AI online interactive development method based on multiple computing platforms and related device

Abstract

This invention discloses an AI online interactive development method and related equipment based on multiple computing platforms. The method includes: an interactive development platform receiving a development task request and selecting suitable intelligent computing center resources; selecting a suitable intelligent computing center according to the platform's scheduling strategy and obtaining the authorization information of the intelligent computing platform of the suitable intelligent computing center; encapsulating the development task request according to the API interface requirements of the suitable intelligent computing center; the intelligent computing platform allocating the resources and environment required by the development task request according to the resource environment allocation request, and sending an access resource link to the interactive development platform; the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user. By establishing a unified interactive development platform, accessing multiple intelligent computing platforms, and providing an AI interactive development environment to users via the Internet, the convenience of development is improved and the development cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an AI online interactive development method and related equipment based on multiple computing platforms. Background Technology

[0002] Development work in the field of artificial intelligence typically requires specific hardware and software support environments. Hardware acceleration resources include various types such as GPUs (Graphics Processing Units, typically suitable for handling large-scale concurrent computations), NPUs (Neural-network Processing Units, employing a "data-driven parallel computing" architecture, particularly adept at processing massive multimedia data like video and images), and MLUs (Machine Learning Units, a type of cloud-based AI chip). Development frameworks include PyTorch, TensorFlow, MindSpore, and PaddlePaddle. Before starting development, developers usually need to spend considerable time configuring the development environment, including software version and underlying hardware acceleration resource version compatibility. Online interactive AI development systems provide developers with out-of-the-box development environments; users only need an internet connection to use the online environment for development.

[0003] The business requirement is to provide an online interactive development environment compatible with multiple computing platforms, supporting various hardware acceleration resources and development frameworks, and providing a configurable online runtime environment to offer users convenient tools for online development.

[0004] Developers in the field of artificial intelligence typically use the following methods when preparing the necessary development environment: 1) purchasing hardware equipment themselves; 2) renting cloud servers; 3) using online development platforms.

[0005] The first method requires significant investment in equipment and the installation of the operating system, drivers, frameworks, and other software. It also restricts the ability to experiment with new hardware and, since the development environment is typically local or within a LAN, remote work is difficult.

[0006] The second method is more flexible than the first, as it eliminates the need to install the underlying operating system and maintain the hardware. However, cloud servers are typically single-environment environments, and if multiple environments are required, multiple cloud servers still need to be purchased. Furthermore, cloud servers are leased for specified time periods, limiting user usage time, and users still need to pay for resources outside of development hours.

[0007] Existing online development environments typically cater to a single hardware resource and a single development framework. For example, AIStudio provides the PaddlePaddle development framework and GPU computing resources, while ModelArts provides the MindSpore development framework and NPU computing resources. When users need to adapt algorithms or models to multiple underlying hardware environments, they have to prepare multiple types of hardware devices or switch between multiple platforms, causing inconvenience.

[0008] Therefore, existing technologies still need to be improved and enhanced. Summary of the Invention

[0009] The main objective of this invention is to provide an AI online interactive development method and related equipment based on multiple computing platforms, aiming to solve the problems of high development threshold and poor convenience caused by the inability of existing online development environments to adapt to multiple underlying hardware environments.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] An AI online interactive development method based on multiple computing platforms includes the following steps:

[0012] The interactive development platform registers and configures multiple intelligent computing centers;

[0013] The interactive development platform receives development task requests created by registered users and selects suitable intelligent computing center resources based on the development task requests, thereby selecting multiple pre-selected intelligent computing centers.

[0014] After selecting an appropriate intelligent computing center from among the multiple pre-selected intelligent computing centers according to the platform scheduling strategy, the interactive development platform obtains the authorization information of the intelligent computing platform of the appropriate intelligent computing center and sends a resource environment allocation request to the intelligent computing platform.

[0015] The interactive development platform encapsulates the development task request according to the API interface requirements of the adaptive intelligent computing center and sends it to the intelligent computing platform in the adaptive intelligent computing center.

[0016] The intelligent computing platform allocates the resources and environment required by the development task request according to the resource allocation environment request, and sends a resource access link to the interactive development platform;

[0017] The interactive development platform associates the access resource link with the development task request and sends the access resource link to the registered user;

[0018] The interactive development platform receives a task stop request from the registered user, selects the adaptive intelligent computing center based on the task stop request, and sends the task stop request to the adaptive intelligent computing center so that the adaptive intelligent computing center releases the occupied resources.

[0019] In the AI ​​online interactive development method based on multiple computing platforms, the steps for the interactive development platform to register and configure multiple intelligent computing centers specifically include:

[0020] The interactive development platform registers and configures multiple intelligent computing centers that can be invoked, and obtains authentication and resource information from each intelligent computing center.

[0021] In the AI ​​online interactive development method based on multiple computing platforms, the interactive development platform receives development task requests established by registered users and selects suitable intelligent computing center resources according to the development task requests. The specific steps of selecting multiple pre-selected intelligent computing centers include:

[0022] The interactive development platform sequentially receives the registration and login information of the user to be registered, and receives the development task request obtained after the registered user performs the operation of configuring the required task parameters;

[0023] The interactive development platform extracts the required hardware resources from the development task request and matches them with the intelligent computing center resources to select multiple pre-selected intelligent computing centers.

[0024] In the AI ​​online interactive development method based on multiple computing platforms, the steps of the interactive development platform selecting an appropriate intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy, obtaining the authorization information of the intelligent computing platform of the appropriate intelligent computing center, and sending a resource environment allocation request to the intelligent computing platform specifically include:

[0025] The interactive development platform selects the compatible intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy.

[0026] The interactive development platform calls the first API interface of the intelligent computing platform, receives the input parameters of the first API interface, and receives the API access authorization information of the intelligent computing platform returned by the first API interface.

[0027] The interactive development platform calls the first API interface based on the API access authorization information and transmits the resource allocation environment request to the intelligent computing platform.

[0028] In the AI ​​online interactive development method based on multiple computing platforms, the step of the interactive development platform encapsulating the development task request according to the API interface requirements of the adapted intelligent computing center and sending it to the intelligent computing platform in the adapted intelligent computing center specifically includes:

[0029] The interactive development platform encapsulates the development task request according to the API interface requirements of the intelligent computing platform;

[0030] The interactive development platform, based on the API access authorization information returned by the first API interface, sends the encapsulated development task request to the intelligent computing platform in the intelligent computing center through the second API interface.

[0031] In the AI ​​online interactive development method based on multiple computing platforms, the step of the intelligent computing platform allocating the resources and environment required by the development task request according to the resource and environment allocation request, and sending access resource links to the interactive development platform specifically includes:

[0032] The intelligent computing platform allocates the resources and environment required by the development task request according to the resource and environment allocation request;

[0033] The intelligent computing platform sends a link to access resources to the interactive development platform.

[0034] In the AI ​​online interactive development method based on multiple computing platforms, the step of the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user specifically includes:

[0035] The interactive development platform associates the access resource link with the development task request;

[0036] The interactive development platform sends the access resource link to the registered user.

[0037] In the AI ​​online interactive development method based on multiple computing platforms, the interactive development platform receives a task stop request from the registered user, selects the appropriate intelligent computing center based on the task stop request, and sends the task stop request to the appropriate intelligent computing center so that the appropriate intelligent computing center can release the occupied resources. The specific steps include:

[0038] The interactive development platform receives the task stop request initiated by the registered user;

[0039] The interactive development platform selects the compatible intelligent computing center based on the task stop request;

[0040] The interactive development platform calls the third API interface of the adapted intelligent computing center to send the stop task request to the adapted intelligent computing center;

[0041] The adaptive intelligent computing center stops the task based on the stop task request and releases the resources occupied by the task.

[0042] In the AI ​​online interactive development method based on multiple computing platforms, the authentication information includes the username and password authorized to log in to the interactive development platform; the resource information includes the type and model of the hardware acceleration chip.

[0043] In the AI ​​online interactive development method based on multiple computing platforms, the required task parameters include the required hardware resources, resource specifications, and development framework.

[0044] In the AI ​​online interactive development method based on multiple computing platforms, the input parameter is the authorization information configured by the intelligent computing platform.

[0045] In the AI ​​online interactive development method based on multiple computing platforms, the resources include hardware acceleration resources, CPU, memory and storage, the environment includes an online accessible container, and the container includes an AI development framework.

[0046] An AI online interactive development system based on a multi-computing platform includes: an interactive development platform and an intelligent computing center; the interactive development platform includes a user management module, an intelligent computing resource management module, a user request preprocessing module, a task management module, and a scheduling and distribution module.

[0047] The user management module is used to receive registration and login information from users to be registered.

[0048] The intelligent computing resource management module is used to register and configure multiple intelligent computing centers;

[0049] The user request preprocessing module is used to receive development task requests created by registered users;

[0050] The scheduling and distribution module is used to select suitable intelligent computing center resources according to the development task request, so as to select multiple pre-selected intelligent computing centers;

[0051] The scheduling and distribution module is used to obtain the authorization information of the intelligent computing platform and send a resource environment allocation request to the intelligent computing platform;

[0052] The scheduling and distribution module is also used to select an adaptive intelligent computing center according to the platform scheduling strategy, and encapsulate the development task request according to the API interface requirements of the adaptive intelligent computing center, and send it to the intelligent computing platform in the adaptive intelligent computing center.

[0053] The intelligent computing platform is used to allocate the resources and environment required by the development task request according to the resource allocation environment request, and send the access resource link to the scheduling and distribution module;

[0054] The task management module is used to associate the access resource link with the development task request and send the access resource link to the registered user;

[0055] The task management module is also used to receive a stop task request from the registered user, select the adaptive intelligent computing center according to the stop task request, and send the stop task request to the adaptive intelligent computing center.

[0056] A computer-readable storage medium stores an AI online interactive development program based on multiple computing platforms, which, when executed by a processor, implements the steps of the AI ​​online interactive development method based on multiple computing platforms as described above.

[0057] Compared to existing technologies, this invention provides an AI online interactive development method and related equipment based on multiple computing platforms. The method includes: an interactive development platform receiving a development task request and selecting suitable intelligent computing center resources; selecting a suitable intelligent computing center according to the platform's scheduling strategy and obtaining the authorization information of the intelligent computing platform of the suitable intelligent computing center; encapsulating the development task request according to the API interface requirements of the suitable intelligent computing center; the intelligent computing platform allocating the resources and environment required by the development task request according to the resource environment allocation request, and sending an access resource link to the interactive development platform; the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user. By establishing a unified interactive development platform, accessing multiple intelligent computing platforms, and providing an AI interactive development environment to users via the Internet, the convenience of development is improved and the development cost is reduced. Attached Figure Description

[0058] Figure 1 A flowchart illustrating a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0059] Figure 2 A flowchart of step S200 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0060] Figure 3 A flowchart of step S300 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0061] Figure 4 A flowchart of step S400 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0062] Figure 5 A flowchart of step S500 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0063] Figure 6 A flowchart of step S600 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0064] Figure 7 A flowchart of step S700 in a preferred embodiment of the AI ​​online interactive development method based on multiple computing platforms provided by the present invention;

[0065] Figure 8 The flowchart for creating an interactive development task provided by this invention;

[0066] Figure 9 Architecture diagram of an online artificial intelligence development system compatible with multiple computing platforms. Detailed Implementation

[0067] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0068] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0069] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0070] To facilitate understanding of the embodiments of this application, the implementation process of the AI ​​online interactive development method based on multiple computing platforms in the embodiments of this invention will be described first:

[0071] By leveraging the open APIs (Application Programming Interfaces) of multiple intelligent computing centers (intelligent computing centers are the primary centers for producing computing power in the intelligent era. They use converged architecture computing systems as platforms and data as resources, enabling powerful computing to drive AI models to deeply process data, continuously generating various intelligent computing services, and providing them to organizations and individuals via the network in the form of cloud services), a unified online interactive AI development platform can be established. This platform will aggregate the capabilities of multiple intelligent computing centers and simultaneously provide development environments for various computing chip hardware facilities and multiple types of AI development frameworks.

[0072] Similar to cloud computing services, the intelligent computing center manages its AI cluster devices in a unified manner, providing them to organizations and individuals as cloud services via the network. Unlike cloud computing, which provides a cloud server environment, users of the intelligent computing center use the cloud service by specifying the required computing resources and submitting an AI task request. The server configures the corresponding task environment based on the resource request and allocates it to the user. Users only use the resources while executing the AI ​​task; the resources are released when the task ends.

[0073] Therefore, this invention provides an AI online interactive development method and related equipment based on multiple computing platforms. In this invention, the interactive development platform selects suitable intelligent computing center resources based on the development task request established by the registered user, choosing from multiple pre-selected intelligent computing centers. After obtaining the authorization information of the intelligent computing platform, it sends a resource allocation environment request to the intelligent computing platform. The intelligent computing platform then allocates the resources and environment required by the development task request according to the resource allocation environment request and sends an access resource link to the interactive development platform. The interactive development platform then associates the access resource link with the development task request and sends the access resource link to the registered user. This provides an online interactive development environment compatible with multiple computing platforms, supporting various hardware acceleration resources and development frameworks, and providing a configurable online runtime environment. It offers convenient tools for users to develop online, effectively lowering the AI ​​development threshold, reducing development costs, improving convenience, and increasing resource utilization.

[0074] The following describes the design scheme of the AI ​​online interactive development method based on multiple computing platforms through specific exemplary embodiments. It should be noted that the following embodiments are only used to explain the technical solution of the invention and are not intended to limit it specifically:

[0075] Please see Figure 1 This invention provides an AI online interactive development method based on multiple computing platforms, which includes the following steps:

[0076] The S100 interactive development platform registers and configures multiple intelligent computing centers.

[0077] Specifically, the interactive development platform connects with multiple intelligent computing centers, that is, the interactive development platform registers and configures the intelligent computing centers that the online artificial intelligence development platform (interactive development platform) can call, thereby informing the interactive development platform how many intelligent computing centers it can use.

[0078] Furthermore, the steps for S100 and the interactive development platform to register and configure multiple intelligent computing centers specifically include:

[0079] S110. The interactive development platform registers and configures multiple callable intelligent computing centers, and obtains authentication and resource information for each center. The authentication information includes the username and password authorized to log in to the interactive development platform; the resource information includes the type and model of the hardware acceleration chip.

[0080] Specifically, the interactive development platform connects with multiple intelligent computing centers. The interactive development platform registers and configures the intelligent computing centers that the online AI development platform can call. This primarily includes authentication information for the AI ​​computing platforms within the intelligent computing centers, such as the username, password, and access token (an credential for API calls used to verify permission) authorized for platform login and API calls. The resource information of the AI ​​computing platform includes the type and model of the hardware acceleration chip (e.g., GPU Nvidia2080Ti or NPU Ascend910), resource specifications (e.g., 1 GPU, 8-core CPU, 16GB RAM), and available development frameworks (e.g., PyTorch, TensorFlow, MindSpore). The AI ​​computing platform is the application that provides computing power services to the intelligent computing centers.

[0081] For further information, please refer to [link / reference]. Figure 1 S200, the interactive development platform receives development task requests established by registered users, and selects suitable intelligent computing center resources according to the development task requests, so as to select multiple pre-selected intelligent computing centers.

[0082] Specifically, after registering and configuring multiple intelligent computing centers, when a registered user creates a development task request, the interactive development platform selects suitable intelligent computing center resources based on the development task request, thereby selecting multiple pre-selected intelligent computing centers. This completes the adaptation of the development task request to the selectable intelligent computing centers provided by the interactive development platform, selecting an selectable intelligent computing center from all intelligent computing centers connected to the interactive development platform, thus obtaining multiple (or possibly one) pre-selected intelligent computing centers. This invention solves the problem of AI developers lacking dedicated environments by establishing a unified AI development platform (interactive development platform), connecting to multiple AI computing platforms, and providing an AI interactive development environment to users via the Internet.

[0083] Furthermore, please refer to Figure 2 S200, the interactive development platform receives development task requests established by registered users, and selects suitable intelligent computing center resources according to the development task requests. The specific steps of selecting multiple pre-selected intelligent computing centers include:

[0084] S210, the interactive development platform sequentially receives the registration and login information of the user to be registered, and receives the development task request obtained after the registered user performs the operation of configuring the required task parameters; wherein, the required task parameters include the required hardware resources, resource specifications and development framework.

[0085] S220. The interactive development platform extracts the required hardware resources from the development task request and matches them with the intelligent computing center resources to select multiple pre-selected intelligent computing centers.

[0086] Specifically, after registering and configuring multiple intelligent computing centers, the process for a registered user to create the interactive development task is as follows: The user registers and logs in on the online AI development platform to complete the platform's user registration and login authentication functions, thus becoming a registered user. Then, the user begins creating the interactive development task and configures the task parameters, including required hardware resources, resource specifications, and the development framework. After completing the task configuration, the user submits the task. Next, the interactive development platform extracts the required hardware resources from the development task request and matches them with the resources of the intelligent computing centers to select from multiple pre-selected intelligent computing centers.

[0087] For further information, please refer to [link / reference]. Figure 1 S300, after the interactive development platform selects an adaptive intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy, it obtains the authorization information of the intelligent computing platform of the adaptive intelligent computing center and sends a resource environment allocation request to the intelligent computing platform.

[0088] Specifically, after selecting multiple pre-selected intelligent computing centers based on the development task request, the interactive development platform determines which intelligent computing center the development task request will be distributed to according to the platform scheduling strategy, i.e., determines that the development task request will be distributed to the adapted intelligent computing center; and the interactive development platform further obtains the authorization information (access token) of the intelligent computing platform (AI computing platform) of the adapted intelligent computing center, thereby completing the specific adaptation to the most suitable adapted intelligent computing center according to the development task request, so as to make the most rational and maximized use of the intelligent computing center resources. Finally, the platform sends the resource allocation environment request to the intelligent computing platform.

[0089] Furthermore, please continue reading Figure 3 S300, the interactive development platform selects an adaptive intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy, obtains the authorization information of the intelligent computing platform of the adaptive intelligent computing center, and sends a resource environment allocation request to the intelligent computing platform. The specific steps include:

[0090] S310. The interactive development platform selects the adaptive intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy.

[0091] S320. The interactive development platform calls the first API interface of the intelligent computing platform, receives the input parameters of the first API interface, and receives the API access authorization information of the intelligent computing platform returned by the first API interface; wherein, the input parameters are the authorization information configured by the intelligent computing platform;

[0092] S330. The interactive development platform calls the first API interface based on the API access authorization information and transmits the resource allocation environment request to the intelligent computing platform.

[0093] Specifically, after selecting multiple pre-selected intelligent computing centers based on the required hardware resources in the development task request, the interactive development platform selects the appropriate intelligent computing center from the multiple pre-selected intelligent computing centers according to the platform scheduling strategy, so as to determine which intelligent computing center the development task request is specifically distributed to; then, the interactive development platform calls the first API interface of the intelligent computing platform, receives the input parameters of the first API interface (the input parameters are the authorization information configured by the intelligent computing platform), and receives the API access authorization information of the intelligent computing platform returned by the first API interface; finally, the interactive development platform calls the first API interface based on the API access authorization information to transmit the resource allocation environment request to the intelligent computing platform.

[0094] For further information, please refer to [link / reference]. Figure 1 S400, the interactive development platform encapsulates the development task request according to the API interface requirements of the adaptive intelligent computing center and sends it to the intelligent computing platform in the adaptive intelligent computing center.

[0095] Specifically, after obtaining the authorization information of the intelligent computing platform of the adaptive intelligent computing center, the interactive development platform encapsulates the development task request according to the API interface requirements of the adaptive intelligent computing center, and sends the encapsulated development task request to the intelligent computing platform in the adaptive intelligent computing center. This unifies the interface specifications of API interfaces on different platforms, reduces the occurrence of unreceived requests due to incompatibility between API interfaces on different AI computing platforms, and facilitates the transmission of development task requests between different interfaces.

[0096] Furthermore, please refer to Figure 4 S400, the step of the interactive development platform encapsulating the development task request according to the API interface requirements of the adapted intelligent computing center and sending it to the intelligent computing platform in the adapted intelligent computing center specifically includes:

[0097] S410. The interactive development platform encapsulates the development task request according to the API interface requirements of the intelligent computing platform;

[0098] S420. The interactive development platform sends the encapsulated development task request to the intelligent computing platform in the intelligent computing center through the second API interface based on the API access authorization information returned by the first API interface.

[0099] Specifically, after obtaining the authorization information from the intelligent computing platform, the interactive development platform encapsulates the development task request according to the API interface requirements of the intelligent computing platform. This is because different AI computing platforms may have different API parameters; if task transmission is required across different AI computing platforms, the request content needs to be encapsulated according to the corresponding API interface specifications. Then, based on the API access authorization information returned by the first API interface of the adapted intelligent computing center, the interactive development platform sends the encapsulated development task request to the intelligent computing platform in the intelligent computing center through the second API interface of the adapted intelligent computing center.

[0100] For further information, please refer to [link / reference]. Figure 1 S500, the intelligent computing platform allocates the resources and environment required by the development task request according to the resource allocation environment request, and sends an access resource link to the interactive development platform. The resources include hardware acceleration resources, CPU, memory, and storage; the environment includes an online-accessible container, and the container includes an AI development framework.

[0101] Specifically, after the interactive development platform sends the resource allocation environment request to the intelligent computing platform, the intelligent computing platform allocates the resources (including hardware acceleration resources, CPU, memory, and storage) and environment (including online accessible containers, including AI development frameworks) required by the development task request according to the resource allocation environment request, and sends an access resource link (resource access URL link) to the interactive development platform, thereby realizing the allocation of the resources and environment required by the development task request according to the resource allocation environment request.

[0102] Furthermore, please refer to Figure 5 S500, the steps of the intelligent computing platform allocating the resources and environment required by the development task request according to the resource allocation environment request, and sending the access resource link to the interactive development platform specifically include:

[0103] S510. The intelligent computing platform allocates the resources and environment required by the development task request according to the resource and environment allocation request.

[0104] S520, The intelligent computing platform sends a resource access link to the interactive development platform.

[0105] Specifically, after receiving the resource allocation request, the intelligent computing platform allocates the resources and environment required by the development task request according to the resource allocation request. Simultaneously, the intelligent computing platform sends the access resource link to the interactive development platform. Each access resource link corresponds one-to-one with a task.

[0106] For further information, please refer to [link / reference]. Figure 1 S600, the interactive development platform associates the access resource link with the development task request and sends the access resource link to the registered user.

[0107] Specifically, after the intelligent computing platform sends the access resource link to the interactive development platform, the interactive development platform associates the access resource link with the development task request and sends the access resource link to the registered user, thereby realizing the association between the access resource link and the development task request, so that the access resource link and the development task request have a specific matching relationship.

[0108] Furthermore, please refer to Figure 6 S600, the step of the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user specifically includes:

[0109] S610, The interactive development platform associates the access resource link with the development task request;

[0110] S620, The interactive development platform sends the access resource link to the registered user.

[0111] Specifically, after the intelligent computing platform receives the access resource link, since the AI ​​development platform's backend database records the URL correspondence between AI development tasks and the resource links allocated to the tasks, the interactive development platform can establish an association between the access resource link and the development task request. Then, the interactive development platform sends the access resource link to the registered user.

[0112] For further information, please refer to [link / reference]. Figure 1 S700, the interactive development platform receives a stop task request from the registered user, selects the adaptive intelligent computing center according to the stop task request, and sends the stop task request to the adaptive intelligent computing center.

[0113] Specifically, after the registered user receives the access resource link, they can initiate a stop task request to the interactive development platform via that link. Then, the interactive development platform selects the appropriate intelligent computing center based on the stop task request and sends the request there. Finally, the intelligent computing center releases the occupied resources, truly stopping the task. This allows the interactive development platform to easily control the intelligent computing center to release the occupied resources by initiating the stop task request via the access resource link, thus ensuring that resources are released as the AI ​​task ends, reducing development costs and improving resource utilization. Furthermore, when the AI ​​task is released, the resources are also released, and the resource link becomes invalid.

[0114] Furthermore, please refer to Figure 7 S700, the interactive development platform receives a stop task request from the registered user, selects the adaptive intelligent computing center according to the stop task request, and sends the stop task request to the adaptive intelligent computing center so that the adaptive intelligent computing center releases the occupied resources. The specific steps include:

[0115] S710, The interactive development platform receives the task stop request initiated by the registered user;

[0116] S720. The interactive development platform selects the adaptive intelligent computing center according to the stop task request.

[0117] S730, The interactive development platform calls the third API interface of the adaptive intelligent computing center to send the stop task request to the adaptive intelligent computing center;

[0118] S740. The adaptive intelligent computing center stops the task according to the stop task request and releases the resources occupied by the task.

[0119] Specifically, after the registered user receives the access resource link, an interactive development environment window will pop up when the user accesses the access resource link through a web page. When the registration needs to be interrupted or the development work has been completed, the user can close the interactive development environment window and select to stop the task by pressing the stop button on the interactive development platform, that is, to initiate the stop task request to the interactive development platform. Then, the interactive development platform selects the corresponding adaptive intelligent computing center according to the stop task request and calls the third API interface of the adaptive intelligent computing center to send the stop task request to the adaptive intelligent computing center. Finally, after receiving the stop task request, the adaptive intelligent computing center stops the task and reclaims the corresponding resources occupied by the task. The final flowchart for creating an interactive development task is as follows: Figure 8 As shown.

[0120] Further, please refer to Figure 9 This invention provides an AI online interactive development system based on multiple computing platforms. The AI ​​online interactive development system based on multiple computing platforms includes: an interactive development platform 10 (… Figure 9 Online AI development platforms and intelligent computing centers (including...) Figure 9 The interactive development platform 10 includes a user management module 11, an intelligent computing resource management module 12, a user request preprocessing module 13, a task management module 14, and a scheduling and distribution module 15.

[0121] The user management module 11 is used to receive registration and login information of users to be registered;

[0122] The intelligent computing resource management module 12 is used to register and configure multiple intelligent computing centers 20;

[0123] The user request preprocessing module 13 is used to receive development task requests established by registered users;

[0124] The scheduling and distribution module 15 is used to select suitable intelligent computing center resources according to the development task request, so as to select multiple pre-selected intelligent computing centers 21;

[0125] The scheduling and distribution module 15 is used to obtain the authorization information of the intelligent computing platform 30 and send a resource environment allocation request to the intelligent computing platform 30.

[0126] The scheduling and distribution module 15 is also used to select the adaptive intelligent computing center 211 according to the platform scheduling strategy, and encapsulate the development task request according to the API interface requirements of the adaptive intelligent computing center 211 and send it to the intelligent computing platform 30 in the adaptive intelligent computing center 211.

[0127] The intelligent computing platform 30 is used to allocate the resources and environment required by the development task request according to the resource allocation environment request, and send the access resource link to the scheduling and distribution module 15;

[0128] The task management module 14 is used to associate the access resource link with the development task request and send the access resource link to the registered user;

[0129] The task management module 14 is also used to receive a stop task request from the registered user, select the adaptive intelligent computing center 211 according to the stop task request, and send the stop task request to the adaptive intelligent computing center 211.

[0130] Specifically, Figure 9 The diagram shows the architecture of an online artificial intelligence development system compatible with multiple computing platforms. It takes as an example that m users (to be registered) register and enroll on the online artificial intelligence development platform (interactive development platform), and two AI computing platforms (also called intelligent computing platforms, AI computing platform 1 and AI computing platform 2) in two intelligent computing centers (intelligent computing center A and intelligent computing center B). The AI ​​computing platform includes an API interface layer, a task scheduling layer and a resource management layer.

[0131] By establishing a unified AI interactive development platform for internet users, the platform receives AI interactive development task requests initiated by users and distributes them to AI computing platforms connected to the platform according to the resource requirements of the tasks. Each AI computing platform manages an AI computing cluster resource (an AI cluster consists of multiple AI servers) and assigns AI tasks to the servers within the cluster.

[0132] The working principle of the AI ​​interactive development platform is explained below: Since the AI ​​online interactive development system based on multiple computing platforms implements an AI online interactive development method based on multiple computing platforms, the specific details of this method are as follows:

[0133] First, the intelligent computing resource management module 12 is used to register and configure multiple intelligent computing centers 20, so as to register and configure the AI ​​computing platforms that can be called by the online artificial intelligence development platform. This mainly includes the authentication information of the AI ​​computing platforms in the intelligent computing centers, namely the username, password, and access token (Access token: an credential for API interface calls, used to verify whether the API interface calls are authorized) authorized to log in and call the platform. The resource information of the AI ​​computing platform includes the type and model of hardware acceleration chips (such as GPU Nvidia 2080Ti or NPU Ascend910), resource specifications (such as 1 GPU, 8-core CPU, 16G memory) and available development frameworks (such as PyTorch, TensorFlow, MindSpore).

[0134] Then, the user to be registered accesses the online artificial intelligence development platform (interactive development platform) through a web page to complete user registration, login, and initiation of interactive development task requests, specifically as follows:

[0135] The user to be registered registers and logs in to the online artificial intelligence development platform. The user management module 11 receives the registration and login information of the user to be registered. That is, based on the user management module 11 of the online artificial intelligence development platform, the platform user registration and login authentication functions are completed. In addition, the registered user submits a new interactive development task to the interactive development platform after creating a new interactive development task on the online artificial intelligence development platform.

[0136] Secondly, the interactive development platform processes the received development task requests and distributes the task requests according to the interactive development task environment requirements set by the registered user. Specifically, it calls the underlying AI computing platform to generate computing resources via API interfaces and returns the obtained resource information to the registered user.

[0137] After a registered user creates an interactive development task on the online AI development platform, the user request preprocessing module 13 receives the development task request created by the registered user. The scheduling and distribution module 15 then selects suitable intelligent computing center resources based on the development task request, in order to select multiple pre-selected intelligent computing centers 21. Subsequently, the scheduling and distribution module 15 obtains the authorization information of the intelligent computing platform 30 and sends a resource allocation environment request to the intelligent computing platform 30. Simultaneously, the scheduling and distribution module 15 is also used to implement the platform's scheduling strategy (the scheduling algorithm within the scheduling and distribution module 15). The algorithm (which can have different implementations) selects an adaptive intelligent computing center 211, and encapsulates the development task request according to the API interface requirements of the adaptive intelligent computing center 211 before sending it to the intelligent computing platform 30 in the adaptive intelligent computing center 211. The intelligent computing platform 30 then allocates the resources and environment required by the development task request according to the resource allocation environment request, and sends an access resource link to the scheduling and distribution module 15. Finally, the task management module 14 associates the access resource link with the development task request and sends the access resource link to the registered user.

[0138] After the registered user receives the access resource link, they can initiate a stop task request to the interactive development platform via the access resource link. The task management module 14 receives the stop task request from the registered user, selects the adaptive intelligent computing center according to the stop task request, and sends the stop task request to the adaptive intelligent computing center. Finally, the adaptive intelligent computing center releases the occupied resources, and the task is truly stopped.

[0139] This invention establishes a unified AI development platform, connects to multiple AI computing platforms, and provides users with an interactive AI development environment via the Internet. This solves the problem of AI developers using dedicated environments and provides an online development environment for AI developers, effectively lowering the threshold for AI development and improving convenience. At the same time, it provides multiple types of AI acceleration chip infrastructure, allowing developers to choose multiple environments as needed on a single platform. Furthermore, resources are released as AI tasks start and end, which can reduce development costs and improve resource utilization.

[0140] Based on a unified online interactive AI development platform, this invention allows users to access the hardware resources needed for AI development simply by connecting to the internet. Users do not need to purchase equipment or configure the basic equipment environment and network infrastructure themselves. Furthermore, it enables work in multiple types of infrastructure environments based on a single platform without switching platforms. Resources can also be used on demand, flexibly turned on and off according to tasks, effectively improving the overall utilization rate of intelligent computing center resources.

[0141] Furthermore, the present invention provides a computer-readable storage medium storing an AI online interactive development program based on multiple computing platforms. When the AI ​​online interactive development program based on multiple computing platforms is executed by a processor, it implements the steps of the AI ​​online interactive development method based on multiple computing platforms as described above. Since the steps of the AI ​​online interactive development method based on multiple computing platforms have been described in detail above, they will not be repeated here.

[0142] In summary, this invention provides an AI online interactive development method and related equipment based on multiple computing platforms. The method includes: an interactive development platform receiving a development task request and selecting suitable intelligent computing center resources; selecting a suitable intelligent computing center according to the platform scheduling strategy and obtaining the authorization information of the intelligent computing platform of the suitable intelligent computing center; encapsulating the development task request according to the API interface requirements of the suitable intelligent computing center; the intelligent computing platform allocating the resources and environment required by the development task request according to the resource environment allocation request, and sending an access resource link to the interactive development platform; the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user. By establishing a unified interactive development platform, accessing multiple intelligent computing platforms, and providing an AI interactive development environment to users via the Internet, the convenience of development is improved and the development cost is reduced.

[0143] It is understood that those skilled in the art can make equivalent substitutions or modifications to the technical solution and inventive concept of the present invention, and all such substitutions or modifications should fall within the protection scope of the appended claims.

Claims

1. A method for online interactive AI development based on multiple computing platforms, characterized in that, The AI ​​online interactive development method based on multiple computing platforms includes: The interactive development platform registers and configures multiple intelligent computing centers; The interactive development platform receives development task requests created by registered users and selects suitable intelligent computing center resources based on the development task requests, thereby selecting multiple pre-selected intelligent computing centers. After selecting an appropriate intelligent computing center from among the multiple pre-selected intelligent computing centers according to the platform scheduling strategy, the interactive development platform obtains the authorization information of the intelligent computing platform of the appropriate intelligent computing center and sends a resource environment allocation request to the intelligent computing platform. The interactive development platform encapsulates the development task request according to the API interface requirements of the adaptive intelligent computing center and sends it to the intelligent computing platform in the adaptive intelligent computing center. The intelligent computing platform allocates the resources and environment required by the development task request according to the resource allocation environment request, and sends a resource access link to the interactive development platform; The interactive development platform associates the access resource link with the development task request and sends the access resource link to the registered user; The interactive development platform receives a task stop request from the registered user, selects the adaptive intelligent computing center based on the task stop request, and sends the task stop request to the adaptive intelligent computing center so that the adaptive intelligent computing center can release the occupied resources.

2. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The interactive development platform's steps for registering and configuring multiple intelligent computing centers specifically include: The interactive development platform registers and configures multiple intelligent computing centers that can be invoked, and configures the authentication information and resource information of each intelligent computing center.

3. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The interactive development platform receives development task requests created by registered users, and selects suitable intelligent computing center resources based on the development task requests. The specific steps for selecting multiple pre-selected intelligent computing centers include: The interactive development platform sequentially receives the registration and login information of the user to be registered, and receives the development task request obtained after the registered user performs the operation of configuring the required task parameters; The interactive development platform extracts the required hardware resources from the development task request and matches them with the intelligent computing center resources to select multiple pre-selected intelligent computing centers.

4. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The interactive development platform selects an compatible intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy, obtains the authorization information of the intelligent computing platform of the compatible intelligent computing center, and sends a resource environment allocation request to the intelligent computing platform. The specific steps include: The interactive development platform selects the compatible intelligent computing center from multiple pre-selected intelligent computing centers according to the platform scheduling strategy. The interactive development platform calls the first API interface of the intelligent computing platform, receives the input parameters of the first API interface, and receives the API access authorization information of the intelligent computing platform returned by the first API interface. The interactive development platform calls the first API interface based on the API access authorization information and transmits the resource allocation environment request to the intelligent computing platform.

5. The AI ​​online interactive development method based on multiple computing platforms according to claim 4, characterized in that, The interactive development platform encapsulates the development task request according to the API interface requirements of the adapted intelligent computing center and sends it to the intelligent computing platform in the adapted intelligent computing center. The specific steps include: The interactive development platform encapsulates the development task request according to the API interface requirements of the intelligent computing platform; The interactive development platform, based on the API access authorization information returned by the first API interface, sends the encapsulated development task request to the intelligent computing platform in the intelligent computing center through the second API interface.

6. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The step of the interactive development platform associating the access resource link with the development task request and sending the access resource link to the registered user specifically includes: The interactive development platform associates the access resource link with the development task request; The interactive development platform sends the access resource link to the registered user.

7. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The interactive development platform receives a task stop request from the registered user, selects the adapted intelligent computing center based on the task stop request, and sends the task stop request to the adapted intelligent computing center so that the adapted intelligent computing center releases the occupied resources. The specific steps include: The interactive development platform receives the task stop request initiated by the registered user; The interactive development platform selects the compatible intelligent computing center based on the task stop request; The interactive development platform calls the third API interface of the adapted intelligent computing center to send the stop task request to the adapted intelligent computing center; The adaptive intelligent computing center stops the task based on the stop task request and releases the resources occupied by the task.

8. The AI ​​online interactive development method based on multiple computing platforms according to claim 2, characterized in that, The authentication information includes the username and password authorized to log in to the interactive development platform; the resource information includes the type and model of the hardware acceleration chip.

9. The AI ​​online interactive development method based on multiple computing platforms according to claim 3, characterized in that, The required task parameters include the required hardware resources, resource specifications, and development framework.

10. The AI ​​online interactive development method based on multiple computing platforms according to claim 4, characterized in that, The input parameters are the authorization information configured in the intelligent computing platform.

11. The AI ​​online interactive development method based on multiple computing platforms according to claim 1, characterized in that, The resources include hardware acceleration resources, CPU, memory, and storage, and the environment includes online-accessible containers, which include AI development frameworks.

12. An AI online interactive development system based on multiple computing platforms, characterized in that, The AI ​​online interactive development system based on multiple computing platforms includes: an interactive development platform and an intelligent computing center; the interactive development platform includes a user management module, an intelligent computing resource management module, a user request preprocessing module, a task management module, and a scheduling and distribution module; The user management module is used to receive registration and login information from users to be registered. The intelligent computing resource management module is used to register and configure multiple intelligent computing centers; The user request preprocessing module is used to receive development task requests created by registered users; The scheduling and distribution module is used to select suitable intelligent computing center resources according to the development task request, so as to select multiple pre-selected intelligent computing centers; The scheduling and distribution module is used to obtain the authorization information of the intelligent computing platform and send a resource environment allocation request to the intelligent computing platform; The scheduling and distribution module is also used to select an adaptive intelligent computing center according to the platform scheduling strategy, and encapsulate the development task request according to the API interface requirements of the adaptive intelligent computing center, and send it to the intelligent computing platform in the adaptive intelligent computing center. The intelligent computing platform is used to allocate the resources and environment required by the development task request according to the resource allocation environment request, and send the access resource link to the scheduling and distribution module; The task management module is used to associate the access resource link with the development task request and send the access resource link to the registered user; The task management module is also used to receive a stop task request from the registered user, select the adaptive intelligent computing center according to the stop task request, and send the stop task request to the adaptive intelligent computing center.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an AI online interactive development program based on multiple computing platforms, which, when executed by a processor, implements the steps of the AI ​​online interactive development method based on multiple computing platforms as described in any one of claims 1-11.

Images