is determined that the historical training state does not belong to the similar training state. When the total deviation of the training data volume meets the requirements, it proceeds to step S122;

[0067] S122 obtains the training data volume under the deviation training model. When the ratio of the training data volume under the deviation training model to the total training data volume under the matching training model is greater than the preset data volume ratio, it is determined that the historical training state does not belong to the similar training state. When the ratio of the training data volume under the deviation training model to the total training data volume under the matching training model is not greater than the preset data volume ratio, it proceeds to step S123;

[0068] S123 determines the data quantity deviation coefficients under different matching training models according to the deviation quantity of the training data volume under different matching training models. When there is a matching training model with a data quantity deviation coefficient that does not meet the requirements, it proceeds to step S124. When there is no matching training model with a data quantity deviation coefficient that does not meet the requirements, it proceeds to step S13;

[0069] When the number of matching training models for which the data volume deviation coefficient does not meet the requirements is greater than the preset model number threshold, it is determined that the historical training state does not belong to the similar training state. When the number of matching training models for which the data volume deviation coefficient does not meet the requirements is not greater than the preset model number threshold, proceed to step S13.

[0070] S13 Determine the comprehensive deviation coefficient based on the data volume deviation coefficient and the model number deviation ratio under different matching training models, and use the comprehensive deviation coefficient to determine the similar training state in the historical training state.

[0071] Specifically, as Figure 3 shown, the method for determining the available working space in the idle working space is:

[0072] Based on the training problem data of the idle working space in the similar training state, determine the similar training state with training problems and use it as the problem training state;

[0073] Determine whether the idle working space is an available working space according to the number of the problem training states.

[0074] Further, when the number of the problem training states is not greater than the preset number threshold, it is determined that the idle working space is an available working space.

[0075] In another embodiment, the method for determining the available working space in the idle working space is:

[0076] Based on the training problem data of the idle working space in the similar training state, determine the similar training state with training problems and use it as the problem training state. When the number of the problem training states is greater than the preset state number, it is determined that the idle working space does not belong to the available working space;

[0077] When the number of the problem training states is not greater than the preset state number:

[0078] Based on the training problem types in different problem training states, determine the number of training problem types in different problem training states. When there is a problem training state in which the number of training problem types does not meet the requirements, it is determined that the idle working space does not belong to the available working space;

[0079] When there is no problem training state in which the number of training problem types does not meet the requirements:

[0080] Determine the training problem anomaly coefficient for different problem training states based on the number of training problem types and the occurrence times of different training problem types in different problem training states. When there is a problem training state where the training problem anomaly coefficient does not meet the requirements, it is determined that the idle workspace does not belong to the available workspace;

[0081] When there is no problem training state where the training problem anomaly coefficient does not meet the requirements:

[0082] Determine the basic deviation coefficient based on the number of the problem training states and the proportion of the number of the problem training states in the similar training states. When the basic deviation coefficient does not meet the requirements:

[0083] Obtain the average value of the training problem anomaly coefficients for different problem training states. When the average value of the training problem anomaly coefficients for different problem training states does not meet the requirements, it is determined that the idle workspace does not belong to the available workspace;

[0084] When the basic deviation coefficient meets the requirements or the average value of the training problem anomaly coefficients for different problem training states meets the requirements:

[0085] Determine the anomaly coefficient evaluation quantity based on the basic deviation coefficient and the average value of the training problem anomaly coefficients for different problem training states, and use the anomaly coefficient evaluation quantity to determine whether the idle workspace is an available workspace.

[0086] Further, the training problem types include insufficient GPU video memory space, insufficient memory space, and excessive GPU load rate.

[0087] S2 Based on the available training resources of the available workspace, determine the matching training requirements of the available workspace, and determine the training frequency coefficients and frequent training requirements for different matching training requirements according to the historical training data corresponding to different matching training requirements;

[0088] Specifically, the matching training requirements of the available workspace are based on the training requirements that can be satisfied by the available training resources corresponding to the available workspace.

[0089] Optionally, as Figure 4 shown, the method for determining the training frequency coefficient of the matching training requirements is:

[0090] Determine the historical training times corresponding to the matching training requirements based on the historical training data corresponding to the matching training requirements;

[0091] Based on the preset frequency coefficient corresponding to the historical training times, determine the training frequency coefficient of the matching training requirements.

[0092] Further, the value range of the training frequency coefficient for matching the training requirements is between 0 and 1. When the training frequency coefficient for matching the training requirements is greater than the preset frequency coefficient, it is determined that the training requirements for matching the training requirements are frequent training requirements.

[0093] S3 Determine the number of available workspaces corresponding to different frequent training requirements, and combine the frequent training coefficients of different matching training requirements to determine the training requirement weight coefficients of different frequent training requirements;

[0094] Specifically, the method for determining the training requirement weight coefficient of the frequent training requirement is as follows:

[0095] Based on the number of available workspaces corresponding to the frequent training requirements, determine the number of available spaces for the frequent training requirements, and use the number of available spaces to determine the preset training availability coefficient for the frequent training requirements;

[0096] Based on the frequent training coefficient of the matching training requirements, determine the training requirement weight coefficient of the frequent training requirements based on the ratio of the frequent training coefficient to the preset training availability coefficient.

[0097] Further, the value range of the training requirement weight coefficient of the frequent training requirement is between 0 and 1. The larger the training requirement weight coefficient of the frequent training requirement, the higher the demand degree of the frequent training requirement for the available work area.

[0098] S4 Based on the training requirement weight coefficients of different available workspaces for different frequent training requirements, determine the matching workspaces in the available workspaces and push the matching workspaces to the customer.

[0099] Specifically, the method for determining the matching workspaces in the available workspaces is as follows:

[0100] Based on the sum of the training requirement weight coefficients of the available workspaces for different frequent training requirements, determine the sum of the weight coefficients of the available workspaces;

[0101] Based on the sum of the weight coefficients, determine whether the available workspace is a matching workspace.

[0102] Further, when the sum of the weight coefficients of the available workspace is greater than the preset coefficient threshold, it is determined that the available workspace does not belong to the matching workspace.

[0103] In another embodiment, the method for determining the matching workspaces in the available workspaces is as follows:

[0104] Based on the training requirement weight coefficients of the available workspaces for different frequent training requirements, determine the screening training requirements for the frequent training requirements corresponding to the available workspaces;

[0105] Based on the quantity of the screening training requirements, determine whether the available workspace is a matching workspace.

[0106] Furthermore, the screening training requirements are frequent training requirements with training requirement weight coefficients greater than the preset weight coefficient.

[0107] Specifically, when the quantity of the screening training requirements is greater than the preset training requirement quantity threshold, it is determined that the available workspace is a matching workspace.

[0108] In another embodiment, the method for determining the matching workspaces in the available workspaces is as follows:

[0109] S41 Obtain the quantity of the matching training requirements of the available workspace, and determine the basic usage requirement coefficient of the available workspace by combining the quantities of the available workspaces corresponding to different matching training requirements;

[0110] Optionally, the above step S41 includes the following content:

[0111] S411 Obtain the quantity of the matching training requirements of the available workspace. When the quantity of the matching training requirements of the available workspace is greater than the preset requirement quantity threshold, it is determined that the available workspace is a matching workspace. When the quantity of the matching training requirements of the available workspace is not greater than the preset requirement quantity threshold, proceed to step S412;

[0112] S412 Obtain the quantities of the available workspaces corresponding to different matching training requirements. When the quantities of the available workspaces corresponding to different matching training requirements are all greater than the preset workspace quantity, it is determined that the available workspace does not belong to the matching workspaces. When there are matching training requirements for which the quantity of the available workspace is not greater than the preset workspace quantity, proceed to step S413;

[0113] S413 Determine the basic usage requirement coefficient of the available workspace by combining the quantity of the matching training requirements of the available workspace and the quantities of the available workspaces corresponding to different matching training requirements. When the basic usage requirement coefficient of the available workspace is greater than the preset requirement coefficient threshold, it is determined that the available workspace belongs to the matching workspaces. When the basic usage requirement coefficient of the available workspace is not greater than the preset requirement coefficient threshold, proceed to step S42.

[0114] S42 determines the sum of the training demand weight coefficients of the available working space based on the training demand weight coefficients of the available working space for different frequent training demands, and uses this sum as the weight coefficient sum;

[0115] S43 determines the usage demand coefficient of the available working space based on the basic usage demand coefficient of the available working space and the weight coefficient sum, and determines whether the available working space is a matching working space using the usage demand coefficient.

[0116] Embodiment 2 Second aspect, as Figure 5 shown, the present invention provides a computer system, including: a memory and a processor connected by communication, and a computer program stored on the memory and capable of running on the processor, where when the processor runs the computer program, it executes the above-mentioned method for scheduling computing power of cloud resources.

[0117] Optionally, the above step S42 includes the following content:

[0118] S421 obtains the frequent training demands of the available working space. When the number of frequent training demands of the available working space is greater than the preset frequent demand number, it is determined that the available working space belongs to the matching working space. When the number of frequent training demands of the available working space is not greater than the preset frequent demand number, it proceeds to step S422;

[0119] S422 determines that when there is a frequent training demand with a training demand weight coefficient greater than the preset weight coefficient based on the training demand weight coefficients of the available working space for different frequent training demands, it proceeds to step S423. When there is no frequent training demand with a training demand weight coefficient greater than the preset weight coefficient, it proceeds to step S424;

[0120] S423 uses the frequent training demands with training demand weight coefficients greater than the preset weight coefficient as the screening training demands. When the number of the screening training demands does not meet the requirements, it is determined that the available working space belongs to the matching working space. When the number of the screening training demands meets the requirements, it proceeds to step S424;

[0121] S424 determines the sum of the training demand weight coefficients of the available working space based on the training demand weight coefficients of the available working space for different frequent training demands, and uses this sum as the weight coefficient sum. When the weight coefficient sum is greater than the preset weight coefficient threshold, it is determined that the available working space belongs to the matching working space. When the weight coefficient sum is not greater than the preset weight coefficient threshold, it proceeds to step S43.

[0122] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of devices, apparatuses, and non-volatile computer storage media, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiments.

[0123] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0124] The above is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of one or more embodiments of this specification shall be included within the scope of the claims of this specification.

Claims

1. A computing power scheduling processing method for cloud resources, characterized in that: Specifically include: Based on the training needs of the customer, determine the training needs characteristic quantity of the customer, use the training needs characteristic quantity of the customer and the existing training data of the existing idle workspace as a basis to determine the similar training state in the historical training state of the existing idle workspace, and determine the available workspace in the idle workspace according to the training problem data of different idle workspaces under similar training states; Based on the available training resources of the available workspace, determining the matching training requirements of the available workspace, and determining the training frequency coefficients and frequent training requirements of different matching training requirements according to the historical training data corresponding to the different matching training requirements; Determine the amount of available workspace corresponding to different frequent training needs, and determine the training demand weight coefficients of different frequent training needs in combination with the frequent training coefficients of different matching training needs; Based on the training demand weight coefficients of different available workspaces under different frequent training demands, a matching workspace in the available workspace is determined, and the matching workspace is pushed to the client.

2. The computing power scheduling processing method for cloud resources according to claim 1, characterized in that: The training demand characteristics of the customer include a training model and a training data volume.

3. The computing power scheduling processing method for cloud resources according to claim 1, characterized in that: The existing training data of the idle workspace includes the number of GPUs in use, existing training models, and the amount of training data of different existing training models.

4. The computing power scheduling processing method for cloud resources according to claim 1, characterized in that: The method for determining the similar training state in the historical training state is: Determining the existing training data of the existing idle workspace as a basis for the existing training model of the idle workspace and the amount of training data of different existing training models; Determine a matching training model based on the training model of the customer and the existing training model, and determine a training deviation coefficient with different matching training models based on the deviation rate of the training data volume in different matching training models in the historical training state; A comprehensive deviation coefficient is determined based on an average value of training deviation coefficients of different matching training models, and a similar training state in the historical training state is determined using the comprehensive deviation coefficient.

5. The computing power scheduling processing method for cloud resources according to claim 4, characterized in that: The similar training states in the historical training states are historical training states whose comprehensive deviation coefficients are within a preset deviation coefficient range.

6. The computing power scheduling processing method for cloud resources according to claim 1, characterized in that: The method for determining the available workspace in the idle workspace is: Based on the training problem data of the idle workspace in the similar training state, a similar training state with a training problem is determined and used as the problem training state; Determining whether the idle workspace is an available workspace is based on the number of the problem training states.

7. The computing power scheduling processing method for cloud resources according to claim 6, characterized in that: When the number of the problem training states is not greater than a preset number threshold, the idle workspace is determined to be an available workspace.

8. The computing power scheduling processing method for cloud resources according to claim 1, characterized in that: The matching training requirements of the available workspace are based on the training requirements that can be met by the available training resources corresponding to the available workspace.

9. A computer system comprising: A memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that when the processor runs the computer program, a computing power scheduling and processing method for cloud resources as described in any one of claims 1-8 is executed.

Images