Method and device for providing computing power resources through computing power packages in an intelligent computing center

Abstract

The present invention provides a method and device for providing computing power resources through computing power packages in an intelligent computing center, which relates to the technical fields of intelligent computing centers, intelligent computing centers and computing power infrastructure. The method includes: Step S1, in the case of needing to call computing power resources, determine the computing power package corresponding to the computing power task to be processed, where the computing power task is a task that needs to call computing power resources for processing, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of computation; Step S2, call the computing power resources based on the application program interface corresponding to the computing power package. The application program interface includes the code program corresponding to the operation of executing the computing power task, and the computing power resources called by the application program interface match the operating environment of the computing power task. The present invention can improve the scheduling efficiency of computing power resources, and measure the computing power according to the amount of computation through the computing power package, realizing the wide application of inclusive computing power.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent computing centers, intelligent computing centers, and computing power infrastructure, and particularly relates to a method and device for an intelligent computing center to provide computing power resources through computing power packages. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged as the times require.

[0003] An "intelligent computing center" refers to a facility that uses large-scale heterogeneous computing power resources, including general computing power and intelligent computing power, to mainly provide the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios like artificial intelligence deep learning model development, model training, and model inference). An intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0004] The "intelligent computing center" includes but is not limited to the "intelligent computing center".

[0005] An "intelligent computing center", that is, an artificial intelligence computing center, is a type of computing power infrastructure based on artificial intelligence theory, adopting an artificial intelligence computing architecture, and providing computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "intelligent computing centers". It is the ability of computer devices or computing / data centers to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to achieve the output of target results through processing information data, and a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] Currently, it is possible to directly provide computing power resources that meet user needs through cloud servers. Before a cloud server executes a computing power task for model development, testers need to build an operation framework according to the specific parameters of the computing power task and set corresponding operation environment parameters, so that the cloud server can execute the computing power task based on the operation framework and operation environment parameters. However, the process of building an operation framework and setting corresponding operation environment parameters is relatively complex and takes a long time, resulting in low efficiency of the intelligent computing center in scheduling computing power resources to process computing power tasks. At the same time, since the GPU resources of the intelligent computing center need to be occupied for a long time during model development, business personnel need to lease the computing power services of the intelligent computing center for a long time, resulting in high costs for model development and making it difficult to achieve the widespread application of inclusive computing power.

[0008] It can be seen that in the prior art, there are problems of low efficiency in scheduling computing power resources to process computing power tasks in an intelligent computing center and high model development costs. Summary of the Invention

[0009] The present invention provides a method and device for an intelligent computing center to provide computing power resources through computing power packages, so as to solve the problems of low efficiency in scheduling and using computing power for computing power tasks and high model development costs in the prior art.

[0010] To solve the above problems, the present invention is implemented as follows:

[0011] In a first aspect, the present invention provides a method for an intelligent computing center to provide computing power resources through computing power packages, including:

[0012] Step S1, in the case of needing to call computing power resources, determine the computing power package corresponding to the computing power task to be processed, where the computing power task is a task that needs to call computing power resources to process, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of calculation;

[0013] Step S2, call the computing power resources based on the application programming interface corresponding to the computing power package, where the application programming interface includes a code program corresponding to the operation of executing the computing power task, and the computing power resources called by the application programming interface match the operating environment of the computing power task.

[0014] In one embodiment, the computing power package corresponds to multiple application programming interfaces, and step S2 includes:

[0015] Step S21, determine the target task type corresponding to the computing power task;

[0016] Step S22, query the target application programming interface corresponding to the target task type based on a preset mapping table, where the preset mapping table includes multiple task types and the multiple application programming interfaces, as well as the mapping relationship between different task types and different application programming interfaces, and the target application programming interface is one of the multiple application programming interfaces;

[0017] Step S23, call the computing power resources based on the target application programming interface.

[0018] In one embodiment, the multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface.

[0019] In one embodiment, after step S2, the method further includes:

[0020] Step S3, process the computing power task based on the computing power resources called by the target application programming interface;

[0021] Among them, the computing power task is the computing power task processed by the target container, and the step S3 includes at least one of the following:

[0022] Step S31: When the load data is greater than or equal to the first preset threshold, scale out the target container based on the target application programming interface, where the load data is the load data of the target container;

[0023] Step S32: When the load data is less than or equal to the second preset threshold, scale in the target container based on the target application programming interface.

[0024] In one embodiment, the step S31 includes:

[0025] Step S311: When the load data is greater than or equal to the first preset threshold, obtain the first computing power resource information of the first server, where the first server is the server to which the target container belongs;

[0026] Step S312: When the first computing power resource information meets the load data, create a first container based on the computing power resources of the first server through the target application programming interface;

[0027] Step S313: Deploy the relevant data of the computing power task to the first container through the target application programming interface;

[0028] Step S314: Execute the computing power task based on the first container.

[0029] In one embodiment, the step S31 further includes:

[0030] Step S315: When the first computing power resource information does not meet the load data, obtain the network information and the second computing power resource information of the second server, where the second server is in the same region or a different region from the first server;

[0031] Step S316: When the second computing power resource information meets the load data and the network information meets the preset network conditions, create a second container based on the computing power resources of the second server;

[0032] Step S317: Deploy the relevant data of the computing power task to the second container;

[0033] Step S318: Execute the computing power task based on the second container.

[0034] In one embodiment, after the step S3, the method further includes:

[0035] Step S4: Measure the computing power resource consumption of the computing power package based on the target container, the first container, or the second container, where the target container, the first container, and the second container are the containers for executing the computing power task.

[0036] In a second aspect, the present invention further provides a device for providing computing power resources through computing power packages in an intelligent computing center, including:

[0037] A determination module, configured to determine a computing power package corresponding to a computing power task to be processed when computing power resources need to be invoked. The computing power task is a task that requires invoking computing power resources for processing, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of computation.

[0038] An invocation module, configured to invoke computing power resources based on the application programming interface corresponding to the computing power package. The application programming interface includes a code program corresponding to the operation of executing the computing power task, and the computing power resources invoked by the application programming interface match the operating environment of the computing power task.

[0039] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in the method for providing computing power resources through computing power packages in the intelligent computing center as described in the first aspect above.

[0040] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of providing computing power resources through computing power packages in the intelligent computing center as described in the first aspect above.

[0041] In a fifth aspect, the present invention further provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the steps in the method for providing computing power resources through computing power packages in the intelligent computing center as described in the first aspect above.

[0042] In the present invention, in the case where computing power resources need to be invoked, a computing power package corresponding to the computing power task to be processed is determined. The computing power task is a task that requires the invocation of computing power resources for processing, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of computation. The computing power resources are invoked based on the application programming interface corresponding to the computing power package. The application programming interface includes a code program corresponding to the operation of executing the computing power task, and the computing power resources invoked by the application programming interface match the operating environment of the computing power task. In this way, when a computing power task needs to be executed, the computing power resources of the intelligent computing center can be quickly scheduled through the application programming interface of the intelligent computing center to process the computing power task, without the user having to configure an operating framework and set operating environment parameters for each computing power task, effectively improving the scheduling efficiency of the computing power resources, and thus effectively improving the execution efficiency of the computing power task. At the same time, by measuring the computing power according to the amount of computation using the computing power package, the problem of high costs caused by business personnel leasing the computing power services of the intelligent computing center for a long time is avoided, and the wide application of inclusive computing power is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] To more clearly illustrate the technical solutions of the present invention, the drawings required for the description of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0044] Figure 1 is a flowchart of a method for an intelligent computing center to provide computing power resources through a computing power package provided by the present invention;

[0045] Figure 2 is a schematic structural diagram of an intelligent computing center provided by the present invention;

[0046] Figure 3 is a structural diagram of a device for an intelligent computing center to provide computing power resources through a computing power package provided by the present invention;

[0047] Figure 4 is a structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The technical solutions in the present invention will be clearly and completely described below with reference to the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0049] The "computing power" described in the present invention refers to: the ability of a computer device or a computing / data center to process information, which is the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, and is the computing ability to process information data and achieve the output of the target result. It is a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through the computing power infrastructure.

[0050] The "computational power" (Computational Power, CP) described in the present invention refers to: the ability of the data center server to process data and achieve the result output, which is a comprehensive index to measure the computing ability of the data center, including general computing ability, supercomputing ability, and intelligent computing ability. The commonly used measurement unit is the number of floating-point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS), and the larger the value, the stronger the comprehensive computing ability. It is estimated that 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream laptops. The calculation formula is: CP = CP 通用 + CP 智能 + CP 超级 .

[0051] The "carrying capacity" (Network Power, NP) described in the present invention refers to: the performance of the data transmission ability of the computing power facility, which is a comprehensive ability including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., and involves network transmission inside and between data centers. It is a comprehensive index to measure the network transmission and scheduling ability.

[0052] The "storage power" (Storage Power, SP) described in the present invention refers to: the comprehensive ability of the data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon. It is a comprehensive index to measure the data storage ability of the data center, including external storage devices such as storage arrays and server internal storage devices. The commonly used measurement unit for storage capacity is exabyte (EB, 1 EB = 2^60 bytes), and the commonly used measurement unit for performance is the number of read and write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB). The disaster recovery ratio is an important manifestation of security and reliability.

[0053] The "computing power infrastructure" described in the present invention refers to: a new type of information infrastructure integrating information computing power, network carrying capacity, and data storage capacity, which can realize the centralized computing, storage, transmission, and application of information, presenting characteristics such as multi-element ubiquitous, intelligent and agile, secure and reliable, green and low-carbon, and is of great significance for boosting industrial transformation and upgrading, empowering China's scientific and technological innovation, meeting people's beautiful life, and realizing high-efficiency social governance.

[0054] The "new information infrastructure" described in the present invention refers to: mainly including network infrastructures such as 5G networks, fiber broadband networks, backbone networks, international communication networks, satellite Internet, etc., computing power infrastructures such as data centers, general computing power centers, intelligent computing centers, supercomputing centers, etc., and new technology facilities such as artificial intelligence, blockchain, and quantum computing. With the emergence and popularization of new general technologies, the form of the new information infrastructure will be more diverse.

[0055] The "computing power" described in the present invention includes: general computing power, intelligent computing power, and super computing power.

[0056] The "general computing power" described in the present invention refers to: the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0057] The "intelligent computing power" described in the present invention refers to: for various artificial intelligence innovation applications, a computing platform deployed on a large scale based on dedicated chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit), such as natural language processing, machine vision, etc.

[0058] The "super computing power" described in the present invention refers to: mainly the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and processes extremely complex or data-intensive problems through a dedicated operating system. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, gene analysis, etc.

[0059] The "intelligent computing center" described in the present invention refers to: a facility that provides the required computing power, data, and algorithms mainly for artificial intelligence applications (such as scenarios of artificial intelligence deep learning model development, model training, and model inference) by using large-scale heterogeneous computing power resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0060] The "intelligent computing center" described in the present invention includes but is not limited to the "intelligent computing center".

[0061] The "Intelligent Computing Center" described in the present invention, namely the artificial intelligence computing center, is a type of computing infrastructure based on artificial intelligence theory, adopting an artificial intelligence computing architecture, and providing computing power services, data services, and algorithm services required for artificial intelligence applications.

[0062] The "Computing Power Center" described in the present invention refers to: facilities mainly composed of infrastructure such as wind, fire, water, and electricity and IT software and hardware devices, with computing power, transportation power, and storage power, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0063] The "Supercomputing Center" described in the present invention refers to: namely the supercomputing data center, which is a data center based on supercomputers or large-scale computing clusters, capable of providing functions such as large-scale computing, storage, and network services, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling, and genome sequencing.

[0064] The "Computing Power Resources" described in the present invention refer to: technologies and facilities with information computing, transmission, storage, and application capabilities required for the development of the digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and support and guarantee resources such as wind, fire, water, and electricity.

[0065] The "Inclusive Computing Power" described in the present invention refers to providing appropriate and effective computing power services to all social strata and groups with computing power service needs at an affordable cost based on the requirements of equal opportunity and the principle of commercial sustainability.

[0066] The "Model" described in the present invention includes but is not limited to "Large Language Model" and "Multimodal Large Model".

[0067] The "Large Language Model" described in the present invention refers to a large-scale language model (LLM), which is a language model with a relatively large number of parameters, aiming to understand and generate human language, trained through a large amount of text data, and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0068] The "Multimodal Large Models" described in the present invention refer to: models that jointly train multimodal information such as text, images, videos, and audio, including but not limited to multimodal large language models.

[0069] The "Agent" in the present invention refers to an agent that can perceive the environment and take actions to achieve specific goals. It can be software, hardware, or a system, and has the capabilities of autonomy, adaptability, and interaction. The agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions based on the knowledge and algorithms it has learned, and then executes actions to affect the environment or achieve the predetermined goals. Agents are widely used in the field of artificial intelligence, commonly found in automated systems, robots, virtual assistants, and game characters, etc. The core lies in the ability to learn autonomously and evolve continuously to better complete tasks and adapt to complex environments.

[0070] The "computing power package" in the present invention refers to the following three aspects simultaneously:

[0071] 1. A service that packages computing power resources with "degree" as the basic measurement unit and provides computing power to users. Users can purchase, use, or lease the computing power of the intelligent computing center through the "computing power providing service". Among them, the measurement of computing power resources with "degree" as the basic measurement unit can refer to the patent document CN 118760575B of the applicant: "1 degree of computing power" is 1 unit of computing power, specifically X TFLOPS·h, that is, X trillion floating-point operation times per second × hour. Among them, "X TFLOPS·h" can be but is not limited to: "1TFLOPS·h", "312 TFLOPS·h", "624TFLOPS·h", "1024 TFLOPS·h", "10000 TFLOPS·h", etc. Any number. The numbers are only illustrative, not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms of extensions without departing from the purpose of the present invention and the scope protected by the claims, all of which belong to the protection scope of the present invention.

[0072] 2. It is used to measure the computing power of the intelligent computing center according to the amount of calculation, realize charging according to the amount of calculation, and only charge when the computing task is running. It can be purchased on demand and used immediately upon purchase.

[0073] 3. It realizes flexible scheduling of GPU resources, elastically expands or contracts computing resources as needed, greatly reducing the waste of computing power resources; realizes out-of-the-box "computing power + algorithm" integrated artificial intelligence services with more economy and high performance, and provides a complete large model + Agent development tool chain; ultimately realizes the wide application of "inclusive computing power".

[0074] Specifically, the "computing power package" in the present invention simultaneously includes the following three functional service components:

[0075] 1. Application Programming Interface (API). The API includes a code program for performing computing power task operations. That is, the API defines a set of application programming interfaces with complete functions, clear interfaces, and easy-to-use features. Through the API, the computing power resources corresponding to the computing power tasks can be quickly called. Among them, the API can automatically match the computing power resources that meet the requirements by calling conditions such as computing resource requirements, machine resource occupancy, GPU type, network interconnection situation, and storage usage, and efficiently complete the user's computing requirements.

[0076] 2. Scaling Service Component. The scaling service component is used to scale the containers executing the computing power tasks, so that the computing power resources occupied by the containers are maintained at a reasonable level. While ensuring the smooth execution of the computing power tasks, less computing power resources of the intelligent computing center are occupied, enabling the intelligent computing center to provide computing power services for more users and realizing the wide application of inclusive computing power.

[0077] 3. Metering Component. The metering component is used to meter the computing power resources consumed during the execution of the computing power tasks. Among them, during the process of metering the consumption of computing power resources, only the computing power resources consumed by the containers when executing the computing power tasks are metered, and the computing power resources consumed during the scaling of the containers are not metered, which improves the accuracy of metering, reduces the computing power resources that users need to lease, and reduces the user's usage cost. Furthermore, the wide application of inclusive computing power is realized.

[0078] The "computing power package" described in the present invention can also be referred to as "computing power set", "computing power library", "computing power group", etc. The name is only illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms of expansions without departing from the purpose of the present invention and the scope protected by the claims, and all belong to the protection scope of the present invention.

[0079] In the prior art, when users develop software or models, they need to use a large amount of computing power resources. In order to reduce the cost of using computing power resources, users usually choose to lease computing power services from cloud servers and use the cloud servers to provide computing power to run software or models when software or model development is needed, so as to reduce the cost of directly purchasing hardware devices.

[0080] Please refer to Figure 1 , Figure 1 which is a flowchart of a method for an intelligent computing center to provide computing power resources through a computing power package provided by the present invention. As Figure 1 shown, it includes the following steps:

[0081] Step S1. In the case of needing to call computing power resources, determine the computing power package corresponding to the computing power task to be processed. The computing power task is a task that needs to call computing power resources to process, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of calculation.

[0082] The above computing power tasks are tasks that users need the intelligent computing center to execute. The computing power tasks can be tasks such as training a model, using the intelligent agents in the intelligent computing center to perform inference on data, and fine-tuning. It should be noted that the computing power tasks are computing power tasks created by users based on the computing power packages in the intelligent computing center, that is, each computing power task corresponds to a computing power package, and the computing power tasks are created through the algorithm framework, data set, and / or preset model provided by the computing power package. The computing power packages in the intelligent computing center can call computing power resources to process the computing power tasks.

[0083] The above intelligent computing center is used to provide computing power resources to execute computing power tasks. Among them, the computing power resources are encapsulated in the form of computing power packages, and the consumed computing power is measured according to the specific amount of computation consumed during the execution of the computing power tasks.

[0084] It should be noted that in the prior art, computing power is mainly provided based on the rental duration, and users determine the rental duration of the computing power service according to the duration of software development required. Specifically, when the software or model is running, computing power resources are consumed, and when the software or model is adjusted, it is not necessary to consume computing power resources, resulting in underutilization of computing power resources and high costs for users to use the computing power service.

[0085] In the present invention, a computing power server is provided based on the amount of computation through a computing power package, which can more fully improve the utilization rate of computing power compared with the method of providing computing power based on the rental duration, and at the same time reduce the usage cost of users, realizing the wide application of inclusive computing power.

[0086] Step S2: Call computing power resources based on the application programming interface corresponding to the computing power package. The application programming interface includes the code program corresponding to the operation of executing the computing power task, and the computing power resources called by the application programming interface match the operating environment of the computing power task.

[0087] The above application programming interface encapsulates the computing power operations of different computing power tasks. Through the application programming interface, computing power resources matching the computing power tasks can be quickly scheduled, and the computing power tasks can be executed based on the computing power resources. It is not necessary for users to configure a running framework and set running environment parameters for each computing power task, which can effectively improve the efficiency of computing power resource scheduling and thus improve the execution efficiency of computing power tasks.

[0088] It should be noted that the application programming interface is the application programming interface corresponding to the computing power package. When the user chooses to purchase, use, or lease computing power resources in the form of a computing power package (that is, measure computing power resources by the amount of computation), the application programming interface needs to be called to process the computing power tasks; when the user does not choose to purchase, use, or lease computing power resources in the form of a computing power package (that is, does not measure computing power resources by the amount of computation), it is not necessary to call the application programming interface to process the computing power tasks.

[0089] Specifically, the application programming interface includes the code program corresponding to the operation of the computing power task, that is, the application programming interface defines a set of application programming interfaces with complete functions, clear interfaces, and easy to use. Through the application programming interface, the computing power resources corresponding to the computing power task can be quickly called. Among them, the application programming interface can automatically match the computing power resources that meet the requirements by calling conditions such as computing resource requirements, machine resource occupancy, GPU type, network interconnection situation, and storage usage, and efficiently complete the user's computing requirements.

[0090] As Figure 2 shown, the intelligent computing center includes a user layer, an application programming interface layer of computing power packages, infrastructure software, and infrastructure hardware. Among them, the user layer is used to provide the user with a large model tool chain, and the user can directly use various tools included in the large model tool chain. Among them, the large model tool chain includes a large model framework, a data set, and various preset models. For example, the large model framework can be PyTorch, vLLM, llama.cpp, etc.; the application programming interface layer of the computing power package is encapsulated with different functions to adapt to different computing power tasks, such as functions of model training, inference, and fine-tuning; the infrastructure software includes various software platforms to provide basic software services, such as the Alaya NeW platform, DingoStack, DingoFS, etc.; the infrastructure hardware includes the servers, GPU acceleration cards, storage systems, and high-speed network hardware of the intelligent computing center.

[0091] It can be understood that by configuring the application programming interface for the intelligent computing center in the above way, the computing power resources of the intelligent computing center can be quickly scheduled through the application programming interface to process the computing power task. It is not necessary for the user to configure the operation framework and set the operation environment parameters for each computing power task, which can effectively improve the efficiency of computing power resource scheduling and then improve the execution efficiency of the computing power task.

[0092] In the present invention, in the case where computing power resources need to be invoked, a computing power package corresponding to the computing power task to be processed is determined. The computing power task is a task that requires the invocation of computing power resources for processing, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of computation. Based on the application programming interface corresponding to the computing power package, the computing power resources are invoked. The application programming interface includes a code program corresponding to the operation of executing the computing power task, and the computing power resources invoked by the application programming interface match the operating environment of the computing power task. In this way, when a computing power task needs to be executed, the computing power resources of the intelligent computing center can be quickly scheduled through the application programming interface of the intelligent computing center to process the computing power task, without the user needing to configure an operating framework and set operating environment parameters for each computing power task, effectively improving the scheduling efficiency of the computing power resources, and thus effectively improving the execution efficiency of the computing power task. At the same time, by measuring the computing power according to the amount of computation through the computing power package, the problem of high costs caused by business personnel leasing the computing power services of the intelligent computing center for a long time can be avoided, and the wide application of inclusive computing power is realized.

[0093] In one embodiment, the computing power package corresponds to multiple application programming interfaces, and step S2 includes:

[0094] Step S21, determining the target task type corresponding to the computing power task;

[0095] Step S22, querying the target application programming interface corresponding to the target task type based on a preset mapping table. The preset mapping table includes multiple task types and the multiple application programming interfaces, as well as the mapping relationship between different task types and different application programming interfaces. The target application programming interface is one of the multiple application programming interfaces;

[0096] Step S23, invoking the computing power resources based on the target application programming interface.

[0097] The above-mentioned computing power package corresponds to multiple application programming interfaces. In the case where the computing power package of the intelligent computing center needs to provide computing power resources for different computing power tasks, the computing power resources can be scheduled for each computing power task through different application programming interfaces, thereby realizing the processing of different computing power tasks.

[0098] It should be noted that the specific content of different computing power tasks is different. For example, the operations performed when the computing power task is a model training task are different from those performed when the computing power task is a data fine-tuning task. If the same application programming interface is used to adapt different types of computing power tasks, the application programming interface needs to set more code to improve compatibility, occupying more resources of the intelligent computing center, and it is difficult to set more application programming interfaces, restricting the number of users of the computing power services that the intelligent computing center can provide.

[0099] Therefore, in this embodiment, different application programming interfaces are set according to the specific types of computing power tasks, so that each application programming interface can provide a computing power resource service for quickly scheduling the intelligent computing center while occupying less resources, thereby increasing the number of users who can use the computing power service provided by the intelligent computing center.

[0100] Specifically, first divide the computing power tasks into different task types according to the specific content of the computing power tasks; create different application programming interfaces, each of which is created for a computing power task of a task type, and the computing power resources of the intelligent computing center can be scheduled through the application programming interface to execute the computing power task; establish a mapping relationship between each application programming interface and different task types, and create a preset mapping table based on the mapping relationship. After creating the preset mapping table, the target application programming interface can be determined according to the task type of the computing power task.

[0101] The above determination of the target task type corresponding to the computing power task (step S21) can specifically be that the user sets the task type corresponding to the computing power task when uploading the computing power task; or, identify the specific content of the computing power task to determine the task type corresponding to the computing power task.

[0102] In the present invention, determine the target task type corresponding to the computing power task; query the target application programming interface corresponding to the target task type based on the preset mapping table, the preset mapping table includes multiple task types and the multiple application programming interfaces, as well as the mapping relationship between different task types and different application programming interfaces, and the target application programming interface is one of the multiple application programming interfaces; call the computing power resources based on the target application programming interface. In this way, the target application programming interface is determined through the target task type of the computing power task, and then the scheduling efficiency of the computing power resources is effectively improved through the target application programming interface, and further the execution efficiency of the computing power task.

[0103] In one embodiment, the multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface.

[0104] In the present invention, the multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface. In this way, at least one application programming interface is set for different task types, so that the intelligent computing center can provide computing power resources for computing power tasks of different task types at the same time, and provide computing power resources for multiple computing power tasks of the same task type.

[0105] It should be noted that in the field of intelligent computing centers, an intelligent computing center can provide a large amount of computing power resources to execute computing power tasks. When the computing power of the intelligent computing center is insufficient, no computing power resource service will be provided to new users. For users who have already been provided with computing power services, there are changes in the consumption of computing power resources. For example, more computing power resources are consumed during the running of a program, while less computing power resources are consumed when the program is adjusted, resulting in a situation where there is redundant computing power resources in the intelligent computing center but it cannot provide computing power services to more users.

[0106] To enable more users to obtain the computing power services of the intelligent computing center, in the present invention, the computing power provided by the intelligent computing center is elastically scaled, so that the intelligent computing center can provide computing power resources to more users.

[0107] Specifically, in one embodiment, after the step S2, the method further includes:

[0108] Step S3: Process the computing power task based on the computing power resources called by the target application programming interface;

[0109] Wherein, the computing power task is the computing power task processed by the target container, and the step S3 includes at least one of the following:

[0110] Step S31: When the load data is greater than or equal to the first preset threshold, expand the target container based on the target application programming interface, where the load data is the load data of the target container;

[0111] Step S32: When the load data is less than or equal to the second preset threshold, scale down the target container based on the target application programming interface.

[0112] The above first preset threshold is used to measure whether the load data of the target container is too large. When the load data is greater than or equal to the first preset threshold, the load of the target container is large. To meet the execution of the computing power task in the target container, it is necessary to further expand the target container to ensure that the computing power task is executed with sufficient computing power resources.

[0113] The above second preset threshold is used to measure whether the load data of the target container is too small. When the load data is less than or equal to the second preset threshold, the load of the target container is small, and there is a situation of excess computing power resources. To improve the utilization rate of the computing power resources of the intelligent computing center and reduce the waste of computing power resources, it is necessary to scale down the target container to release more free computing power, so that the intelligent computing center can provide computing power services to more users, thereby further realizing the wide application of inclusive computing power.

[0114] Among them, the first preset threshold is greater than the second preset threshold. When the load data of the target container is between the first preset threshold and the second preset threshold, the computing power resources occupied by the target container are at a relatively stable level, and there is no need to scale the target container up or down, avoiding the situation where the target container needs to be frequently scaled up or down due to setting the first preset threshold and the second preset threshold to be the same, making the computing power resources of the intelligent computing center in a relatively stable state.

[0115] Furthermore, when it is necessary to scale the target container up or down, it can be directly implemented through the target application programming interface corresponding to the computing power package, realizing the automatic scheduling of computing power resources without the need for users to set or adjust.

[0116] In the present invention, when the load data is greater than or equal to the first preset threshold, the target container is expanded based on the target application programming interface, where the load data is the load data of the target container; when the load data is less than or equal to the second preset threshold, the target container is scaled down based on the target application programming interface. In this way, the target container is scaled up or down according to the load data of the target container, enabling the target container to occupy less computing power resources of the intelligent computing center while ensuring sufficient computing power resources to execute computing power tasks, so that the intelligent computing center can provide computing power services for more users, further realizing the wide application of inclusive computing power.

[0117] In some embodiments, the computing power resources of the intelligent computing center include local computing power resources and off-site computing power resources. The methods for the intelligent computing center to schedule computing power resources in different regions for scaling up or down are also different, and parameters such as the network and the remaining computing power of different servers need to be considered to make the delay provided by the intelligent computing center lower and the computing power resources sufficient.

[0118] Among them, first determine whether the local computing power resources are sufficient, and then schedule off-site computing power resources when the local resources are insufficient.

[0119] Specifically, in one embodiment, step S31 includes:

[0120] Step S311, when the load data is greater than or equal to the first preset threshold, obtain the first computing power resource information of the first server, where the first server is the server to which the target container belongs;

[0121] Step S312, when the first computing power resource information meets the load data, create a first container based on the computing power resources of the first server through the target application programming interface;

[0122] Step S313, deploy the relevant data of the computing power task to the first container through the target application programming interface;

[0123] Step S314, execute the computing power task based on the first container.

[0124] The above-mentioned first server is the server to which the target container belongs. That is, in the case where the target container needs to be expanded, the local computing power resources are preferentially considered for scheduling. When the local computing power resources meet the scheduling requirements, the local computing power resources are directly scheduled; when the local computing power resources do not meet the scheduling requirements, the computing power resources in different locations are then scheduled.

[0125] Specifically, in the case of needing to expand, obtain the first computing power resource information of the first server, and the first computing power resource information is used to characterize the remaining computing power resources of the first server. When the first computing power resource information meets the load data, it is considered that the first server meets the expansion requirements of the target container, and at this time, the computing power resources of the first server are directly scheduled for expansion.

[0126] Among them, the first computing power resource information meets the load data, which may be that the first computing power resource information includes the remaining computing power of the first server, the load data includes the required computing power, and when the remaining computing power is greater than the required computing power, the first computing power resource information meets the load data.

[0127] The relevant data of the above-mentioned computing power task specifically includes at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task. Deploying the relevant data of the computing power task to the first container specifically means deploying at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task into the first container, so that the first container can process the computing power task according to the deployed relevant data.

[0128] In the present invention, when the load data is greater than or equal to the first preset threshold, obtain the first computing power resource information of the first server, and the first server is the server to which the target container belongs; when the first computing power resource information meets the load data, create a first container based on the computing power resources of the first server through the target application program interface; deploy the relevant data of the computing power task to the first container through the target application program interface; execute the computing power task based on the first container. In this way, the expansion of the target container is realized through the computing power resources of the first server, the corresponding computing service can be quickly started, and the computing process can be diverted to a new machine, thereby realizing the efficient execution of the computing power task.

[0129] In the case where the local computing power resources are insufficient, the intelligent computing center needs to schedule the computing power resources in different locations to process the computing power task. Specifically, in one embodiment, the step S31 further includes:

[0130] Step S315: When the first computing power resource information does not meet the load data, obtain the network information and the second computing power resource information of the second server, where the second server is in the same area or a different area from the first server;

[0131] Step S316: When the second computing power resource information meets the load data and the network information meets the preset network conditions, create a second container based on the computing power resource of the second server;

[0132] Step S317: Deploy the relevant data of the computing power task to the second container;

[0133] Step S318: Execute the computing power task based on the second container.

[0134] The above-mentioned second server is a server other than the first server. When the computing power resources of the first server are insufficient, the intelligent computing center needs to schedule computing power resources through a second server in a different location to process the computing power task.

[0135] In some embodiments, the second server may be a server in the same area as the first server. It should be noted that the second server and the first server being in the same area may mean that the hardware devices of the second server and the first server are in the same or adjacent computer rooms, that is, the physical distance between the second server and the first server is relatively close. In this case, the latency between the second server and the first server is relatively low, and the computing power resources of the second server can be preferentially called to expand the target container.

[0136] In some embodiments, the second server may be a server in a different area from the first server. It should be noted that the second server and the first server being in different areas may mean that the hardware devices of the second server and the first server are in computer rooms with a relatively long spatial distance. For example, the hardware device of the first server is located within City A, and the hardware device of the second server is located within City B. At this time, the latency between the second server and the first server is relatively high, and it can be used as a backup method for remotely invoking computing power resources.

[0137] Specifically, step S315 is specifically as follows:

[0138] When the first computing power resource information does not meet the load data, obtain the computing power resource information and network information of the second server in the same area;

[0139] When the computing power resource information of the second server in the same region does not meet the load data, or when the network information of the second server in the same region meets the preset network conditions, obtain the computing power resource information and network information of the second server in different regions;

[0140] The specific step S316 is as follows:

[0141] When the computing power resource information of the second server in the same region meets the load data and the network information of the second server in the same region meets the preset network conditions, create the second container based on the computing power resources of the second server in the same region;

[0142] When the computing power resource information of the second server in different regions meets the load data and the network information of the second server in different regions meets the preset network conditions, create the second container based on the computing power resources of the second server in different regions.

[0143] The above preset network conditions are used to measure whether the network information of the second server meets the requirement of scheduling computing power resources with low latency. Among them, the preset network conditions may include at least one of a latency threshold, a bandwidth threshold, and a connectivity parameter threshold, and the network situation of the second server is determined by at least one of latency, bandwidth, and connectivity parameters.

[0144] In some embodiments, when the second server and the first server are in the same region, obtain the network information of all servers in the first region, where the first region is the region where the first server is located;

[0145] When the network information of the second server is the best among the network information of all servers, the second server meets the preset network conditions.

[0146] In some embodiments, when the second server and the first server are in different regions, obtain at least one of the latency, bandwidth, and connectivity parameters of the second server;

[0147] When the latency of the second server meets the latency threshold, and / or the bandwidth of the second server meets the bandwidth threshold, and / or the connectivity of the second server meets the connectivity threshold, the second server meets the preset network conditions.

[0148] Deploying the relevant data of the computing power task to the second container (step S313) mentioned above, when the latency is low (the second server and the first server are in the same region), the relevant data of the computing power task can be synchronously deployed to the second container, while when the latency is high (the second server and the first server are in different regions), the relevant data of the computing power task needs to be asynchronously deployed to the second container.

[0149] Specifically, when the second server and the first server are in the same region, at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task is synchronously copied to the storage hardware of the second server, then a computing power device is selected to create a second container, and at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task is deployed to the second container, so that the second container can deploy relevant data to process the computing power task.

[0150] While when the second server and the first server are in different regions, at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task is asynchronously copied to the storage hardware of the second server, then a computing power device is selected to create a second container, and at least one of the environment parameters, data to be processed, model, operation parameters, and application program of the computing power task is deployed to the second container, so that the second container can deploy relevant data to process the computing power task.

[0151] In the present invention, when the first computing power resource information does not meet the load data, the network information and the second computing power resource information of the second server are obtained, and the second server and the first server are in the same region or different regions; when the second computing power resource information meets the load data and the network information meets the preset network conditions, a second container is created based on the computing power resource of the second server through the target application programming interface; the relevant data of the computing power task is deployed to the second container through the target application programming interface; the computing power task is executed based on the second container. In this way, when the local computing power resources are insufficient, the intelligent computing center can schedule the computing power resources of the second server to process the computing power task.

[0152] Furthermore, when the load data of the target container is low, it is necessary to scale down the target container to release the computing power resources of the intelligent computing center, so that the intelligent computing center can provide more computing power resources to users.

[0153] It should be noted that different from scaling up the target container, when the target container needs to be scaled down, it does not need to consider other servers, and only the computing power resources of the first server where the target container is located need to be released.

[0154] In some embodiments, when the load data is less than or equal to a second preset threshold, scaling down the target container based on the target application programming interface may be to release the computing power resources of the target container according to a preset ratio.

[0155] In some embodiments, when the load data is less than or equal to a second preset threshold, scaling down the target container based on the target application programming interface may be to release the computing power resources of the target container according to the value of the preset computing power resources.

[0156] In one embodiment, after the step S3, the method further includes:

[0157] Step S4, measuring the consumption of the computing power resources of the computing power package based on the target container, the first container or the second container, where the target container, the first container and the second container are the containers for executing the computing power task.

[0158] It should be noted that during the process of executing the computing power task in the intelligent computing center, the situation of scaling the target container up or down may occur, and the computing power resources of the intelligent computing center will be consumed during the scaling process. For example, deploying the relevant data of the computing power task to the first container requires the use of the computing power resources of the intelligent computing center. However, since this process is not an execution of the computing power task, when measuring the computing power consumption, only the computing power resources consumed by the target container, the first container or the second container for executing the computing power task are measured to improve the accuracy of the measurement.

[0159] In the present invention, measuring the consumption of the computing power resources of the computing power package based on the target container, the first container or the second container, where the target container, the first container and the second container are the containers for executing the computing power task, can improve the measurement accuracy of the computing power resources, thereby reducing the user's usage cost and further realizing the wide application of inclusive computing power.

[0160] Specifically, the measuring the consumption of the computing power resources of the computing power package based on the target container, the first container or the second container includes:

[0161] Obtaining the target ratio of the processor resources occupied by the target container, the first container or the second container, and the total time of occupying the processor, where the processor is the processor of the intelligent computing center;

[0162] Obtaining the computing power value of the processor, where the computing power value is used to represent the computing power resources provided by the processor per unit time;

[0163] Calculating the consumption of the computing power resources based on the target ratio, the total time and the computing power value.

[0164] Exemplarily, the computing power value of the processor is 3 degrees / h, and the computing power task is executed by the target container for a total of 2 hours. During the execution of the computing power task, the target proportion of the processor resources occupied is 1 / 3. In this case, directly multiplying the computing power value of the processor, the total time of occupying the processor, and the target proportion can obtain the computing power value of the computing power package consumed by the execution of the computing power task, which is 2 degrees. Here, "degree" is the unit of computing power resources.

[0165] Further, in the case where the computing power task is executed by multiple containers, it is necessary to sum the computing power resources consumed by different containers to obtain the total computing power resource consumption of the execution of the computing power task.

[0166] In some embodiments, the obtaining the target proportion of the target container, the first container, or the second container occupying the processor resources includes:

[0167] Obtaining at least one of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity of the processor occupied by the target container, the first container, or the second container;

[0168] Calculating the target proportion based on at least one of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity.

[0169] In the present invention, calculating the target proportion through at least one of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity improves the accuracy of the target proportion, thereby improving the accuracy of the calculated computing power consumption, reducing the user's usage cost, and realizing the wide application of inclusive computing power.

[0170] In some embodiments, the calculating the target proportion based on at least one of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity includes:

[0171] Obtaining at least one of the preset half-precision floating-point computing power number, the preset video memory bandwidth, and the preset video memory capacity of the processor;

[0172] Calculating a first ratio of the first half-precision floating-point computing power number to the preset half-precision floating-point computing power number, and / or calculating a second ratio of the first video memory bandwidth to the preset video memory bandwidth, and / or calculating a third ratio of the first video memory capacity to the preset video memory capacity;

[0173] Calculating the target proportion based on the first ratio, the second ratio, and the third ratio.

[0174] Wherein, in the case of only obtaining one of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity, the corresponding first ratio, second ratio, or third ratio is directly set as the target proportion.

[0175] When obtaining at least two of the first half-precision floating-point computing power number, the first video memory bandwidth, and the first video memory capacity, at least two of the corresponding first ratio, second ratio, and third ratio are weighted to obtain the target ratio.

[0176] Please refer to Figure 3 , Figure 3 FIG. is a structural diagram of a device for providing computing power resources through computing power packages in an intelligent computing center provided by the present invention. As Figure 3 shown, the device 300 for providing computing power resources through computing power packages in the intelligent computing center includes:

[0177] A determination module 301, configured to determine a computing power package corresponding to a computing power task to be processed when it is necessary to call computing power resources, where the computing power task is a task that needs to call computing power resources for processing, and the computing power package is used to measure the computing power of the intelligent computing center according to the amount of calculation;

[0178] A call module 302, configured to call computing power resources based on an application programming interface corresponding to the computing power package, where the application programming interface includes a code program corresponding to an operation for executing the computing power task, and the computing power resources called by the application programming interface match the operating environment of the computing power task.

[0179] In one embodiment, the computing power package corresponds to multiple application programming interfaces, and the call module 302 includes:

[0180] A determination unit, configured to determine a target task type corresponding to the computing power task;

[0181] A query unit, configured to query a target application programming interface corresponding to the target task type based on a preset mapping table, where the preset mapping table includes multiple task types and the multiple application programming interfaces, as well as a mapping relationship between different task types and different application programming interfaces, and the target application programming interface is one of the multiple application programming interfaces;

[0182] A call unit, configured to call computing power resources based on the target application programming interface.

[0183] In one embodiment, the multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface.

[0184] In one embodiment, after the call module 302, the device 300 for providing computing power resources through computing power packages in the intelligent computing center further includes:

[0185] A processing module processes the computing power task by applying the computing power resources called based on the target application programming interface;

[0186] Wherein, the computing power task is a computing power task processed by a target container, and the processing module includes at least one of the following:

[0187] An expansion sub-module, configured to expand the target container based on the target application programming interface when the load data is greater than or equal to a first preset threshold, where the load data is the load data of the target container;

[0188] A reduction sub-module, configured to reduce the target container based on the target application programming interface when the load data is less than or equal to a second preset threshold.

[0189] In one embodiment, the expansion sub-module includes:

[0190] A first acquisition unit, configured to acquire first computing power resource information of a first server when the load data is greater than or equal to the first preset threshold, where the first server is the server to which the target container belongs;

[0191] A first creation unit, configured to create a first container based on the computing power resources of the first server through the target application programming interface when the first computing power resource information meets the load data;

[0192] A first deployment unit, configured to deploy the relevant data of the computing power task to the first container through the target application programming interface;

[0193] A first execution unit, configured to execute the computing power task based on the first container.

[0194] In one embodiment, the expansion sub-module further includes:

[0195] A second acquisition unit, configured to acquire network information and second computing power resource information of a second server when the first computing power resource information does not meet the load data, where the second server is in the same region or a different region from the first server;

[0196] A second creation unit, configured to create a second container based on the computing power resources of the second server when the second computing power resource information meets the load data and the network information meets the preset network conditions;

[0197] A second deployment unit, configured to deploy the relevant data of the computing power task to the second container;

[0198] A second execution unit, configured to execute the computing power task based on the second container.

[0199] In one embodiment, after the processing module, the device 300 for providing computing power resources through computing power packages in the intelligent computing center further includes:

[0200] A metering module, configured to meter the consumption of computing power resources of the computing power package based on a target container, a first container, or a second container, where the target container, the first container, and the second container are containers for executing the computing power task.

[0201] The device for providing computing power resources through computing power packages in the intelligent computing center provided by the present invention can implement each process of the above-mentioned method for providing computing power resources through computing power packages in the intelligent computing center. The technical features correspond one by one and can achieve the same technical effects. To avoid repetition, they will not be elaborated here.

[0202] It should be noted that the device for providing computing power resources through computing power packages in the intelligent computing center of the present invention can be a device, or a component, an integrated circuit, or a chip in an electronic device.

[0203] The present invention also provides an electronic device. Refer to Figure 4 , Figure 4 which is a schematic structural diagram of an electronic device provided in an embodiment of the present invention. The electronic device includes a memory 401, a processor 402, and a program or instruction running on the memory 401. When the program or instruction is executed by the processor 402, it can implement Figure 1 any step in the corresponding method embodiment of providing computing power resources through computing power packages in the intelligent computing center and achieve the same beneficial effects. They will not be elaborated here.

[0204] Among them, the processor 402 can be a CPU, an ASIC, an FPGA, or a GPU.

[0205] Those of ordinary skill in the art can understand that all or part of the steps of implementing the method embodiment of providing computing power resources through computing power packages in the intelligent computing center can be completed by hardware related to program instructions, and the program can be stored in a readable medium.

[0206] The present invention also provides a readable storage medium. A computer program is stored on the readable storage medium. When the computer program is executed by a processor, it can implement any step in the above-mentioned Figure 1 corresponding method embodiment of providing computing power resources through computing power packages in the intelligent computing center and can achieve the same technical effects. To avoid repetition, they will not be elaborated here. The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0207] The present invention also provides a computer program product, including computer instructions, which implement each process of the method embodiment of the corresponding intelligent computing center providing computing power resources through computing power packages when executed by a processor, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Figure 1

[0208] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. In addition, the terms "include" and "have" and any of their variants are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the present invention means at least one of the connected objects. For example, A and / or B and / or C means including A alone, B alone, C alone, as well as the cases where A and B exist, B and C exist, A and C exist, and A, B, and C all exist, a total of 7 cases.

[0209] It should be noted that in this article, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, article or device. Without more limitations, an element defined by the statement "including one..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0210] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better embodiment. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or a second terminal device, etc.) to execute the methods of various embodiments of the present invention.

[0211] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms of expansions without departing from the spirit of the present invention and the scope protected by the claims, and all of them fall within the protection scope of the present invention.

Claims

1. A method for an intelligent computing center to provide computing resources through a computing package, characterized in that: include: Step S1: when computing power resources need to be called, determine the computing power package corresponding to the computing power task to be processed, where the computing power task is a task that needs to be processed by calling computing power resources, and the computing power package is used to measure the computing power of the intelligent computing center according to the computing amount; Step S2: calling computing resources based on the application program interface corresponding to the computing package, the application program interface including a code program corresponding to the operation of executing the computing task, and the computing resources called by the application program interface match the operating environment of the computing task; The computing power package corresponds to a plurality of application program interfaces, and the step S2 includes: Step S21, determining the target task type corresponding to the computing task; Step S22: querying a target application program interface corresponding to the target task type based on a preset mapping table, wherein the preset mapping table includes a plurality of task types and the plurality of application program interfaces, and a mapping relationship between different task types and different application program interfaces, and the target application program interface is one of the plurality of application program interfaces; Step S23: calling computing resources based on the target application program interface; The multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface.

2. The method according to claim 1, characterized in that After step S2, the method further includes: Step S3: Processing the computing task based on the computing resources called by the target application program interface; The computing task is a computing task processed by the target container, and step S3 includes at least one of the following: Step S31: when the load data is greater than or equal to a first preset threshold, expand the target container based on the target application program interface, and the load data is the load data of the target container; Step S32: When the load data is less than or equal to a second preset threshold, shrink the target container based on the target application program interface.

3. The method according to claim 2, characterized in that The step S31 comprises: Step S311: When the load data is greater than or equal to the first preset threshold, obtain first computing resource information of a first server, where the first server is the server to which the target container belongs; Step S312: When the first computing power resource information satisfies the load data, create a first container based on the computing power resources of the first server through the target application program interface; Step S313: deploy the relevant data of the computing task to the first container through the target application program interface; Step S314: Execute the computing task based on the first container.

4. The method according to claim 3, characterized in that The step S31 further includes: Step S315: When the first computing power resource information does not satisfy the load data, obtain network information and second computing power resource information of a second server, where the second server and the first server are in the same area or in different areas; Step S316: When the second computing power resource information satisfies the load data and the network information satisfies the preset network condition, create a second container based on the computing power resources of the second server; Step S317: deploy the relevant data of the computing task to the second container; Step S318: Execute the computing task based on the second container.

5. The method according to any one of claims 2 to 4, characterized in that After step S3, the method further includes: Step S4: Measure the computing resource consumption of the computing package based on the target container, the first container or the second container, where the target container, the first container and the second container are containers for executing the computing task.

6. A device for providing computing resources through computing packages in an intelligent computing center, characterized in that: include: A determination module, used to determine a computing power package corresponding to a computing power task to be processed when computing power resources need to be called, wherein the computing power task is a task that needs to be processed by calling computing power resources, and the computing power package is used to measure the computing power of the intelligent computing center according to the computing amount; A calling module, used to call computing resources based on an application program interface corresponding to the computing package, where the application program interface includes a code program corresponding to an operation of executing the computing task, and the computing resources called by the application program interface match the operating environment of the computing task; The computing power package corresponds to multiple application programming interfaces, and the calling module includes: A determination unit, used to determine a target task type corresponding to the computing task; a query unit, configured to query a target application program interface corresponding to the target task type based on a preset mapping table, wherein the preset mapping table includes a plurality of task types and the plurality of application program interfaces, and a mapping relationship between different task types and different application program interfaces, and the target application program interface is one of the plurality of application program interfaces; A calling unit, wherein the application calls computing resources based on the target application program interface; The multiple task types include at least one of a model training task type, an inference task type, and a data fine-tuning task type, and each task type corresponds to at least one application programming interface.

7. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the method for providing computing resources through a computing power package by an intelligent computing center as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for an intelligent computing center to provide computing resources through a computing power package as described in any one of claims 1 to 5.

9. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the method for an intelligent computing center to provide computing resources through a computing power package as described in any one of claims 1 to 5.

Images