Remote Development Method and Device for Intelligent Computing Center Model Oriented to Inclusive Computing Power

Abstract

The present invention provides a remote development method and device for an intelligent computing center model for inclusive computing power, which relates to the technical fields of intelligent computing centers, intelligent computing centers, and computing power infrastructure. The method includes: when receiving operation information, adjusting the operating environment of a first container based on the operation information and generating a target environment file, where the target environment file is used to characterize the relevant situation of the operating environment; when receiving a code adjustment instruction, creating the operating environment of a second container based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center; adjusting the model code in the operating environment of the second container based on the code adjustment instruction; scheduling the GPU resources of the intelligent computing center to process the adjusted model code. The present invention can greatly improve the utilization rate of computing power resources of the intelligent computing center and greatly reduce the economic cost of model development.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent computing centers, intelligent computing centers, and computing power infrastructure, and specifically relates to a method and device for remotely developing an intelligent computing center model for inclusive computing power. 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, mainly to provide the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios for developing, training, and inferring artificial intelligence deep learning models). The 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 that provides computing power services, data services, and algorithm services required for artificial intelligence applications based on artificial intelligence theory and using an artificial intelligence computing architecture.

[0006] "Computing power" is the core of the "intelligent computing center" and the "intelligent computing center", which is the ability of a computer device or a computing / data center 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 the target result by processing information data, and a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0007] Currently, the intelligent computing center provides computing power resources for users through Graphics Processing Unit (GPU) acceleration cards, so that users can develop models based on the computing power resources of the intelligent computing center and reduce the cost of model development. However, in the prior art, the model development process is long, and users need to frequently adjust the model during the development process. The adjustment process still needs to occupy the computing power resources of the intelligent computing center, resulting in the long-term occupation of the computing power resources of the intelligent computing center and very low utilization rate of the computing power resources. At the same time, users need to lease the computing power services of the intelligent computing center for a long time, resulting in a high cost of model development and making it difficult to achieve the widespread application of inclusive computing power.

[0008] It can be seen that there are problems of very low utilization rate of computing power resources and high cost of model development in the prior art. Summary of the Invention

[0009] The present invention provides a method and device for remotely developing an intelligent computing center model for inclusive computing power, so as to solve the problems of low utilization rate of computing power resources and high cost of model development 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 remotely developing an intelligent computing center model for inclusive computing power, including:

[0012] Step S1: When receiving operation information, adjust the running environment of the first container based on the operation information, and generate a target environment file, where the target environment file is used to characterize the relevant situation of the running environment, and the first container is a container created based on the central processing unit (CPU) resources of the intelligent computing center;

[0013] Step S2: When receiving a code adjustment instruction, create the running environment of the second container based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center;

[0014] Step S3: Adjust the model code in the running environment of the second container based on the code adjustment instruction;

[0015] Step S4: Schedule the graphics processing unit (GPU) resources of the intelligent computing center to process the adjusted model code.

[0016] In one embodiment, step S1 includes:

[0017] Step S11: Obtain an initial environment file, where the initial environment file is the file corresponding to the container when the model code was last adjusted;

[0018] Step S12: When receiving the operation information, adjust the running environment of the first container based on the operation information to obtain an adjustment result;

[0019] Step S13: When the adjustment result meets a first preset condition, save the adjustment result to the initial environment file in layers to obtain the target environment file.

[0020] In one embodiment, the first preset condition includes at least one of the following:

[0021] The data volume of the adjustment result is greater than or equal to a first set data threshold;

[0022] The time since the last save of the adjustment result is greater than or equal to a set time threshold.

[0023] In one embodiment, the initial environment file includes at least one storage layer, and step S13 includes:

[0024] Step S131, traverse at least one storage layer of the initial environment file;

[0025] Step S132, determine a target storage layer, where the target storage layer is a storage layer in the at least one storage layer that meets a second preset condition;

[0026] Step S133, save the adjustment result to the target storage layer to obtain the target environment file.

[0027] In one embodiment, the second preset condition includes at least one of the following:

[0028] The data volume of the target storage layer is less than or equal to a second set data threshold;

[0029] The sum of the data volume of the target storage layer and the data volume of the adjustment result is less than or equal to a third set data threshold;

[0030] The number of layers of the target storage layer is less than or equal to a set number threshold.

[0031] In one embodiment, step S1 includes:

[0032] Step S14, when the operation information is received and a save instruction is received, adjust the running environment of the first container based on the operation information to obtain an adjustment result;

[0033] Step S15, save multiple processes of the first container after adjusting the running environment, and memory information corresponding to each process to obtain the target environment file, where at least one process in the multiple processes is a process corresponding to the adjustment result.

[0034] In one embodiment, step S2 includes at least one of the following:

[0035] Step S21, when the first container and the second container meet the affinity rule, perform affinity scheduling on the target environment file, and create the running environment of the second container based on the target environment file obtained by the scheduling;

[0036] Step S22, when the first container and the second container do not meet the affinity rule, schedule the target environment file based on the peer-to-peer (P2P) network, and create the running environment of the second container based on the target environment file obtained by the scheduling.

[0037] In one embodiment, step S4 includes:

[0038] Step S41: Create a third container based on the GPU resources of the intelligent computing center;

[0039] Step S42: Process the adjusted model code based on the third container to obtain a processing result;

[0040] After the step S4, the method further includes:

[0041] Step S5: Send the processing result to the terminal, close the third container, and release the GPU resources of the intelligent computing center occupied by the third container.

[0042] In a second aspect, the present invention further provides an intelligent computing center model remote development device for inclusive computing power, including:

[0043] A generation module, configured to, when receiving operation information, adjust the running environment of a first container based on the operation information and generate a target environment file, where the target environment file is used to characterize the relevant situation of the running environment, and the first container is a container created based on the CPU resources of the intelligent computing center;

[0044] A creation module, configured to, when receiving a code adjustment instruction, create the running environment of a second container based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center;

[0045] An adjustment module, configured to adjust the model code in the running environment of the second container based on the code adjustment instruction;

[0046] A processing module, configured to schedule the GPU resources of the intelligent computing center to process the adjusted model code.

[0047] 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, where when the computer program is executed by the processor, the steps in the intelligent computing center model remote development method for inclusive computing power described in the first aspect above are implemented.

[0048] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the intelligent computing center model remote development method for inclusive computing power described in the first aspect above are implemented.

[0049] In a fifth aspect, the present invention further provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps in the remote development method of the intelligent computing center model for inclusive computing power as described in the first aspect above.

[0050] In the present invention, when an operation information is received, the operating environment of the first container is adjusted based on the operation information, and a target environment file is generated, where the target environment file is used to characterize the relevant situation of the operating environment. The first container is a container created based on the central processing unit (CPU) resources of the intelligent computing center; when a code adjustment instruction is received, the operating environment of the second container is created based on the target environment file, and the second container is a container created based on the CPU resources of the intelligent computing center; the model code is adjusted in the operating environment of the second container based on the code adjustment instruction; and the graphics processing unit (GPU) resources of the intelligent computing center are scheduled to process the adjusted model code. In this way, by generating the target environment file, when the code adjustment instruction is received, the CPU resources are occupied to create the second container, and the operating environment of the second container is created through the target environment file, enabling the user to adjust the model code in the second container through the code adjustment instruction without occupying the computing power resources of the intelligent computing center, greatly improving the utilization rate of the computing power resources of the intelligent computing center. Further, since there is no need for the user to lease the computing power service of the intelligent computing center for a long time, the economic cost of the user for model development is greatly reduced, and the wide application of inclusive computing power is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] 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 without creative efforts based on these drawings.

[0052] Figure 1 is a flowchart of a remote development method of an intelligent computing center model for inclusive computing power provided by the present invention;

[0053] Figure 2 is a structural diagram of a remote development device of an intelligent computing center model for inclusive computing power provided by the present invention;

[0054] Figure 3 is a structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The following will clearly and completely describe the technical solutions in the present invention in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0056] The "computing power" as described in the present invention refers to: the ability of a computer device or a computing / data center 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 a target result through processing information data, a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly providing services to society through computing power infrastructure.

[0057] The "computational power" (Computational Power, CP) as described in the present invention refers to: the ability of a data center server to process data and achieve result output, a comprehensive indicator for measuring the computing ability of a 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 .

[0058] The "carrying capacity" (Network Power, NP) as described in the present invention refers to: the performance of the data transmission ability of computing power facilities, a comprehensive ability including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., involving network transmission within and between data centers, and a comprehensive indicator for measuring network transmission scheduling ability.

[0059] The "storage power" (Storage Power, SP) as described in the present invention refers to: the comprehensive ability of a data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon, a comprehensive indicator for measuring the data storage ability of a 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), 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), and the disaster recovery ratio is an important manifestation of security and reliability.

[0060] The "computing power infrastructure" described in the present invention refers to: a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity, and can realize the centralized computing, storage, transmission, and application of information.

[0061] The "new type of information infrastructure" described in the present invention mainly includes network infrastructures such as 5G networks, fiber broadband networks, backbone networks, international communication networks, and satellite Internet, computing power infrastructures such as data centers, general computing power centers, intelligent computing centers, and supercomputing centers, and new technology infrastructures such as artificial intelligence, blockchain, and quantum computing.

[0062] The "computing power" described in the present invention includes: general computing power, intelligent computing power, and super computing power.

[0063] 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.

[0064] The "intelligent computing power" described in the present invention refers to: for various artificial intelligence innovation applications, a computing platform is 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, and so on.

[0065] The "super computing power" described in the present invention mainly refers to: 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.

[0066] The "intelligent computing center" described in the present invention refers to: a facility that mainly provides 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) 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.

[0067] The "intelligent computing center" described in the present invention includes but is not limited to the "intelligent computing center".

[0068] 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.

[0069] The "Computing Power Center" described in the present invention refers to: a facility mainly composed of infrastructure such as wind, fire, water, and electricity and IT software and hardware devices, with computing power, transportation capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0070] 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.

[0071] The "Computing Power Resources" described in the present invention refers 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.

[0072] 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.

[0073] The "Model" described in the present invention includes but is not limited to "Large Language Model" and "Multimodal Large Model".

[0074] The "Large Language Model" described in the present invention refers to the large 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.

[0075] The "Multimodal Large Model" described in the present invention (Multimodal Large Models) refers to: a model that jointly trains multimodal information such as text, images, videos, and audio, including but not limited to multimodal large language models.

[0076] Please refer to Figure 1 , Figure 1 which is a flowchart of a remote development method for an intelligent computing center model for inclusive computing power provided by the present invention. As Figure 1 shown, it includes the following steps:

[0077] Step S1: When receiving operation information, adjust the running environment of the first container based on the operation information, and generate a target environment file, which is used to characterize the relevant situation of the running environment. The first container is a container created based on the CPU resources of the intelligent computing center.

[0078] The above operation information is the information sent by the user when remotely adjusting the running environment of the first container. When the intelligent computing center receives the operation information, it parses the operation information to obtain the specific content of the adjustment of the running environment, and then realizes the adjustment of the running environment of the first container through the specific content to obtain an adjustment result.

[0079] Among them, the operation information packet is used to characterize at least one operation, such as operations on the system like installation, update, or uninstallation of the development package of the running environment.

[0080] The above adjustment result is the adjustment result obtained by the intelligent computing center after receiving the operation information and adjusting the running environment of the first container. Among them, the adjustment result corresponds to the received operation information. When multiple operation information is received, the first container will be adjusted multiple times to obtain multiple adjustment results. The multiple adjustment results correspond to the multiple operation information one by one. By saving the multiple adjustment results, the target environment file can be obtained, and the running environment of the first container can be restored through the target environment file.

[0081] In some embodiments, before the user sends operation information, it is necessary to install Integrated Development Environment (IDE) software on the terminal. Through the IDE software, the operation information can be sent to the intelligent computing center to realize the adjustment of the running environment of the first container. Among them, the IDE software can be the code integrated development software provided by the intelligent computing center or the industry standard IDE extension plug-in. Through the code integrated development software or the industry standard IDE extension plug-in, the creation of the running environment can be realized, and the computing power resources of the intelligent computing center can be scheduled to process the adjusted model code.

[0082] Specifically, the user logs in on the terminal, creates the running environment of the first container through the code integrated development software or the IDE extension plug-in, and realizes the adjustment of the running environment of the first container through the operation information, so that the running environment of the first container of the intelligent computing center meets the requirements of the running model code.

[0083] Further, the user can upload the model code and running data to be adjusted to the storage layer of the intelligent computing center through the IDE software, or directly upload the model code and running data to be adjusted to the second container through the IDE software, improving the flexibility of model development.

[0084] The above-mentioned first container is a container created based on the CPU resources of the intelligent computing center. It should be noted that since the intelligent computing center provides computing power resources to run the model code through GPU acceleration cards, and the number of GPU acceleration cards that can be deployed in the intelligent computing center is limited and cannot provide computing power resources indefinitely. To reduce the occupancy of the computing power resources of the intelligent computing center during the model adjustment process, in the present invention, the first container is created based on the CPU resources. By setting the running environment for model operation in the first container and remotely updating or adjusting the model code in the first container, the occupancy of the computing power resources of the intelligent computing center is reduced, thereby improving the utilization rate of the computing power resources of the intelligent computing center; further, since it is not necessary for the user to rent the computing power service of the intelligent computing center for a long time, the economic cost of the user for model development is greatly reduced, and the wide application of inclusive computing power is realized.

[0085] Among them, the running environment of the first container is remotely created and adjusted by the user. Specifically, the user sends operation information to the intelligent computing center, and the intelligent computing center creates and adjusts the running environment of the first container through the operation information.

[0086] The above-mentioned target environment file is used to characterize the relevant situation of the running environment of the first container. Based on the target environment file, the running environment of the first container can be restored in other containers, so that when the user does not need to adjust the model code, the first container can be closed, and when the user needs to adjust the model code, the second container can be temporarily created, and the running environment of the second container is created based on the target environment file, enabling the user to adjust the model code through the temporarily created second container.

[0087] Step S2, in the case of receiving a code adjustment instruction, create the running environment of the second container based on the target environment file. The second container is a container created based on the CPU resources of the intelligent computing center.

[0088] The above-mentioned code adjustment instruction is an instruction sent by the user when the model code needs to be adjusted. Through the code adjustment instruction, the user can remotely adjust the model code of the model deployed in the intelligent computing center.

[0089] The running environment of the above-mentioned second container is created based on the target environment file, and the environment of the second container is consistent with that of the first container, enabling the second container to deploy the model code, and the user can remotely adjust the model code through the second container.

[0090] It should be noted that both the first container and the second container can enable users to remotely adjust the model code. However, since users may have other requirements during the model development process and cannot continuously adjust the model code, in order to reduce the occupancy of CPU resources in the intelligent computing center, when the model does not need to be adjusted, the first container needs to be closed, and then the second container is started later to enable users to remotely adjust the model code.

[0091] That is, in some embodiments, when the first container is running in the intelligent computing center, there is no need to create a second container, but instead, the model code is directly adjusted in the running environment of the first container based on the code adjustment instruction.

[0092] In some embodiments, when the first container is not running in the first intelligent computing center, the running environment of the second container is created based on the target environment file, and then the model code is adjusted in the running environment of the second container based on the code adjustment instruction.

[0093] Step S3: Adjust the model code in the running environment of the second container based on the code adjustment instruction.

[0094] It should be noted that the model code is arranged in the second container, and the running environment of the second container is the environment corresponding to the model code. Through the second container, users can remotely adjust the model code in the second container.

[0095] Specifically, after the intelligent computing center receives the code adjustment instruction, a second container is created, and the running environment of the second container is created based on the target environment file. The model code is arranged in the second container; the code adjustment instruction is parsed to obtain the code adjustment content; and the model code in the second container is adjusted based on the code adjustment content.

[0096] In some embodiments, the model code is stored in the storage layer of the intelligent computing center. When the model code does not need to be adjusted, the container is closed, and the model code is stored in the storage layer. At this time, the model code only occupies the storage resources of the intelligent computing center and does not occupy CPU resources and GPU resources; when the model code needs to be adjusted, a second container is created, and the model code is obtained from the storage layer and arranged in the second container. Then, the user adjusts the model code in the second container through the code adjustment instruction, improving the utilization rate of CPU resources and GPU resources in the intelligent computing center.

[0097] Step S4: Schedule the graphics processing unit (GPU) resources of the intelligent computing center to process the adjusted model code.

[0098] After the adjustment of the model code is completed in the second container, the graphics processing unit (GPU) resources of the intelligent computing center are then scheduled to process the adjusted model code, thereby avoiding long-term occupation of the computing power resources of the intelligent computing center and improving the utilization rate of the computing power resources of the intelligent computing center.

[0099] In the present invention, in the case of receiving operation information, the operating environment of the first container is adjusted based on the operation information, and a target environment file is generated, where the target environment file is used to characterize the relevant situation of the operating environment. The first container is a container created based on the central processing unit (CPU) resources of the intelligent computing center; in the case of receiving a code adjustment instruction, the operating environment of the second container is created based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center; the model code is adjusted in the operating environment of the second container based on the code adjustment instruction; the GPU resources of the intelligent computing center are scheduled to process the adjusted model code. In this way, by generating the target environment file, when receiving the code adjustment instruction, the CPU resources are occupied to create the second container, and the operating environment of the second container is created through the target environment file, enabling the user to adjust the model code in the second container through the code adjustment instruction without occupying the computing power resources of the intelligent computing center, greatly improving the utilization rate of the computing power resources of the intelligent computing center. Further, since there is no need for the user to lease the computing power service of the intelligent computing center for a long time, the economic cost of the user for model development is significantly reduced, realizing the wide application of inclusive computing power.

[0100] In one embodiment, the step S1 includes:

[0101] Step S11, obtaining an initial environment file, where the initial environment file is the file corresponding to the container when the model code was last adjusted;

[0102] Step S12, in the case of receiving the operation information, adjusting the operating environment of the first container based on the operation information to obtain an adjustment result;

[0103] Step S13, in the case where the adjustment result meets a first preset condition, storing the adjustment result in layers in the initial environment file to obtain the target environment file.

[0104] The above initial environment file is the file corresponding to the container when the model code was last adjusted, that is, the initial environment file is the latest environment file corresponding to the model code stored in the intelligent computing center. In the case of receiving operation information, the initial environment file needs to be updated so that the target environment file obtained through the update can obtain the updated operating environment.

[0105] It should be noted that when creating the environment file for the first time, the intelligent computing center pre-creates the first container, and then the user remotely creates the running environment through the operation information. Specifically, after the intelligent computing center receives the operation information, it adjusts the running environment of the first container to obtain an adjustment result. When the adjustment result meets the first preset condition, the adjustment result is saved to generate the first environment file. When new operation information is received subsequently, a new adjustment result is generated based on the new operation information, and the new adjustment results are continuously saved on the basis of the first environment file to obtain the target environment file. The running environment of the model code can be restored through the target environment file.

[0106] In the present invention, an initial environment file is obtained, and the initial environment file is the file corresponding to the container during the last adjustment of the model code. When the operation information is received, the running environment of the first container is adjusted based on the operation information to obtain an adjustment result. When the adjustment result meets the first preset condition, the adjustment result is hierarchically saved to the initial environment file to obtain the target environment file. In this way, the periodic update of the environment file is realized through the first preset condition.

[0107] In one embodiment, the first preset condition includes at least one of the following:

[0108] The data volume of the adjustment result is greater than or equal to the first set data threshold;

[0109] The time since the last saving of the adjustment result is greater than or equal to the set time threshold.

[0110] The above first set data threshold is used to determine whether the data volume of the adjustment result meets the saving condition. It should be noted that during the process of the user remotely adjusting the running environment of the container, the running environment may need to be frequently adjusted. At this time, if the adjustment result corresponding to the operation information is saved as the target environment file every time an operation information is received, it will occupy a large amount of computing resources. Therefore, in the present invention, the first set data threshold is set, and whether to save the adjustment result is determined by whether the data volume of the adjustment result meets the first set data threshold.

[0111] For example, the first set data threshold is set to 100M, the operation information is continuously received, and the running environment is adjusted based on the received operation information to obtain an adjustment result. When the data volume of the received adjustment result reaches 100M, the adjustment results corresponding to all the received operation information are saved to the initial environment file to obtain the target environment file.

[0112] In addition to the above methods, a set time threshold can also be set to determine whether to save the adjustment result by the set time threshold. That is, in the present invention, during the process of continuously receiving operation information, the adjustment result corresponding to the operation information is saved once every fixed time by setting the time threshold, which can also reduce the computing power resources occupied by saving the adjustment result.

[0113] For example, set the set time threshold to 30s, continuously receive operation information, and adjust the operating environment based on the received operation information to obtain an adjustment result; when the time since the last save of the adjustment result reaches 30s, save the adjustment results corresponding to all the received operation information to the initial environment file to obtain the target environment file.

[0114] In the present invention, by setting the first set data threshold or the set time threshold, the adjustment result is saved regularly, avoiding occupying too much computing power resources of the intelligent computing center due to frequent saving of the adjustment result.

[0115] In one embodiment, the initial environment file includes at least one save layer, and step S13 includes:

[0116] Step S131, traverse at least one save layer of the initial environment file;

[0117] Step S132, determine the target save layer, where the target save layer is the save layer that meets the second preset condition among the at least one save layer;

[0118] Step S133, save the adjustment result to the target save layer to obtain the target environment file.

[0119] It should be noted that since the user remotely adjusts the operating environment of the container is a continuous process, in order to ensure the stability of the operating environment created by the environment file, in the present invention, the adjustment results are saved in layers. Among them, the initial environment file includes at least one save layer, and each save layer is used to save the adjustment result, and the adjustment results are saved successively through layered saving.

[0120] The above target save layer is the save layer that meets the second preset condition. It should be noted that the initial environment file includes at least one save layer, and by traversing the at least one save layer, the target save layer that meets the second preset condition is determined, and the adjustment result can be saved through the target save layer.

[0121] Among them, the second preset condition includes at least one of the following:

[0122] The data volume of the target save layer is less than or equal to the second set data threshold;

[0123] The sum of the data volume of the target storage layer and the data volume of the adjustment result is less than or equal to a third set data threshold;

[0124] The number of layers of the target storage layer is less than or equal to a set number threshold.

[0125] It should be noted that each storage layer of the target environment file cannot store the adjustment result infinitely, and the target environment file cannot store the adjustment result infinitely either. Therefore, in the present invention, the data volume stored in each storage layer and the number of layers of the target environment file are restricted.

[0126] Among them, the second set data threshold and the third set data threshold are used to restrict the data volume stored in each storage layer, and the second set data threshold and the third set data threshold may be the same or different; the set number threshold is used to restrict the number of layers of the target environment file.

[0127] For example, the second set data threshold and the third set data threshold are the same, both set to 10G, and the set number threshold is 128 layers, so that the size of each storage layer in the target environment file does not exceed 10G after storing the adjustment result, and the size of the target environment file does not exceed 1280G.

[0128] In the present invention, by using the second set data threshold and the third set data threshold to restrict the data volume stored in each storage layer, and the set number threshold to restrict the number of layers of the target environment file, the target environment file does not occupy too much storage resources of the intelligent computing center.

[0129] In one embodiment, the step S1 includes:

[0130] Step S14, when receiving the operation information and a save instruction, adjusting the running environment of the first container based on the operation information to obtain an adjustment result;

[0131] Step S15, saving multiple processes of the first container after adjusting the running environment and the memory information corresponding to each process to obtain the target environment file, and at least one process among the multiple processes is the process corresponding to the adjustment result.

[0132] It should be noted that, in addition to the solution of periodically saving the adjustment result in the above text, the instant saving of the environment information can also be implemented according to the user's saving instruction. Since the user usually sends the saving instruction when the phased adjustment of the running environment is not completed, that is, a complete adjustment result has not been formed, and at this time, the adjustment result cannot be directly saved. Therefore, in the present invention, the processes of the first container and the memory information corresponding to each process are directly saved to obtain a target environment file, so that the processes in the first container can be restored through the target environment file, and the user can continue to complete the adjustment of the running environment on the basis of restoring the processes.

[0133] Among them, the saving instruction is an instruction sent by the user through the IDE software. When the saving instruction is received, the intelligent computing center saves multiple processes of the first container and the memory information corresponding to each process as a target environment file. At this time, the target environment file is a binary file and is used to restore the processes of the container.

[0134] Furthermore, since the first container is a container created based on CPU resources, the above multiple processes do not include GPU processes, thereby reducing the size of the target environment file.

[0135] In one embodiment, the step S2 includes at least one of the following:

[0136] Step S21, when the first container and the second container satisfy the affinity rule, perform affinity scheduling on the target environment file, and create the running environment of the second container based on the scheduled target environment file;

[0137] Step S22, when the first container and the second container do not satisfy the affinity rule, schedule the target environment file based on a point-to-point (P2P) network, and create the running environment of the second container based on the scheduled target environment file.

[0138] The above affinity rule is used to characterize the affinity degree between the first container and the second container. Specifically, the affinity rule can be whether the first container and the second container are located on the same node, or whether the first container and the second container use the computing power resources of the same GPU. It should be noted that when the first container and the second container satisfy the affinity rule, the affinity scheduling of the target environment file can be directly performed to quickly deploy the target environment file in the second container, improve the creation speed of the running environment of the second container, and reduce the latency when the user uses it.

[0139] Further, in the case where the first container and the second container do not satisfy the affinity rule, affinity scheduling cannot be directly used. To improve the speed of creating the running environment of the second container, in the present invention, the target environment file is quickly obtained through the P2P network, and then the target environment file is quickly deployed in the second container to improve the speed of creating the running environment of the second container and reduce the latency when the user uses it.

[0140] In one embodiment, step S4 includes:

[0141] Step S41, creating a third container based on the GPU resources of the intelligent computing center;

[0142] Step S42, processing the adjusted model code based on the third container to obtain a processing result;

[0143] After step S4, the method further includes:

[0144] Step S5, sending the processing result to the terminal, and closing the third container to release the GPU resources of the intelligent computing center occupied by the third container.

[0145] In the present invention, a third container is created based on the GPU resources of the intelligent computing center; the adjusted model code is processed based on the third container to obtain a processing result; the processing result is sent to the terminal, and the third container is closed to release the GPU resources of the intelligent computing center occupied by the third container. In this way, by using the GPU computing power resources of the intelligent computing center by the third container to run the adjusted model code and releasing the computing power resources of the intelligent computing center after the operation is completed, the utilization rate of the computing power resources of the intelligent computing center is improved.

[0146] Further, after obtaining the processing result, the computing power resources of the intelligent computing center are released, reducing the time for the user to lease computing power resources and realizing the wide application of inclusive computing power.

[0147] In some embodiments, the processing result includes a running log, enabling the user to determine the running situation of the updated model code through the running log for further optimization of the model code.

[0148] It should be noted that the method of the present invention realizes the consistency guarantee of the running environment, code, and data, and the user can quickly use inclusive computing power with one key. Specifically, before step S41, the method includes:

[0149] Step S40, receiving a parameter instruction sent by the terminal, where the parameter instruction includes the required parameters for running the model code;

[0150] Step S42 includes:

[0151] Step S421: Process the updated model code based on the container and the requirement parameters to obtain a processing result.

[0152] Among them, the requirement parameters may include operation parameters. The intelligent computing center can infer the required operating environment and resource size for the development task based on the operation parameters and the updated model code.

[0153] Furthermore, the model code can be updated based on the mirror rapid loading technology and the updated model code can be processed, which can achieve millisecond-level startup, making users unaware of the existence of the remote intelligent computing center and improving the user experience.

[0154] In some embodiments, the intelligent computing center can share the computing power resources of one GPU acceleration card among multiple containers, which can further reduce the user's usage cost and the system load of the intelligent computing center.

[0155] Please refer to Figure 2 , Figure 2 which is the structural diagram of an intelligent computing center model remote development device for inclusive computing power provided by the present invention. As Figure 2 shown, the intelligent computing center model remote development device 200 for inclusive computing power includes:

[0156] A generation module 201, configured to, when receiving operation information, adjust the operating environment of the first container based on the operation information and generate a target environment file, where the target environment file is used to characterize the relevant situation of the operating environment, and the first container is a container created based on the central processing unit (CPU) resources of the intelligent computing center;

[0157] A creation module 202, configured to, when receiving a code adjustment instruction, create the operating environment of the second container based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center;

[0158] An adjustment module 203, configured to adjust the model code in the operating environment of the second container based on the code adjustment instruction;

[0159] A processing module 204, configured to schedule the graphics processing unit (GPU) resources of the intelligent computing center to process the adjusted model code.

[0160] In one embodiment, the generation module 201 includes:

[0161] An acquisition unit, configured to acquire an initial environment file, where the initial environment file is the file corresponding to the container when the model code was last adjusted;

[0162] A first adjustment unit, configured to, when receiving the operation information, adjust the running environment of a first container based on the operation information to obtain an adjustment result;

[0163] A first storage unit, configured to, when the adjustment result meets a first preset condition, hierarchically store the adjustment result into an initial environment file to obtain the target environment file.

[0164] In one embodiment, the first preset condition includes at least one of the following:

[0165] The data volume of the adjustment result is greater than or equal to a first set data threshold;

[0166] The time since the last storage of the adjustment result is greater than or equal to a set time threshold.

[0167] In one embodiment, the initial environment file includes at least one storage layer, and the first storage unit includes:

[0168] A traversal subunit, configured to traverse at least one storage layer of the initial environment file;

[0169] A determination subunit, configured to determine a target storage layer, where the target storage layer is a storage layer that meets a second preset condition among the at least one storage layer;

[0170] A storage subunit, configured to store the adjustment result into the target storage layer to obtain the target environment file.

[0171] In one embodiment, the second preset condition includes at least one of the following:

[0172] The data volume of the target storage layer is less than or equal to a second set data threshold;

[0173] The sum of the data volume of the target storage layer and the data volume of the adjustment result is less than or equal to a third set data threshold;

[0174] The number of layers of the target storage layer is less than or equal to a set number threshold.

[0175] In one embodiment, the generation module 201 includes:

[0176] A second adjustment unit, configured to, when receiving the operation information and a storage instruction, adjust the running environment of a first container based on the operation information to obtain an adjustment result;

[0177] A second storage unit, configured to store multiple processes of the first container after adjusting the running environment and the memory information corresponding to each process to obtain the target environment file, where at least one process among the multiple processes is the process corresponding to the adjustment result.

[0178] In one embodiment, the creation module 202 includes at least one of the following:

[0179] A first creation unit, configured to perform affinity scheduling on the target environment file when the first container and the second container meet the affinity rule, and create a running environment for the second container based on the scheduled target environment file;

[0180] A second creation unit, configured to schedule the target environment file based on a peer-to-peer (P2P) network when the first container and the second container do not meet the affinity rule, and create a running environment for the second container based on the scheduled target environment file.

[0181] In one embodiment, the processing module 204 includes:

[0182] A third creation unit, configured to create a third container based on the GPU resources of the intelligent computing center;

[0183] A processing unit, configured to process the adjusted model code based on the third container to obtain a processing result;

[0184] After the processing module 204, the intelligent computing center model remote development device 200 for inclusive computing power further includes:

[0185] A release module, configured to send the processing result to a terminal, close the third container, and release the GPU resources of the intelligent computing center occupied by the third container.

[0186] The intelligent computing center model remote development device provided by the present invention can implement each process of the above-mentioned intelligent computing center model remote development method for inclusive computing power. The technical features correspond one by one and can achieve the same technical effects. To avoid repetition, details are not described here again.

[0187] It should be noted that the intelligent computing center model remote development device in the present invention can be a device, or a component, an integrated circuit, or a chip in an electronic device.

[0188] The present invention also provides an electronic device. Refer to Figure 3 , Figure 3 which is a schematic structural diagram of an electronic device provided in an embodiment of the present invention. The electronic device includes a memory 301, a processor 302, and a program or instruction stored in the memory 301 and running on the processor 302. When the program or instruction is executed by the processor 302, it can implement Figure 1Any steps in the corresponding embodiment of the remote development method for the intelligent computing center model for inclusive computing power and the same beneficial effects achieved thereby will not be elaborated here.

[0189] Among them, the processor 302 can be a CPU, ASIC, FPGA or GPU.

[0190] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned embodiment of the remote development method for the intelligent computing center model for inclusive computing power can be completed by hardware related to program instructions, and the said program can be stored in a readable medium.

[0191] The present invention also provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it can implement any step in the corresponding Figure 1 embodiment of the remote development method for the intelligent computing center model for inclusive computing power, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. The said storage medium, such as a Read-Only Memory (ROM), Random Access Memory (RAM), magnetic disk or optical disc, etc.

[0192] The present invention also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement each process in the corresponding Figure 1 embodiment of the remote development method for the intelligent computing center model for inclusive computing power, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0193] 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 "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, 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 application 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, and both A and B exist, both B and C exist, both A and C exist, and all of A, B, and C exist, a total of 7 situations.

[0194] It should be noted that in this text, the term "including", "comprising" 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 expressly listed, or further includes elements inherent in such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including that element.

[0195] From the description of the above embodiments, those skilled in the art can clearly understand that the above-described 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 implementation. Based on such an understanding, the technical solution of the present application, 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 may be a mobile phone, computer, server, air conditioner, or a second terminal device, etc.) to execute the methods of the various embodiments of the present application.

[0196] The embodiments of the present application are described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Those of ordinary skill in the art, under the inspiration of the present application and without departing from the purpose of the present application and the scope protected by the claims, can also make many forms, all of which fall within the protection scope of the present application.

Claims

1. A remote development method of an intelligent computing center model for universal computing power, characterized in that: include: Step S1: When operation information is received, the operating environment of the first container is adjusted based on the operation information, and a target environment file is generated, where the target environment file is used to characterize relevant conditions of the operating environment, and the first container is a container created based on the central processing unit CPU resources of the intelligent computing center; Step S2: when receiving the code adjustment instruction, creating an operating environment of a second container based on the target environment file, where the second container is a container created based on the CPU resources of the intelligent computing center; Step S3: adjusting the model code in the running environment of the second container based on the code adjustment instruction; Step S4, dispatching the graphics processor GPU resources of the intelligent computing center to process the adjusted model code; The step S1 comprises: Step S11, obtaining an initial environment file, wherein the initial environment file is a file corresponding to the container when the model code is last adjusted; Step S12: when receiving the operation information, adjusting the operating environment of the first container based on the operation information to obtain an adjustment result; Step S13: if the adjustment result meets the first preset condition, save the adjustment result in layers to the initial environment file to obtain the target environment file; The first preset condition includes at least one of the following: The data volume of the adjustment result is greater than or equal to a first set data threshold; The time since the last adjustment result was saved is greater than or equal to the set time threshold.

2. The method according to claim 1, characterized in that The initial environment file includes at least one storage layer, and the step S13 includes: Step S131, traversing at least one storage layer of the initial environment file; Step S132: determining a target preservation layer, wherein the target preservation layer is a preservation layer that satisfies a second preset condition among the at least one preservation layer; Step S133: Save the adjustment result to the target storage layer to obtain the target environment file.

3. The method according to claim 2, characterized in that The second preset condition includes at least one of the following: The data volume of the target storage layer is less than or equal to a second set data threshold; The sum of the data volume of the target storage layer and the data volume of the adjustment result is less than or equal to a third set data threshold; The number of layers of the target preservation layer is less than or equal to a set number threshold.

4. The method according to claim 1, characterized in that The step S1 comprises: Step S14: when the operation information is received and a save instruction is received, the operating environment of the first container is adjusted based on the operation information to obtain an adjustment result; Step S15: Save the multiple processes of the first container after adjusting the running environment and the memory information corresponding to each process to obtain the target environment file, wherein at least one process among the multiple processes is a process corresponding to the adjustment result.

5. The method according to claim 1 or 4, characterized in that The step S2 includes at least one of the following: Step S21: When the first container and the second container satisfy the affinity rule, affinity scheduling is performed on the target environment file, and a running environment of the second container is created based on the target environment file obtained by scheduling; Step S22: If the first container and the second container do not satisfy the affinity rule, schedule the target environment file based on a peer-to-peer P2P network, and create a running environment for the second container based on the scheduled target environment file.

6. The method according to claim 1 or 4, characterized in that: The step S4 comprises: Step S41: Create a third container based on the GPU resources of the intelligent computing center; Step S42: Processing the adjusted model code based on the third container to obtain a processing result; After step S4, the method further includes: Step S5: Send the processing result to the terminal, close the third container, and release the GPU resources of the intelligent computing center occupied by the third container.

7. A remote development device for an intelligent computing center model for universal computing power, characterized in that: include: A generation module, configured to adjust the operating environment of the first container based on the operating information when the operating information is received, and generate a target environment file, wherein the target environment file is used to characterize relevant conditions of the operating environment, wherein the first container is a container created based on the central processing unit (CPU) resources of the intelligent computing center; A creation module, configured to create an operating environment of a second container based on the target environment file when a code adjustment instruction is received, wherein the second container is a container created based on the CPU resources of the intelligent computing center; An adjustment module, configured to adjust the model code in the running environment of the second container based on the code adjustment instruction; A processing module, used for scheduling the GPU resources of the intelligent computing center to process the adjusted model code; The generation module comprises: An acquisition unit, used to acquire an initial environment file, where the initial environment file is a file corresponding to the container when the model code is last adjusted; A first adjustment unit, configured to adjust the operating environment of the first container based on the operation information when receiving the operation information, to obtain an adjustment result; A first saving unit, configured to save the adjustment result in layers to an initial environment file to obtain the target environment file if the adjustment result meets a first preset condition; The first preset condition includes at least one of the following: The data volume of the adjustment result is greater than or equal to a first set data threshold; The time since the last adjustment result was saved is greater than or equal to the set time threshold.

8. 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 remote development method of an intelligent computing center model for inclusive computing power as described in any one of claims 1 to 6 are implemented.

9. 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 remote development method of an intelligent computing center model for inclusive computing power as described in any one of claims 1 to 6.

10. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the remote development method of an intelligent computing center model for universal computing power as described in any one of claims 1 to 6.

Images