Intelligent computing center model development method and device for inclusive computing power

Abstract

The present invention provides a method and device for developing a model of an intelligent computing center for inclusive computing power, and relates to the technical field of computing power infrastructure. The method comprises: step S1, receiving first image data of a model to be developed; step S2, obtaining a first node, the first node being a node occupying GPU resources of a graphics processor of an intelligent computing center, the first node comprising a debugging service component; step S3, adjusting the first image data based on the debugging service component; step S4, after the adjustment of the first image data is completed, closing the first node, and releasing the GPU resources occupied by the first node. The present invention can avoid continuous occupation of GPU resources, reduce the load of the intelligent computing center, enable the intelligent computing center to provide computing power services to more users, and realize the inclusive application of computing power.

Description

Technical Field

[0001] The present invention relates to the field of computing power infrastructure technology, and in particular to a method and device for developing an intelligent computing center model for universal computing power. Background Art

[0002] "Intelligent computing center" refers to a facility that provides the required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model reasoning) by using large-scale heterogeneous computing resources, including general computing power and intelligent computing power. The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from bottom-level computing power to top-level application enablement.

[0003] "Inclusive computing power" means providing appropriate and effective computing power services to all social classes and groups that need computing power services at an affordable cost based on the requirements of equal opportunity and the principle of commercial sustainability. To achieve inclusive computing power, on the one hand, it is necessary to strengthen the construction of computing power infrastructure to increase the total amount of computing power that can be provided, and on the other hand, it is necessary to improve the utilization rate of computing power, avoid the waste of computing power and reduce the cost of using computing power.

[0004] At present, during the model development process, testers are required to repeatedly adjust the code and use the artificial intelligence (AI) acceleration card of the intelligent computing center to provide computing power to run the adjusted code until the running results meet expectations. Since the model development process requires a large amount of computing power resources, the model development process is usually set up in the intelligent computing center. In the actual model development process, business personnel upload the data corresponding to the model to the intelligent computing center and repeatedly modify the data corresponding to the model. During the entire model development period, the GPU resources of the intelligent computing center need to be occupied for a long time, resulting in a high load on the intelligent computing center, making it difficult to achieve universal application of computing power. At the same time, since the GPU resources of the intelligent computing center need to be occupied for a long time during the model development period, business personnel need to rent the computing power services of the intelligent computing center for a long time, resulting in a high cost of model development.

[0005] It can be seen that in the relevant technologies, there are problems that model development leads to a high load on the intelligent computing center and high model development costs. Summary of the invention

[0006] The embodiments of the present invention provide a method and device for developing an intelligent computing center model for universal computing power, so as to solve the problems in related technologies that model development leads to a high load on the intelligent computing center and a high model development cost.

[0007] To solve the above problems, the present invention is achieved as follows:

[0008] In a first aspect, an embodiment of the present invention provides a method for developing an intelligent computing center model for inclusive computing power, including:

[0009] Step S1, receiving first image data of a model to be developed;

[0010] Step S2: acquiring a first node, where the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component;

[0011] Step S3: adjusting the first image data based on the debugging service component;

[0012] Step S4: After the first mirror data is adjusted, shut down the first node and release the GPU resources occupied by the first node.

[0013] In one embodiment, step S1 includes:

[0014] Step S101: receiving code data of the model to be developed and environment information for running the code data sent by a terminal;

[0015] Step S102: creating a second node based on the central processing unit CPU resources of the intelligent computing center and the environmental information;

[0016] Step S103: Generate the first mirror data based on the code data through the second node;

[0017] Step S104: shut down the second node and release the CPU resources occupied by the second node.

[0018] In one embodiment, step S3 includes:

[0019] Step S301: receiving breakpoint debugging information sent by a terminal through a port, wherein the port is a port corresponding to the debugging service component and the port is used for external access to the debugging service component;

[0020] Step S302: adjusting the first mirror data based on the breakpoint debugging information through the first node;

[0021] Step S303: Run the adjusted first mirror data through the first node to obtain a running result;

[0022] Step S304: Send the operation result to the terminal through the first node.

[0023] In one embodiment, step S1 includes:

[0024] Step S111: creating a third node based on the central processing unit CPU resources of the intelligent computing center;

[0025] Step S112: establishing a communication connection between the third node and the terminal;

[0026] Step S113: receiving the initial code data sent by the terminal through the third node;

[0027] Step S114: Generate the first mirror data based on the initial code data through the third node.

[0028] In one embodiment, step S3 includes:

[0029] Step S311: receiving a code modification instruction sent by the terminal through the third node;

[0030] Step S312: Generate breakpoint debugging information based on the code modification instruction through the third node;

[0031] Step S313: the third node sends the breakpoint debugging information to the first node through a port, where the port is a port corresponding to the debugging service component and is used for external access to the debugging service component;

[0032] Step S314: Adjust the first mirror data based on the breakpoint debugging information through the first node.

[0033] In one embodiment, step S2 includes:

[0034] Step S21: when the intelligent computing center stores a first node corresponding to the model to be developed, start the first node;

[0035] Step S22: When the intelligent computing center does not store the first node corresponding to the model to be developed, create the first node based on the GPU resources of the intelligent computing center, and create the debugging service component in the first node.

[0036] In a second aspect, an embodiment of the present invention further provides an intelligent computing center model development device for inclusive computing power, including:

[0037] A receiving module, used for receiving first image data of a model to be developed;

[0038] An acquisition module, configured to acquire a first node, wherein the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component;

[0039] An adjustment module, configured to adjust the first image data based on the debugging service component;

[0040] The closing module is used to close the first node and release the GPU resources occupied by the first node after the adjustment of the first mirror data is completed.

[0041] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps in the method for developing an intelligent computing center model for inclusive computing power as described in the first aspect above are implemented.

[0042] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the method for developing an intelligent computing center model for inclusive computing power as described in the first aspect above are implemented.

[0043] In a fifth aspect, an embodiment of the present invention further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps in the method for developing an intelligent computing center model for inclusive computing power as described in the first aspect above.

[0044] In an embodiment of the present invention, the first mirror data of the model to be developed is received; the first node is obtained, the first node is a node occupying the GPU resources of the graphics processor of the intelligent computing center, and the first node includes a debugging service component; the debugging service component is used to adjust the first mirror data; after the adjustment of the first mirror data is completed, the first node is closed, and the GPU resources occupied by the first node are released. In this way, in the process of model development, the first mirror data is adjusted by the debugging service component of the first node, and after the adjustment is completed, the first node is closed and the GPU resources occupied by the first node are released, so as to avoid the continuous occupation of GPU resources, reduce the load of the intelligent computing center, and enable the intelligent computing center to provide computing power services to more users, realizing the universal application of computing power. At the same time, after the adjustment is completed, the first node is closed and the GPU resources occupied by the first node are released, so that the computing power service of the intelligent computing center does not need to be rented for a long time during the entire model development process, thereby greatly reducing the cost of model development. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0046] Figure 1 It is a flow chart of a method for developing an intelligent computing center model for inclusive computing power provided by an embodiment of the present invention;

[0047] Figure 2 is an interactive flow chart of a local encoding method provided by an embodiment of the present invention;

[0048] Figure 3 is an interactive flow chart of a remote encoding method provided by an embodiment of the present invention;

[0049] Figure 4 It is a structural diagram of an intelligent computing center model development device for inclusive computing power provided by an embodiment of the present invention;

[0050] Figure 5 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The "computing power" mentioned in the present invention is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to execute certain computing requirements. It is the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to the society through computing power infrastructure.

[0053] The "computing power" (Computational Power, CP) mentioned in the present invention is the ability of a data center server to process data and output results. It is a comprehensive indicator to measure the computing power of a data center, including general computing power, super computing power and intelligent computing power. The commonly used unit of measurement is the number of floating point operations per second (FLOPS: Floating Point Operations Per Second, 1EFLOPS=10^18FLOPS). The larger the value, the stronger the comprehensive computing power. According to calculations, 1EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream notebooks. The calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 .

[0054] The "Network Power" (NP) described in the present invention is a manifestation of the data transmission capability of computing facilities, including comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, etc. The carrying capacity involves network transmission within and between data centers, and is a comprehensive indicator for measuring network transmission scheduling capabilities. In the embodiment of the present invention, the carrying capacity uses the memory bandwidth.

[0055] The "Storage Power" (SP) described in the present invention is the comprehensive ability of a data center in terms of data storage capacity, performance, safety and reliability, and green and low-carbon. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and server built-in storage devices. The commonly used unit of measurement for storage capacity is exabyte (EB, 1EB=2^60bytes), and the commonly used unit of measurement for performance is the number of reads and writes per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB). The disaster recovery ratio is an important manifestation of safety and reliability.

[0056] The "computing power infrastructure" described in the present invention is a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity. It can realize centralized computing, storage, transmission and application of information, and presents characteristics such as diversity and ubiquity, intelligence and agility, security and reliability, and green and low-carbon.

[0057] The “computing power” mentioned in the present invention includes general computing power, intelligent computing power and super computing power.

[0058] The “general computing power” mentioned in the present invention refers to the computing power provided by servers based on central processing unit (CPU) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0059] The "intelligent computing power" described in the present invention is a computing platform for large-scale deployment of special chips such as graphics processing units (GPUs), field programmable gate arrays (FPGAs), and application-specific integrated circuits (ASICs) for various innovative artificial intelligence applications, such as natural language processing and machine vision.

[0060] The "super computing power" mentioned in the present invention is 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 uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for calculations in cutting-edge scientific fields, such as planetary simulation, drug molecule design, and gene analysis.

[0061] The "intelligent computing center" mentioned in the present invention refers to a facility that provides the required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model reasoning) by using large-scale heterogeneous computing resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from bottom-level computing power to top-level application enablement.

[0062] The "computing resources" mentioned in the present invention refer to the technologies and facilities with information calculation, 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 supporting and guarantee resources such as wind, fire, water and electricity.

[0063] The “model” mentioned in the present invention includes but is not limited to “large language model” and “multimodal large model”.

[0064] The “large language model” mentioned in the present invention refers to a large language model (LLM), which is a language model with a large parameter scale, designed to understand and generate human language, trained with a large amount of text data, and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0065] The "Multimodal Large Models" mentioned in the present invention refer to models that combine multimodal information such as text, images, videos, audio, etc. for training, including but not limited to multimodal large language models.

[0066] The "inclusive computing power" mentioned in the present invention refers to providing appropriate and effective computing power services at an affordable cost to all social classes and groups that have computing power service needs based on the requirements of equal opportunity and the principle of commercial sustainability.

[0067] The "mirror data" described in the present invention refers to the copy data generated by copying, compressing and format conversion of all or part of the code, and all or part of the code can be obtained by parsing the mirror data. In the present invention, after the intelligent computing center receives the first mirror data, it can parse the first mirror data to obtain the code data corresponding to the model to be developed.

[0068] See also Figure 1 , Figure 1 is a flow chart of a method for developing an intelligent computing center model for inclusive computing power provided by an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0069] Step S1, receiving first image data of a model to be developed.

[0070] The first mirror data may be mirror data generated based on complete code or partial code, wherein the code is written or adjusted by business personnel for specific business and can be flexibly adjusted according to the progress of model development.

[0071] In some implementations, when the intelligent computing center receives the first mirror data for the first time, the first mirror data is full data, that is, the first mirror data is mirror data generated based on all codes of the current model.

[0072] In some embodiments, when the intelligent computing center receives the first mirror data not for the first time, the first mirror data is incremental data, that is, the first mirror data is the difference between the entire code based on the current model and the entire code of the model before adjustment. The first node can obtain the complete first mirror data by receiving the first mirror data multiple times.

[0073] For example, the first mirror data is incremental data. Before the first node receives the first mirror data, it also receives the second mirror data. The second mirror data is full data. The complete mirror data can be obtained by combining the first mirror data and the second mirror data. In this way, the complete first mirror data can be obtained only by synchronizing the incremental data part, thereby improving the synchronization efficiency.

[0074] Step S2: Obtain a first node, where the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component.

[0075] The first node is a node (pod) in the container of the intelligent computing center. Different functional components can be preset in the first node to achieve different functions, such as adjusting the model, running the model, etc.

[0076] In an embodiment of the present invention, the first node is used to run the first mirror data of the model to be developed and generate a running result, which can reflect the relevant situation of the model to be developed, such as the accuracy and loss value of the model prediction. Among them, the first node includes a debugging service component, and the model is adjusted by the debugging service component. Optionally, the debugging service component can be DebugServer.

[0077] In one embodiment, step S2 includes:

[0078] Step S21: when the intelligent computing center stores a first node corresponding to the model to be developed, start the first node;

[0079] Step S22: When the intelligent computing center does not store the first node corresponding to the model to be developed, create the first node based on the GPU resources of the intelligent computing center, and create the debugging service component in the first node.

[0080] It should be noted that the model to be developed may have multiple development stages, and there will be a long period of time between different development stages when the first node does not need to be debugged. During this period, the first node is only retained in the intelligent computing center, but it does not need to be run. When the model to be developed needs to be debugged again, the first node is re-enabled to reduce the GPU resource usage of the intelligent computing center. Therefore, when the first node corresponding to the model to be developed is stored in the intelligent computing center, the first node is directly enabled, which can achieve rapid activation and reduce resource usage during non-use.

[0081] At the same time, when debugging the model to be developed for the first time, the intelligent computing center does not store the first node corresponding to the model to be developed, so a new first node needs to be established to complete the debugging of the model to be developed through the first node. Furthermore, when the current development stage of the model to be developed is completed and there is a new development stage in the future, the first node can be closed and retained to release the GPU resources occupied by the first node; when there is no new development stage, the first node can be directly closed and deleted.

[0082] In some implementations, after the first node completes the adjustment of the model to be developed, the first node is closed and the GPU resources occupied by the first node are released, the first node is retained, and if the first node is not enabled or does not receive a postponement instruction within a preset time period, the first node is deleted; further, if a postponement instruction is received, the time is recalculated, and if the first node is not enabled or does not receive a postponement instruction within the preset time period, the first node is deleted. In this way, by setting the preset time, it is prevented that the developer forgets to delete the first node, causing the first node to occupy the storage resources of the intelligent computing center.

[0083] Step S3: adjusting the first image data based on the debugging service component.

[0084] The above adjustment of the first mirror data by the debugging service component may be adjusting the first mirror data by receiving an external adjustment modification instruction, or may be adjusting the first mirror data by receiving a modification instruction sent by other nodes, as described in detail in subsequent embodiments.

[0085] Step S4: After the first mirror data is adjusted, shut down the first node and release the GPU resources occupied by the first node.

[0086] It should be noted that after the first mirror data is adjusted, there is no modification instruction that needs to adjust the first mirror data temporarily. At this time, the operation of the first node will still occupy GPU resources, resulting in a waste of computing power. In an embodiment of the present invention, after the first mirror data is adjusted, the first node is shut down, and the GPU resources occupied by the first node are not used, so that other users can occupy the released GPU resources to perform tasks such as model development, thereby improving the utilization rate of GPU resources.

[0087] In an embodiment of the present invention, the first mirror data of the model to be developed is received; the first node is obtained, the first node is a node occupying the GPU resources of the graphics processor of the intelligent computing center, and the first node includes a debugging service component; the debugging service component is used to adjust the first mirror data; after the adjustment of the first mirror data is completed, the first node is closed, and the GPU resources occupied by the first node are released. In this way, in the process of model development, the first mirror data is adjusted by the debugging service component of the first node, and after the adjustment is completed, the first node is closed and the GPU resources occupied by the first node are released, so as to avoid the continuous occupation of GPU resources, reduce the load of the intelligent computing center, and enable the intelligent computing center to provide computing power services to more users, realizing the universal application of computing power. At the same time, after the adjustment is completed, the first node is closed and the GPU resources occupied by the first node are released, so that the computing power service of the intelligent computing center does not need to be rented for a long time during the entire model development process, thereby greatly reducing the cost of model development.

[0088] It should be noted that in the process of debugging the model to be developed, the code of the model to be developed can be completed directly on the terminal, and then the complete code data can be debugged (i.e. local coding); or the terminal can be remotely connected to the intelligent computing center, and the business personnel can adjust the code in real time, and then debug the adjusted code data (i.e. remote coding). Different methods need to be processed separately to achieve that both methods can reduce the GPU resource usage of the intelligent computing center.

[0089] Specifically, when the local encoding method is adopted, the step S1 includes:

[0090] Step S101: receiving code data of the model to be developed and environment information for running the code data sent by a terminal;

[0091] Step S102: creating a second node based on the central processing unit CPU resources of the intelligent computing center and the environmental information;

[0092] Step S103: Generate the first mirror data based on the code data through the second node;

[0093] Step S104: shut down the second node and release the CPU resources occupied by the second node.

[0094] In the embodiment of the present invention, the code data of the model to be developed and the environment information for running the code data sent by the receiving terminal are used; a second node is created based on the CPU resources and environment information of the central processing unit of the intelligent computing center; the first mirror data is generated based on the code data through the second node; the second node is closed, and the CPU resources occupied by the second node are released. In this way, the first mirror data is generated through the second node, and the operating environment of the first mirror data is consistent with the local operating environment of the terminal, so that the first node can be consistent with the local operating environment of the terminal when running the first mirror data.

[0095] It should be noted that if Figure 2 As shown, if the local operating environment of the terminal is consistent with the operating environment of the intelligent computing center, there is no need to create a second node through CPU resources. Instead, the first mirror data is directly generated based on the complete code data locally in the terminal and then sent to the first node. Since the operating environment of the first node is consistent with the local operating environment of the terminal in this case, the first node can directly run the first mirror data and adjust the first mirror data.

[0096] In one embodiment, step S3 includes:

[0097] Step S301: receiving breakpoint debugging information sent by a terminal through a port, wherein the port is a port corresponding to the debugging service component and the port is used for external access to the debugging service component;

[0098] Step S302: adjusting the first mirror data based on the breakpoint debugging information through the first node;

[0099] Step S303: Run the adjusted first mirror data through the first node to obtain a running result;

[0100] Step S304: Send the operation result to the terminal through the first node.

[0101] In an embodiment of the present invention, the breakpoint debugging information sent by the terminal is received through a port, the port is a port corresponding to the debugging service component, and the port is used to externally access the debugging service component; the first mirror data is adjusted based on the breakpoint debugging information through the first node; the adjusted first mirror data is run through the first node to obtain an operation result; and the operation result is sent to the terminal through the first node. In this way, the interaction between the terminal and the first node is realized through the port, and the terminal can debug the first mirror data by sending the breakpoint debugging information.

[0102] For details, see Figure 2When the model to be developed needs to be debugged, the first node is started in the intelligent computing center, and the local code of the terminal is synchronized to the first node (i.e. Figure 2 1. Start the first node and synchronize the code); start the debugging service component DebugServer in the first node (that is, Figure 2 2. Start the debugging service component in the terminal; the terminal communicates with the debugging service component through the port (i.e. Figure 2 3. Connecting and debugging service components), adjusting the first mirror data through breakpoint debugging information, so that the first node can run the adjusted first mirror data to obtain the running result.

[0103] The above breakpoint debugging information is generated by the terminal by modifying the configuration of the local code component (VSCode Debug). Breakpoint debugging is achieved by modifying the configuration of the VSCode Debug component.

[0104] The above operation result is the result obtained by the first node running the adjusted first mirror data. The first node is configured with different service components, and different operation results, such as accuracy and loss value, can be generated through different service components. After obtaining the operation result, the first node sends the operation result to the terminal, so that the business personnel can make further adjustments to the model to be developed according to the operation result.

[0105] In the case of adopting the remote encoding method, the step S1 includes:

[0106] Step S111: creating a third node based on the central processing unit CPU resources of the intelligent computing center;

[0107] Step S112: establishing a communication connection between the third node and the terminal;

[0108] Step S113: receiving the initial code data sent by the terminal through the third node;

[0109] Step S114: Generate the first mirror data based on the initial code data through the third node.

[0110] In an embodiment of the present invention, a third node is created based on the CPU resources of the intelligent computing center; a communication connection is established between the third node and the terminal; the initial code data sent by the terminal is received through the third node; and the first mirror data is generated through the third node based on the initial code data. In this way, through the communication connection between the third node and the terminal, business personnel can implement remote coding in the third node without occupying the GPU resources of the intelligent computing center; after receiving the initial code data, the first mirror data is generated based on the initial code data to implement the code update of the model to be developed.

[0111] The operating environment of the third node is the same as the local operating environment of the terminal, so as to facilitate modification of the code of the model to be developed.

[0112] like Figure 3 As shown, in the case of remote coding, a third node is first created based on the CPU resources of the intelligent computing center, and a communication connection is established between the third node and the terminal (i.e. Figure 3 1. Connecting to a third node), so that the terminal can send relevant instructions for modifying the code to the third node, thereby realizing remote coding of the code of the model to be developed.

[0113] It should be noted that, when remote coding is adopted, the third node is used to update the code, so there is no need to shut down the third node; after debugging is completed, the first node and the third node are shut down at the same time, and the GPU resources occupied by the first node and the CPU resources occupied by the third node are released.

[0114] Specifically, step S3 includes:

[0115] Step S311: receiving a code modification instruction sent by the terminal through the third node;

[0116] Step S312: Generate breakpoint debugging information based on the code modification instruction through the third node;

[0117] Step S313: the third node sends the breakpoint debugging information to the first node through a port, where the port is a port corresponding to the debugging service component and is used for external access to the debugging service component;

[0118] Step S314: Adjust the first mirror data based on the breakpoint debugging information through the first node.

[0119] In an embodiment of the present invention, the code modification instruction sent by the terminal is received by the third node; breakpoint debugging information is generated by the third node based on the code modification instruction; the third node sends the breakpoint debugging information to the first node through a port, the port is a port corresponding to the debugging service component, and the port is used for external access to the debugging service component; the first mirror data is adjusted by the first node based on the breakpoint debugging information. In this way, after the third node receives the code modification instruction sent by the terminal, breakpoint debugging information is generated, and the breakpoint debugging information is sent to the debugging service component of the first node through the port to achieve the adjustment of the model to be developed.

[0120] Furthermore, if Figure 3 As shown, before debugging the model to be developed, the third node starts the VSCodeServer component (i.e. Figure 3 2. Start code component in the code to generate breakpoint debugging information; when starting to debug the model to be developed, start the first node and synchronize the code (i.e. Figure 3 3. Start the first node and synchronize the code), and start the debugging service component DebugServer of the first node (that is, Figure 3 4. Start the debugging service component in the third node), the third node connects to the debugging service component of the first node through the port (i.e. Figure 3 5. Connecting the debugging service component) to synchronize the first mirror data to the first node, and sending the breakpoint debugging information to the debugging service component to adjust the first mirror data corresponding to the model to be developed.

[0121] See also Figure 4 , Figure 4 is a structural diagram of an intelligent computing center model development device for inclusive computing power provided by an embodiment of the present invention, such as Figure 4 As shown, the intelligent computing center model development device 400 for inclusive computing power includes:

[0122] A receiving module 401 is used to receive first image data of a model to be developed;

[0123] An acquisition module 402 is used to acquire a first node, where the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component;

[0124] An adjustment module 403, configured to adjust the first image data based on the debugging service component;

[0125] The closing module 404 is used to close the first node and release the GPU resources occupied by the first node after the adjustment of the first mirror data is completed.

[0126] In one embodiment, the receiving module 401 includes:

[0127] A first receiving unit, configured to receive code data of the model to be developed and environment information for running the code data sent by a terminal;

[0128] A first creation unit, configured to create a second node based on the central processing unit CPU resources of the intelligent computing center and the environmental information;

[0129] A first generating unit, configured to generate the first mirror data based on the code data through the second node;

[0130] The shutting down unit is used to shut down the second node and release the CPU resources occupied by the second node.

[0131] In one embodiment, the adjustment module 403 includes:

[0132] A second receiving unit, configured to receive breakpoint debugging information sent by a terminal through a port, wherein the port is a port corresponding to the debugging service component, and the port is used for external access to the debugging service component;

[0133] A first adjustment unit, configured to adjust the first mirror data based on the breakpoint debugging information through the first node;

[0134] an operating unit, configured to operate the adjusted first mirror data through the first node to obtain an operating result;

[0135] A sending unit is used to send the operation result to the terminal through the first node.

[0136] In one embodiment, the receiving module 401 includes:

[0137] A second creation unit is used to create a third node based on the central processing unit CPU resources of the intelligent computing center;

[0138] An establishing unit, configured to establish a communication connection between the third node and the terminal;

[0139] A third receiving unit, configured to receive the initial code data sent by the terminal through the third node;

[0140] The second generating unit is configured to generate the first mirror data based on the initial code data through the third node.

[0141] In one embodiment, the adjustment module 403 includes:

[0142] a fourth receiving unit, configured to receive, through the third node, a code modification instruction sent by the terminal;

[0143] A third generating unit, configured to generate breakpoint debugging information based on the code modification instruction through the third node;

[0144] A sending unit, configured for the third node to send the breakpoint debugging information to the first node through a port, wherein the port is a port corresponding to the debugging service component, and the port is used for external access to the debugging service component;

[0145] The second adjusting unit is configured to adjust the first mirror data based on the breakpoint debugging information through the first node.

[0146] In one embodiment, the acquisition module 402 includes:

[0147] A starting unit, configured to start the first node when the first node corresponding to the model to be developed is stored in the intelligent computing center;

[0148] The third creation unit is used to create the first node based on the GPU resources of the intelligent computing center when the intelligent computing center does not store the first node corresponding to the model to be developed, and to create the debugging service component in the first node.

[0149] The intelligent computing center model development device for inclusive computing power provided in the embodiment of the present invention is capable of implementing the various processes of the various embodiments of the above-mentioned intelligent computing center model development method for inclusive computing power. The technical features correspond one to one and can achieve the same technical effects. To avoid repetition, they will not be described here.

[0150] It should be noted that the intelligent computing center model development device for inclusive computing power in the embodiment of the present invention can be a device, or a component, integrated circuit, or chip in an electronic device.

[0151] The embodiment of the present invention further provides an electronic device, see Figure 5 , Figure 5 is a schematic diagram of the structure of an electronic device provided by the present invention, the electronic device includes a memory 501, a processor 502 and a program or instruction stored in the memory 501 and running on the memory 501. When the program or instruction is executed by the processor 502, Figure 1 Any steps in the corresponding embodiment of the intelligent computing center model development method for inclusive computing power and the same beneficial effects achieved will not be repeated here.

[0152] The processor 502 may be a CPU, an ASIC, an FPGA or a GPU.

[0153] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned embodiment of the intelligent computing center model development method for inclusive computing power can be completed through hardware related to program instructions, and the program can be stored in a readable medium.

[0154] The embodiment of the present invention further provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above Figure 1 Any steps in the corresponding embodiments of the method for developing a model of an intelligent computing center for inclusive computing power can achieve the same technical effect, and will not be repeated here to avoid repetition. The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0155] The terms "first", "second" etc. in the embodiments of the present invention are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. In addition, the terms "include" and "have" and any variation thereof are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment. In addition, "and / or" is used in the present application to represent at least one of the connected objects, such as A and / or B and / or C, indicating that A alone, B alone, C alone, and A and B all exist, B and C all exist, A and C all exist, and 7 situations in which A, B and C all exist.

[0156] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0157] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a second terminal device, etc.) to execute the methods of each embodiment of the present application.

[0158] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for developing an intelligent computing center model for inclusive computing power, characterized in that: include: Step S1, receiving first image data of a model to be developed; Step S2: acquiring a first node, where the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component; Step S3: adjusting the first image data based on the debugging service component; Step S4: after the first mirror data adjustment is completed, shut down the first node and release the GPU resources occupied by the first node; The step S1 comprises: Step S111: creating a third node based on the central processing unit CPU resources of the intelligent computing center; Step S112: establishing a communication connection between the third node and the terminal; Step S113: receiving the initial code data sent by the terminal through the third node; Step S114: Generate the first mirror data based on the initial code data through the third node; The step S3 comprises: Step S311: receiving a code modification instruction sent by the terminal through the third node; Step S312: Generate breakpoint debugging information based on the code modification instruction through the third node; Step S313: the third node sends the breakpoint debugging information to the first node through a port, where the port is a port corresponding to the debugging service component and is used for external access to the debugging service component; Step S314: Adjust the first mirror data based on the breakpoint debugging information through the first node.

2. The method according to claim 1, characterized in that The step S1 comprises: Step S101: receiving code data of the model to be developed and environment information for running the code data sent by a terminal; Step S102: creating a second node based on the central processing unit CPU resources of the intelligent computing center and the environmental information; Step S103: Generate the first mirror data based on the code data through the second node; Step S104: shut down the second node and release the CPU resources occupied by the second node.

3. The method according to claim 2, characterized in that The step S3 comprises: Step S301: receiving breakpoint debugging information sent by a terminal through a port, wherein the port is a port corresponding to the debugging service component and the port is used for external access to the debugging service component; Step S302: adjusting the first mirror data based on the breakpoint debugging information through the first node; Step S303: Run the adjusted first mirror data through the first node to obtain an operation result; Step S304: Send the operation result to the terminal through the first node.

4. The method according to any one of claims 1 to 3, characterized in that The step S2 comprises: Step S21: when the intelligent computing center stores a first node corresponding to the model to be developed, start the first node; Step S22: When the intelligent computing center does not store the first node corresponding to the model to be developed, create the first node based on the GPU resources of the intelligent computing center, and create the debugging service component in the first node.

5. An intelligent computing center model development device for universal computing power, characterized in that: include: A receiving module, used for receiving first image data of a model to be developed; An acquisition module, configured to acquire a first node, wherein the first node is a node occupying a graphics processor GPU resource of an intelligent computing center, and the first node includes a debugging service component; An adjustment module, configured to adjust the first image data based on the debugging service component; A closing module, used to close the first node and release the GPU resources occupied by the first node after the adjustment of the first mirror data is completed; The receiving module comprises: A first receiving unit, configured to receive code data of the model to be developed and environment information for running the code data sent by a terminal; A first creation unit, configured to create a second node based on the central processing unit CPU resources of the intelligent computing center and the environmental information; A first generating unit, configured to generate the first mirror data based on the code data through the second node; A shutting down unit, used to shut down the second node and release the CPU resources occupied by the second node; The adjustment module comprises: A second receiving unit, configured to receive breakpoint debugging information sent by a terminal through a port, wherein the port is a port corresponding to the debugging service component, and the port is used for external access to the debugging service component; A first adjustment unit, configured to adjust the first mirror data based on the breakpoint debugging information through the first node; an operating unit, configured to operate the adjusted first mirror data through the first node to obtain an operating result; A sending unit is used to send the operation result to the terminal through the first node.

6. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method for developing an intelligent computing center model for inclusive computing power as described in any one of claims 1 to 4.

7. 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 developing an intelligent computing center model for inclusive computing power as described in any one of claims 1 to 4.

8. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the intelligent computing center model development method for universal computing power as described in any one of claims 1 to 4.

Images