Intelligent computing platform deployment and management method, system and equipment based on multiple architectures and medium

Abstract

The invention provides an intelligent computing platform deployment and management method, system and device based on multiple architectures and a medium, and the method comprises the steps that a plurality of heterogeneous computing units configured by a target host are recognized based on an aip-deploy tool, and each heterogeneous computing unit comprises an image processing unit and a neural network processing unit; a preset offline package is obtained, a Kubernetes cluster is established for each heterogeneous computing unit according to the offline package, deployment and configuration of specified components are performed on the Kubernetes cluster to generate an intelligent computing platform, and the intelligent computing platform is used for performing unified management on all the heterogeneous computing units in the Kubernetes cluster. According to the method, a kubernetes cluster is automatically deployed through an aip-deploy tool, one-key deployment and configuration of intelligent calculation related components are carried out on a target host, the method is suitable for off-line rapid framework of resource environments such as a GPU and an NPU, the time cost of deployment personnel is reduced, and errors are reduced; and heterogeneous computing units such as the GPU and the NPU are managed in a unified manner through the intelligent computing platform, so that various intelligent computing resources at the bottom layer are managed in a unified manner, resource waste is avoided, and the utilization rate of the resources is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent computing resource management, and in particular to an intelligent computing platform deployment and management method, system, device and medium based on multiple architectures. BACKGROUND

[0002] With the development of large language models and AI training and inference, the original data center needs to deploy a large number of intelligent computing related resources, including graphics processing units (GPUs), neural network processing units (NPUs), etc. In traditional intelligent computing platforms, different types of computing resources such as GPUs and NPUs need to be configured independently by humans, which is complex and has a long deployment cycle. Moreover, the traditional intelligent computing platform may not support heterogeneous environments well, and cannot handle different types of computing resources such as CPUs, GPUs, and NPUs simultaneously, resulting in the inability to manage different types of computing resources uniformly and poor resource utilization. SUMMARY

[0003] The embodiments of the present application provide an intelligent computing platform deployment and management method, system, device and medium based on multiple architectures to solve the problems of related technologies. The technical solutions are as follows:

[0004] In a first aspect, the embodiments of the present application provide an intelligent computing platform deployment and management method based on multiple architectures, comprising:

[0005] Identifying a plurality of heterogeneous computing units configured by a target host based on an aip-deploy tool, the heterogeneous computing units including image processing units and neural network processing units;

[0006] Obtaining a preset offline package, building a Kubernetes cluster for each heterogeneous computing unit according to the offline package, and deploying and configuring specified components of the Kubernetes cluster to generate an intelligent computing platform, the intelligent computing platform being used to uniformly manage all heterogeneous computing units in the Kubernetes cluster.

[0007] In an embodiment, the offline package includes a driver installation package for each heterogeneous computing unit, all software components required for building and managing the Kubernetes cluster, and dependencies required for maintaining the normal operation of the cluster components.

[0008] In an embodiment, the method further comprises:

[0009] Obtaining the driver and dependency components of each heterogeneous computing unit according to the offline package;

[0010] Installing the driver and dependency components in the target host through the aip-deploy tool.

[0011] In an embodiment, the Kubernetes cluster is built for each heterogeneous computing unit according to the offline package, including:

[0012] The Kubernetes components required for installing the Kubernetes cluster are installed according to the offline package, and the etcd node of the Kubernetes cluster and the control plane of the Kubernetes cluster are deployed to implement the building of the Kubernetes cluster.

[0013] In an embodiment, the method further includes:

[0014] Architecture information of the heterogeneous computing unit is obtained, node labels of the Kubernetes cluster are set according to the architecture information, and device plugins related to the heterogeneous computing unit are deployed according to the labels of each node, the device plugins being used to report resource information of each heterogeneous computing unit to the Kubernetes cluster.

[0015] In an embodiment, the method further includes:

[0016] According to the node labels of each node in the Kubernetes cluster, a monitoring data collector corresponding to the heterogeneous computing unit is deployed, the monitoring data collector being used to provide a monitoring service for the Kubernetes cluster to collect monitoring data of the heterogeneous computing unit.

[0017] In an embodiment, the method further includes:

[0018] A computing power task is obtained, the computing power task being obtained through a webpage interface provided by the intelligent computing platform; the Kubernetes cluster is accessed according to the computing power task, the computing power task is scheduled to a target node having required heterogeneous computing resources in the Kubernetes cluster, so that the Kubernetes cluster creates a container on the target node to run the computing power task.

[0019] In a second aspect, an embodiment of the present application provides a multi-architecture-based intelligent computing platform deployment and management system, which executes the multi-architecture-based intelligent computing platform deployment and management method as described above.

[0020] In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory and a processor. The memory and the processor communicate with each other through an internal connection path. The memory is used to store instructions, and the processor is used to execute the instructions stored in the memory. When the processor executes the instructions stored in the memory, the processor executes the method in any one of the embodiments of each aspect described above.

[0021] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium storing a computer program, when the computer program is run on a computer, the method in any one of the above aspects is executed.

[0022] The advantages or beneficial effects of the above technical solutions at least include:

[0023] The aip-deploy tool is used for automatically deploying the kubernetes cluster, one-key deployment and configuration of the intelligent computing related components are performed on the target host, and the resource environment such as GPU and NPU is suitable for offline rapid construction, time cost of deployment personnel is reduced, and errors are reduced.

[0024] The above summary is only for the purpose of the description and is not intended to limit in any way. In addition to the above described illustrative aspects, embodiments, and features, further aspects, embodiments, and features of the present application will be readily apparent from the drawings and detailed description below. BRIEF DESCRIPTION OF DRAWINGS

[0025] In the drawings, like reference numerals refer to same or similar functionalities throughout the several views. The drawings are not necessarily to scale. It is to be understood that these drawings only depict several embodiments of the disclosure and are not to be considered as limiting its scope.

[0026] Figure 1 A flowchart of a method for deploying and managing an intelligent computing platform based on multiple architectures according to the present application;

[0027] Figure 2 A schematic diagram of building a kubernetes cluster through aip-deploy according to the present application;

[0028] Figure 3 A schematic diagram of creating computing power tasks through an intelligent computing platform according to the present application;

[0029] Figure 4 An architecture schematic diagram of a deployment and management system of an intelligent computing platform based on multiple architectures according to the present application;

[0030] Figure 5 A structure block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In the following, only certain exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and description are considered to be exemplary in nature rather than limiting.

[0032] With the development of large language models and AI training inference, the original data center needs to deploy a large number of intelligent computing related resources, including graphics processing units (GPUs), neural network processing units (NPUs), etc. In traditional intelligent computing platforms, different types of computing resources such as GPUs and NPUs need to be configured independently by manual configuration, which has high complexity and long deployment period. Moreover, the traditional intelligent computing platform may not support heterogeneous environments, and cannot handle different types of computing resources such as CPUs, GPUs, and NPUs simultaneously, resulting in the inability to manage different types of computing resources uniformly and leading to poor resource utilization.

[0033] To solve the above problems, the embodiments of the present application provide a multi-architecture-based intelligent computing platform deployment and management method, which can deploy kubernetes-related components with one key and manage GPUs and NPUs through a unified platform to meet various intelligent scheduling needs.

[0034] As shown in Figures 1-3 The multi-architecture-based intelligent computing platform deployment and management method of the present embodiment specifically includes the following steps:

[0035] Step S1: Identify a plurality of heterogeneous computing units configured by a target host based on an aip-deploy tool, wherein the heterogeneous computing units include image processing units and neural network processing units.

[0036] It should be noted that the aip-deploy tool is a tool for automatically deploying and managing AI applications, which can help users automatically obtain GPU information and corresponding architectures of a host, and automatically install GPU drivers and related dependencies.

[0037] The aip-deploy tool first identifies the heterogeneous computing units installed in the host, such as image processing units (GPUs) and neural network processing units (NPUs). The identification of the heterogeneous computing units can be achieved by querying the PCI device list of the system, because GPUs and NPUs are hardware devices that communicate with the system through PCI interfaces.

[0038] Further, after identifying the heterogeneous computing units, the aip-deploy tool further obtains specific information of the heterogeneous computing units, including model and architecture, by reading the information of the PCI configuration space.

[0039] According to the obtained model architecture information, the aip-deploy tool determines the driver version of the heterogeneous computing unit to be installed and other related dependencies such as CUDA, and automatically downloads and installs these components through the aip-deploy tool.

[0040] Through the above steps, the aip-deploy tool can automatically configure the required GPU / NPU environment for the host, thereby simplifying the deployment process of the AI application.

[0041] It should be noted that the drivers and dependencies called by the aip-deploy tool are pre-stored in the offline package. Therefore, an offline package containing the required components needs to be pre-built, which includes the driver installation package of each heterogeneous computing unit, all software components required for building and managing the Kubernetes cluster, and the dependencies required for maintaining the normal operation of the cluster components. Specifically, the components included in the offline package include:

[0042] GPU / NPU installation package: GPU / NPU driver adapted to your operating system;

[0043] Cluster components: various components required by Kubernetes;

[0044] Related dependency package: other software packages and libraries required for cluster operation.

[0045] After identifying the model and architecture of the heterogeneous computing unit, the driver and dependency components of each heterogeneous computing unit in the offline package are obtained, and the aip-deploy tool is used to install the driver and dependency components in the target host.

[0046] Step S2: Obtain the pre-set offline package, build a Kubernetes cluster for each heterogeneous computing unit according to the offline package, and deploy and configure the specified components of the Kubernetes cluster to generate an intelligent computing platform. The intelligent computing platform is used to uniformly manage all heterogeneous computing units in the Kubernetes cluster.

[0047] Install the Kubernetes components required by the Kubernetes cluster from the offline package, and deploy the etcd node of the Kubernetes cluster and the control plane of the Kubernetes cluster to implement the building of the Kubernetes cluster.

[0048] It should be noted that the Kubernetes components mainly include the following parts, which together constitute the core functions of the Kubernetes cluster:

[0049] kubeadm: instructions for initializing the cluster.

[0050] kubelet: A component that runs on each node in the cluster to launch Pods and containers, etc.

[0051] kubectl: A command-line tool used to communicate with the cluster.

[0052] The roles of these components are:

[0053] kubeadm: Helps users initialize a Kubernetes cluster, including creating control plane nodes and joining worker nodes to the cluster.

[0054] kubelet: Runs on each node, responsible for launching containers, monitoring the health of containers, and reporting the status of nodes and Pods to the control plane.

[0055] kubectl: Provides a command-line interface for users to deploy and manage resources in the Kubernetes cluster.

[0056] etcd is a distributed key-value storage system that stores the configuration data, state, and metadata of the Kubernetes cluster. The steps to deploy etcd nodes are as follows:

[0057] Initialize etcd: Deploy etcd instances on control plane nodes, which will serve as the central storage of cluster data.

[0058] Configure etcd: Specify the configuration of etcd in kubeadminit, such as data directory and network settings.

[0059] Start etcd: Start the etcd service through the kubeadminit command, which will communicate with the API Server and save the state of the cluster.

[0060] The control plane is the brain of Kubernetes, responsible for managing and coordinating the entire cluster. The main components of the control plane include:

[0061] API Server (kube-apiserver): Handles all API requests, performs verification, authorization, and data persistence, and interacts with etcd to save the configuration data and state of the cluster.

[0062] Scheduler (kube-scheduler): Responsible for assigning newly created Pods to appropriate nodes.

[0063] Controller Manager (kube-controller-manager): Manages various controllers, monitors the state of the cluster, and ensures that the actual state is consistent with the desired state.

[0064] The steps to deploy the control plane are as follows:

[0065] Initialize the control plane node: Use the kubeadminit command to initialize the control plane node, which will start the APIServer, scheduler, and controller manager.

[0066] Set up kubectl configuration: Copy the administrator configuration of the cluster to $HOME / .kube / config to interact with the cluster using kubectl.

[0067] Deploy network plugins: Choose and deploy network plugins to enable communication between Pods.

[0068] Join the worker nodes: Use the kubeadm join command to join the worker nodes to the cluster.

[0069] Through the above steps, the components required for installing the Kubernetes cluster are installed, and the etcd node and control plane of the cluster are deployed, thus building a complete Kubernetes cluster.

[0070] Further, after building the Kubernetes cluster, obtain the architecture information of the heterogeneous computing unit, which includes the type and architecture of the heterogeneous computing unit; according to the architecture information, label the nodes of the Kubernetes cluster, and give different node labels to each type of node. Node labels are used to identify specific attributes or capabilities of nodes, such as GPU model or NPU type.

[0071] According to different node labels, deploy device plugins related to heterogeneous computing units (such as nvidia-device-plugin plugin for GPU and ascend-device-plugin for NPU), which are used to report resource information of each heterogeneous computing unit to the Kubernetes cluster, including the number and model of resources, so that they can be recognized and used by the scheduler, thus being controlled by Kubernetes.

[0072] In addition, after building the Kubernetes cluster, generate the intelligent computing platform, which can provide a web interface for accessing the intelligent computing platform. The intelligent computing platform realizes the unified management and scheduling of heterogeneous computing resources, including GPU, NPU, etc., through Kubernetes, while providing performance optimization, cost control, and user experience improvement. The Kubernetes cluster extends its application in the AI field through the intelligent computing platform, achieving more efficient resource management and task scheduling.

[0073] Further, according to the node labels of each node in the Kubernetes cluster, deploy the monitoring data collectors (such as GPU: dcgm-export, NPU: npu-export, etc.) corresponding to the heterogeneous computing units, which are used to provide monitoring services for the Kubernetes cluster to collect and uniformly manage GPU and NPU resource data from different nodes, and realize all-around monitoring and optimized scheduling of resources.

[0074] After the Kubernetes cluster and the intelligent computing platform have been built and delivered for use, if the business demand increases and more computing resources are needed to support more workloads or improve processing capacity, the cluster resources can be expanded through the aip-deploy tool, so that the Kubernetes cluster can dynamically adjust the scale according to the business demand to support the changing workloads.

[0075] Further, after the Kubernetes cluster and the intelligent computing platform have been built and delivered for use, as shown in Figure 3 , the user can log in to the Web interface of the intelligent computing platform and use the tools and functions provided on the Web interface to view, configure and manage cluster resources, and also can perform operations such as cluster upgrade, node expansion, cluster health monitoring, etc. Specifically:

[0076] The AI developer uploads datasets, uploads or develops algorithms and performs corresponding resource management through the intelligent computing platform, so that the AI developer can conveniently access and use the resources of the intelligent computing platform to develop and test AI algorithms.

[0077] The AI application applies for intelligent computing resources on the intelligent computing platform, selects corresponding algorithms, datasets and corresponding computing power resources to create a training task.

[0078] The specific method is: the AI application applies for the required intelligent computing resources through the Web interface, selects appropriate algorithms and datasets to set the training task, configures computing power resources such as GPU or NPU, and submits the training task. The intelligent computing platform schedules the task to the corresponding resource pool and creates a container to run the AI training task. The AI application allows the AI application to train the AI model using the resources of the intelligent computing platform, improves the training efficiency and effect, and accelerates the development and iteration of the AI model.

[0079] The AI application applies for intelligent computing resources on the intelligent computing platform, selects corresponding AI models and corresponding computing power resources to create an inference task.

[0080] The specific method is that the AI applicant applies for the required intelligent computing resources through the Web interface, selects a pre-trained AI model, configures the required computing resources, submits the inference task, the platform schedules the task to the corresponding resource pool, and creates a container to run the AI inference task, so that the AI applicant can use the resources of the intelligent computing platform to execute the inference of the AI model, and a platform is provided to test and deploy the AI model, realizing the practical application of AI technology.

[0081] The operation and maintenance personnel view the overall resource allocation of the platform, platform monitoring, and daily maintenance of the cluster through the front-end Web interface.

[0082] The specific method is that the AI applicant applies for the required intelligent computing resources through the Web interface, selects a pre-trained AI model, configures the required computing resources, submits the inference task, the platform schedules the task to the corresponding resource pool, and creates a container to run the AI inference task, so that the AI applicant can use the resources of the intelligent computing platform to execute the inference of the AI model, and a platform is provided to test and deploy the AI model, realizing the practical application of AI technology.

[0083] The operation and maintenance personnel login the Web interface of the intelligent computing platform, view the resource allocation of the cluster, monitor the performance and health status of the cluster, and perform daily cluster maintenance tasks such as updating software, fixing problems, and optimizing configuration; can ensure the stable operation and performance optimization of the intelligent computing platform, timely discover and solve possible problems, and guarantee the high availability of the platform.

[0084] Through the above steps, the intelligent computing platform provides a comprehensive AI development and deployment environment, from resource management to algorithm development, to model training and inference, and daily operation and maintenance management, forming a closed-loop AI development and application process.

[0085] In this embodiment, the method of the AI applicant applying for the computing power task in the intelligent computing platform includes the following steps:

[0086] Obtain the computing power task, which is obtained by uploading the Web interface provided by the intelligent computing platform.

[0087] The intelligent computing platform backend service receives the task request, performs verification and preprocessing, and prepares the metadata and configuration information of the task, which is equivalent to the intelligent computing platform verifying whether the uploaded task meets the requirements, including the integrity of the data set and code, and the rationality of the resource demand.

[0088] The user specifies the resource demand when submitting the task, or the intelligent computing platform automatically estimates the resource demand according to the task, and according to the resource demand, the resource type and quantity required by the computing power task are determined, such as CPU, memory, GPU or NPU, etc.

[0089] The intelligent computing platform encapsulates the task as a Kubernetes job (Job), deployment (Deployment) or its suitable workload type, and submits it to the Kubernetes API server to realize the access of the Kubernetes cluster.

[0090] The Kubernetes scheduler schedules tasks according to the resource requirements of the tasks and the resource status of the nodes. The scheduler checks the resource usage and node labels of all nodes in the cluster and selects the most suitable node to run the task.

[0091] The computing power task is scheduled to the target node in the Kubernetes cluster that has the required heterogeneous computing resources. The Kubernetes cluster creates a container on the target node, which loads the environment and dependencies required by the task to execute the computing power task.

[0092] The smart computing platform collects the running data of the task through the monitoring and logging system of Kubernetes and displays it to the user on the Web interface. The user can monitor the running status of the task, including progress, performance and resource usage, so as to understand the task execution in real time.

[0093] Once the task is completed, the Kubernetes stops the related Pod, and the smart computing platform saves the task result and provides it for the user to download or further analyze. The smart computing platform of the embodiment can provide a complete closed-loop process from task submission to execution for the user, so that the user can conveniently use the powerful computing power of the Kubernetes cluster to execute various computing power intensive tasks.

[0094] Another embodiment of the present application provides a multi-architecture-based smart computing platform deployment and management system, which executes the multi-architecture-based smart computing platform deployment and management method as described above. As shown in the system, Figure 4 the AI application layer uniformly manages the resources of each physical resource layer, while meeting the needs of AI researchers and AI application personnel. A complete, automated, and heterogeneous smart computing hardware-adapted cluster deployment and management platform is provided, which improves deployment efficiency, reduces deployment time cost, labor cost, and reduces deployment errors through standard configuration and deployment.

[0095] The functions of each module in the system embodiment can be referred to the corresponding description in the above method, which will not be repeated here.

[0096] Another embodiment of the present application provides an electronic device, Figure 5 a structural block diagram of an electronic device according to an embodiment of the present application is shown. As shown in the figure, Figure 5 the electronic device includes a memory 100 and a processor 200, and the memory 100 stores a computer program that can run on the processor 200. The processor 200 executes the computer program to implement the multi-architecture-based smart computing platform deployment and management method in the above embodiment. The number of the memory 100 and the processor 200 can be one or more.

[0097] The electronic device further includes

[0098] The communication interface 300 is configured to communicate with external devices and transmit data.

[0099] If the memory 100, the processor 200 and the communication interface 300 are independently implemented, the memory 100, the processor 200 and the communication interface 300 can be connected with each other through a bus and complete communication with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0100] Optionally, in a specific implementation, if the memory 100, the processor 200 and the communication interface 300 are integrated on a chip, the memory 100, the processor 200 and the communication interface 300 can complete communication with each other through an internal interface.

[0101] The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the method provided in the embodiment of the present application.

[0102] The embodiment of the present application further provides a chip, which comprises a processor, and the processor is configured to call and run instructions stored in a memory, so that a communication device installed with the chip executes the method provided in the embodiment of the present application.

[0103] The embodiment of the present application further provides a chip, which comprises an input interface, an output interface, a processor and a memory, and the input interface, the output interface, the processor and the memory are connected through an internal connection path. The processor is configured to execute code in the memory, and when the code is executed, the processor is configured to execute the method provided in the embodiment of the present application.

[0104] It is to be understood that the above-described processor can be a central processing unit (CPU), but can also be other general purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general purpose processor can be a microprocessor or any conventional processor, etc. It is to be noted that the processor can be an advanced RISC machine (ARM) architecture processor.

[0105] Further, the memory can include a read-only memory and a random access memory, and can further include a non-volatile random access memory. The memory can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can include a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically EPROM (EEPROM), or a flash memory. The volatile memory can include a random access memory (RAM) used as an external cache. By way of example, but not limitation, many forms of RAM can be used. For example, a static RAM (SRAM), a dynamic RAM (DRAM), a synchronous DRAM (SDRAM), a double data rate synchronous DRAM (DDR SDRAM), an enhanced SDRAM (ESDRAM), a synchlink DRAM (SLDRAM), and a direct rambus RAM (DR RAM) can be used.

[0106] In the above-described embodiments, all or part can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part generates the flow or function according to the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium.

[0107] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, different embodiments or examples described in the present specification and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0108] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0109] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for deploying and managing a multi-architecture based intelligent computing platform, characterized in that, Comprising: Identifying a plurality of heterogeneous computing units configured on a target host based on an aip-deploy tool, the heterogeneous computing units comprising image processing units and neural network processing units; Obtaining a preset offline package, building a Kubernetes cluster for each of the heterogeneous computing units according to the offline package, and deploying and configuring specified components of the Kubernetes cluster to generate an intelligent computing platform, the intelligent computing platform being configured to uniformly manage all of the heterogeneous computing units in the Kubernetes cluster. 2.The method of claim 1, wherein, The offline package contains a driver installation package for each of the heterogeneous computing units, all software components required for building and managing the Kubernetes cluster, and dependencies required for maintaining normal operation of cluster components. 3.The method of claim 2, wherein, Further comprising: Obtaining the driver and dependency components of each of the heterogeneous computing units according to the offline package; Installing the driver and the dependency components in the target host through the aip-deploy tool. 4.The method of claim 3, wherein, The building of the Kubernetes cluster for each of the heterogeneous computing units according to the offline package comprises: Installing Kubernetes components required by the Kubernetes cluster according to the offline package, deploying an etcd node of the Kubernetes cluster, and deploying a control plane of the Kubernetes cluster to implement the building of the Kubernetes cluster. 5.The method of claim 1, wherein, Further comprising: Obtaining architecture information of the heterogeneous computing units, labeling nodes of the Kubernetes cluster according to the architecture information, and deploying device plugins related to the heterogeneous computing units according to labels of each node, the device plugins being configured to report resource information of each of the heterogeneous computing units to the Kubernetes cluster. 6.The method of claim 5, wherein, Further comprising: Deploying a monitoring data collector corresponding to the heterogeneous computing units according to node labels of each node in the Kubernetes cluster, the monitoring data collector being configured to provide monitoring services for the Kubernetes cluster to collect monitoring data of the heterogeneous computing units.

7. The multi-architecture based intelligent computing platform deployment and management method of claim 1, wherein, Further comprising: Obtaining a computing power task, the computing power task being obtained through a web interface provided by the intelligent computing platform; Accessing the Kubernetes cluster according to the computing power task, scheduling the computing power task to a target node having required heterogeneous computing resources in the Kubernetes cluster, so that the Kubernetes cluster creates a container on the target node to run the computing power task.

8. A multi-architecture-based intelligent computing platform deployment and management system, characterized in that, Performing the multi-architecture-based intelligent computing platform deployment and management method according to any one of claims 1-7.

9. An electronic device, comprising: Comprising: A processor and a memory, the memory storing instructions, the instructions being loaded and executed by the processor to implement the multi-architecture-based intelligent computing platform deployment and management method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the multi-architecture-based intelligent computing platform deployment and management method in any one of claims 1-7.

Images