Service invocation method and apparatus, processing device, storage medium, and program product

Abstract

The present disclosure provides a service invocation method and apparatus, a processing device, a storage medium, and a program product, applied to an RIC platform, wherein the RIC platform supports multiple services, and the multiple services are used by a first application to perform a machine learning workflow. The method comprises: receiving configuration information sent by a first application, the configuration information being used for creating tasks related to the machine learning workflow on the multiple services; invoking the multiple services to execute the tasks to obtain a machine learning result; and sending the machine learning result to the first application.

Description

Service calling method, device, processing equipment, storage medium and program product

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure is based on and claims the priority of Chinese patent application with application number 202410370783.6 and application date March 28, 2024. The entire content of the Chinese patent application is hereby incorporated into this disclosure as a reference. Technical Field

[0003] The present disclosure relates to the field of wireless communication technologies, and in particular to a service invoking method, apparatus, processing device, storage medium, and program product. Background Art

[0004] The Near-Real-Time RAN Intelligent Controller (O-RAN Near-RT RIC) is a functional entity defined by the O-RAN Alliance based on the existing 5G Radio Access Network (5G RAN) standard. The O-RAN Near-RT RIC is further divided into two functional entities: a single Radio Access Network Intelligent Controller (RAN Intelligent Controller, RIC) platform and multiple intelligent applications (xApps). In related technologies, when an xApp utilizes RIC platform services to implement the "data collection-data preprocessing-model inference / training" workflow, the xApp must participate in each step of the workflow's loop. Summary of the Invention

[0005] Embodiments of the present disclosure provide a service invoking method, apparatus, processing device, storage medium, and program product.

[0006] The present disclosure provides a service invocation method, which is applied to a RIC platform. The RIC platform supports multiple services, and the multiple services are used by a first application to perform a machine learning workflow. The method includes:

[0007] receiving configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on the multiple services;

[0008] Calling the multiple services to perform the task and obtain machine learning results;

[0009] Send the machine learning result to the first application.

[0010] The present disclosure further provides a service invocation method, which is applied to a first application and includes:

[0011] Sending configuration information to a RIC platform, where the RIC platform supports multiple services, the multiple services are used for the first application to perform a machine learning workflow, and the configuration information is used to create tasks related to the machine learning workflow on the multiple services;

[0012] Receive the machine learning results sent by the RIC platform, where the machine learning results are obtained by the RIC platform calling the multiple services to perform the task.

[0013] The present disclosure provides a service invocation device, which is applied to a RIC platform. The RIC platform supports multiple services, and the multiple services are used by a first application to perform a machine learning workflow. The device includes:

[0014] a first receiving unit, configured to receive configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on the multiple services;

[0015] A first calling unit is configured to call the multiple services to execute the task and obtain a machine learning result;

[0016] A first sending unit is used to send the machine learning result to the first application.

[0017] The present disclosure further provides a service invoking device, which is applied to a first application and includes:

[0018] a second sending unit, configured to send configuration information to a RIC platform, wherein the RIC platform supports multiple services, the multiple services are used by the first application to perform a machine learning workflow, and the configuration information is used to create tasks related to the machine learning workflow on the multiple services;

[0019] The second receiving unit is used to receive the machine learning results sent by the RIC platform, where the machine learning results are obtained by the RIC platform calling the multiple services to perform the task.

[0020] An embodiment of the present disclosure provides a processing device, including: a processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory, and execute any one of the above-mentioned service calling methods.

[0021] An embodiment of the present disclosure provides a computer-readable storage medium for storing a computer program, wherein the computer program enables a computer to execute any one of the above-mentioned service calling methods.

[0022] An embodiment of the present disclosure provides a computer program product, including computer program instructions, which enable a computer to execute any one of the above-mentioned service calling methods.

[0023] In the technical solution of the embodiment of the present disclosure, the RIC platform receives configuration information sent by the first application. This configuration information is used to create tasks related to the machine learning workflow on multiple services supported by the RIC platform. Then, the RIC platform calls multiple services to perform tasks related to the machine learning workflow, obtains machine learning results, and then sends the machine learning results to the first application. In this way, in the process of implementing the machine learning workflow, the first application pre-configures the relevant services, enabling continuous calls between multiple services without requiring the participation of the first application at every step. This greatly reduces the signaling overhead of the first application and the transmission pressure on the first application-side interface when forwarding data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] FIG1 is a schematic diagram of the logical architecture of an O-RAN provided by an embodiment of the present disclosure;

[0025] FIG2 is a schematic diagram of the internal architecture of the RIC provided in an embodiment of the present disclosure;

[0026] FIG3 is a schematic diagram of a baseline solution for implementing a machine learning workflow using the RIC platform provided by an xApp according to an embodiment of the present disclosure;

[0027] FIG4 is a flow chart of a service invocation method applied to a RIC platform according to an embodiment of the present disclosure;

[0028] FIG5 is a schematic diagram of a workflow scheme of “data collection - data preprocessing - model reasoning / training” in which an upstream service calls a downstream service provided by an embodiment of the present disclosure;

[0029] FIG6 is a schematic diagram of a workflow scheme of “data collection - data preprocessing - model reasoning / training” in which a downstream service calls an upstream service according to an embodiment of the present disclosure;

[0030] FIG7 is a schematic diagram of a solution for implementing a multi-service serial workflow through a single service entry provided by an embodiment of the present disclosure;

[0031] FIG8 is a flow chart of a service calling method applied to a first application according to an embodiment of the present disclosure;

[0032] FIG9 is a schematic structural diagram of a service invocation device applied to a RIC platform according to an embodiment of the present disclosure;

[0033] FIG10 is a schematic structural diagram of a service invoking apparatus for a first application provided by an embodiment of the present disclosure;

[0034] FIG11 is a schematic structural diagram of a processing device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] The following will describe the technical solutions in the embodiments of the present disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0036] It should be noted that in the embodiments of the present disclosure, the term "and / or" is merely a description of the association relationship between associated objects, indicating that three possible relationships exist. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the embodiments of the present disclosure, the character " / " generally indicates that the associated objects are in an "or" relationship.

[0037] In the description of the embodiments of the present disclosure, the term "correspondence" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and indication, configuration and configuration, etc.

[0038] To facilitate understanding of the technical solutions of the embodiments of the present disclosure, the relevant technologies of the embodiments of the present disclosure are described below. The following relevant technologies are optional solutions that can be arbitrarily combined with the technical solutions of the embodiments of the present disclosure, and they all fall within the protection scope of the embodiments of the present disclosure.

[0039] The O-RAN Near-RT RIC (hereafter referred to as RIC) is a functional entity defined by the O-RAN Alliance based on existing 5G RAN standards. Its fundamental characteristic is the ability to optimize radio resource management through near-real-time, fine-grained data collection and control. Figure 1 shows the logical architecture of O-RAN. In the current O-RAN architecture, the RIC connects to other network functions via the A1, O1, Y1, and E2 interfaces. Specifically, the RIC collects radio resource data (such as cell configuration, user context, measurement reports, etc.) and service enhancement information (A1-EI) via the E2 and A1 interfaces, respectively.

[0040] The RIC is further divided into two functional entities: a single RIC platform and multiple wireless intelligent applications (xApps). Figure 2 shows the internal architecture of the RIC. xApps are lightweight software applications running on the RIC platform. They can be developed and deployed independently of the RIC platform. They are typically customized based on business applications or operator requirements and utilize artificial intelligence (AI) / machine learning (ML) algorithms to achieve specific optimization goals, such as optimizing the Quality of Service (QoS) / Quality of Experience (QoE) for a specific slice. The RIC platform is responsible for terminating external interfaces and providing general service support for xApps.

[0041] The RIC platform already supports / is expected to support the following services, allowing xApps to be called on demand:

[0042] (1) E2 subscription service: supports xApp to collect wireless resource data from E2 interface;

[0043] (2) A1 Subscription Service: supports xApp to collect business enhancement information from A1 interface;

[0044] (3) Model management service: supports xApp to register and deploy its own AI / ML model files to the RIC platform (RIC platform provides a runtime environment such as PyTorch) for calling, or discover AI / ML models provided by the RIC platform;

[0045] (4) Data pipeline service: supports preprocessing of raw E2 / A1 data to generate input data that can be directly used by AI / ML models;

[0046] (5) Model inference service;

[0047] (6) Model training services.

[0048] Figure 3 shows a baseline solution diagram for xApp using the RIC platform to implement a machine learning workflow. This diagram shows xApp using RIC platform services to implement the "data acquisition - data preprocessing - AI / ML model inference" workflow.

[0049] However, in the aforementioned baseline solution's loop (steps 3, 4, and 5), each step requires the participation of the xApp. When E2 / A1 data reports are frequent (10ms period), the xApp's signaling overhead is high. Furthermore, when signaling carries data (such as E2 / A1 data or pre-processed data), the xApp's interface forwarding data is also subject to significant transmission pressure. In other words, when data reports are frequent, the xApp's signaling overhead is high, and when signaling carries data, the xApp's interface forwarding data is also subject to significant transmission pressure.

[0050] To solve the above problems, the present disclosure proposes a process and interface design that supports xApp to combine and use the above services, thereby reducing the signaling overhead and transmission pressure of xApp when implementing the "data collection-data preprocessing-AI / ML model inference (or training)" workflow.

[0051] To facilitate understanding of the technical solutions of the embodiments of the present disclosure, the technical solutions of the present disclosure are described in detail below through specific embodiments. The above related technologies can be combined with the technical solutions of the embodiments of the present disclosure as optional solutions, and all of them fall within the scope of protection of the embodiments of the present disclosure. The embodiments of the present disclosure include at least part of the following contents.

[0052] The present disclosure provides a method for invoking a service on a RIC platform, wherein the RIC platform supports multiple services, and the multiple services are used by a first application to perform a machine learning workflow. FIG4 is a flow chart of the method for invoking a service on a RIC platform. As shown in FIG4 , the method may include the following steps:

[0053] Step 401: Receive configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on multiple services.

[0054] In the disclosed embodiments, the RIC platform receives configuration information from an xApp (first application) and uses this information to create tasks related to the machine learning workflow across multiple services. These services include a model processing service, a data pipeline service, and a data subscription service. Specifically, the model processing service can be a model inference service or a model training service, and the machine learning workflow is a "data acquisition - data preprocessing - model inference / training" workflow. In this disclosure, model inference / training is uniformly referred to as model processing.

[0055] In some embodiments, the multiple services also include a model management service. The RIC platform can receive a model file sent by a first application through the model management service, allowing the first application to deploy a model on the RIC platform based on the model file. The model is uniquely identified by a model identifier (ID). In some embodiments, the RIC platform can receive a model file sent by the first application through the model management service before receiving configuration information from the first application, allowing the first application to deploy a model on the RIC platform based on the model file.

[0056] In some implementations, step 401 may be implemented in two different ways (way 1 and way 2), where way 1 is implemented through steps s1-s3, and way 2 is implemented through steps t1-t4, specifically including:

[0057] Method 1

[0058] Step s1: Receive a first request sent by a first application through the model processing service, where the first request carries first configuration information and is used to request the model processing service to create a model processing task based on the first configuration information.

[0059] In an embodiment of the present disclosure, the model processing service of the RIC platform receives a first request sent by an xApp to request the creation of a model processing task, wherein the model processing task is uniquely identified by a model processing task ID and corresponds to a set of configuration information (first configuration information), wherein the first configuration information includes, but is not limited to: the model ID used, the processing result push or notification address, and the model version number, etc., wherein the processing result push or notification address includes, for example, a combination of an Internet Protocol (IP) address and a port number. That is, the first request is used to request the model processing task to create a model processing task based on the first configuration information, wherein the model processing task can be a model inference task or a model training task.

[0060] Step s2: Receive a second request sent by the first application through the data pipeline service, where the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information.

[0061] In an embodiment of the present disclosure, the data pipeline service of the RIC platform receives a second request from an xApp requesting the creation of a data pipeline task. The data pipeline task is uniquely identified by a data pipeline task ID and corresponds to a set of configuration information (second configuration information). The second configuration information includes, but is not limited to, instructions for preprocessing data (e.g., normalization, deduplication, and multi-source data association), a model processing task ID used when invoking the model processing service, and the like. In other words, the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information.

[0062] Step s3: Receive a third request sent by the first application through the data subscription service, where the third request carries third configuration information, and the third request is used to request the data subscription service to create a data subscription based on the third configuration information.

[0063] In the disclosed embodiment, the E2 subscription service / A1 subscription service of the RIC platform receives a third request sent by an xApp to request the creation of one or more E2 data subscriptions / A1 data subscriptions. The data subscription is uniquely identified by an E2 subscription ID / A1 subscription ID and corresponds to a set of configuration information (third configuration information). The third configuration information includes, but is not limited to, the type and reporting period / conditions of data collected by the E2 / A1 interface, and the data pipeline task ID used when invoking the data pipeline service. That is, the third request is used to request the data subscription service to create one or more E2 data subscriptions / A1 data subscriptions based on the third configuration information.

[0064] It can be understood that method 1 introduces the concepts of model processing tasks and data pipeline tasks in the model processing service and data pipeline service definitions of the RIC platform, and uses model processing task ID and data pipeline task ID for identification respectively, thereby realizing the series connection of the three services of "data acquisition-data preprocessing-model inference / training".

[0065] Method 2

[0066] Step t1: The model processing service receives a fourth request sent by the first application, where the fourth request carries configuration information and is used to request the model processing service to create a workflow task based on the configuration information.

[0067] Among them, the workflow tasks include model processing tasks, data pipeline tasks and data subscriptions. The model processing tasks can be model inference tasks or model training tasks. The configuration information includes: first configuration information, second configuration information and third configuration information.

[0068] In the disclosed embodiment, the RIC platform's model processing service receives a fourth request from the xApp to create a workflow task. The workflow task corresponds to a set of configuration information, including parameters such as, but not limited to, the model ID used, the model version used, the address to which the processing results are pushed, parameters required by the data pipeline service (e.g., commands for pre-processing data), and parameters required by the E2 / A1 subscription service (e.g., the type of data collected by the E2 / A1 interface and the reporting period / conditions). In other words, the fourth request is used to request the model processing service to create a workflow task based on the configuration information.

[0069] Step t2: Allocate a workflow task identifier through the model processing service and create a model processing task based on the first configuration information.

[0070] In an embodiment of the present disclosure, in response to the model processing service creating a workflow task, the model processing service may allocate a workflow task ID to itself and create the model processing task based on the first configuration information.

[0071] Step t3: Send second configuration information to the data pipeline service through the model processing service. The second configuration information is used by the data pipeline service to create a data pipeline task.

[0072] In an embodiment of the present disclosure, the model processing service interacts with the data pipeline service, and the model processing service sends second configuration information to the data pipeline service, so that the data pipeline service creates a data pipeline task based on the second configuration information, wherein the configuration parameters corresponding to the data pipeline task include but are not limited to: workflow task ID, operation instructions for pre-processing data in the configuration information, etc.

[0073] Step t4: Send third configuration information to the data subscription service through the model processing service. The third configuration information is used by the data subscription service to create a data subscription.

[0074] In an embodiment of the present disclosure, the model processing service interacts with the E2 subscription service / A1 subscription service, and sends third configuration information to the E2 subscription service / A1 subscription service through the model processing service, so that the E2 subscription service / A1 subscription service creates a data subscription based on the third configuration information, wherein the configuration parameters corresponding to the data subscription include but are not limited to: workflow task ID, the type of E2 / A1 data in the configuration information, and the reporting cycle / conditions.

[0075] It should be noted that in response to the model processing service creating a workflow task and creating a model processing task, as well as creating a data pipeline task and data subscription through interaction with the data pipeline service and the data subscription service, the model processing service can return the assigned workflow task ID to the xApp.

[0076] It can be understood that the second method introduces the concept of workflow tasks in the case of new or existing service definitions on the RIC platform, and uses workflow task IDs for identification, thereby realizing the concatenation of the three services of "data collection-data preprocessing-model reasoning / training".

[0077] Step 402: Call multiple services to perform tasks and obtain machine learning results.

[0078] In an embodiment of the present disclosure, in response to the RIC platform using configuration information to create a task related to a machine learning workflow on multiple services, the RIC platform can call multiple services in the RIC platform to execute the task to obtain a machine learning result, where the machine learning result is an inference result or a training result.

[0079] In some embodiments, step 402 may be implemented in two different ways (way 3 and way 4). Way 3 is implemented through steps w1-w3, and way 4 is implemented through steps z1-z6, specifically including:

[0080] Method 3

[0081] Step w1: Subscribe to data through the data subscription service to obtain the collected data.

[0082] In the embodiment of the present disclosure, data subscription is performed through the E2 subscription service / A1 subscription service to obtain the collected E2 / A1 data.

[0083] Step w2: Send the collected data and the data pipeline task identifier or workflow task identifier to the data pipeline service through the data subscription service, and pre-process the collected data based on the data pipeline task identifier or workflow task identifier through the data pipeline service.

[0084] In the disclosed embodiment, the E2 subscription service / A1 subscription service of the RIC platform can send the collected E2 / A1 data (or the address identifier where the data is stored) and the data pipeline task ID or workflow task ID used to process the data to the data pipeline service. Therefore, the data pipeline service can pre-process the collected E2 / A1 data based on the data pipeline task ID or workflow task ID.

[0085] That is to say, in some embodiments, the E2 subscription service / A1 subscription service of the RIC platform can send the collected E2 / A1 data (or the address identifier where the data is stored) and the data pipeline task ID used to process the data to the data pipeline service. Therefore, the data pipeline service can pre-process the collected E2 / A1 data based on the data pipeline task ID.

[0086] In some embodiments, the E2 subscription service / A1 subscription service of the RIC platform can send the collected E2 / A1 data (or the address identifier where the data is stored) and the workflow task ID used to process the data to the data pipeline service. Therefore, the data pipeline service can pre-process the collected E2 / A1 data based on the workflow task ID.

[0087] Step w3: Send the preprocessed data and the model processing task identifier or workflow task identifier to the model processing service through the data pipeline service. The model processing service performs model processing on the preprocessed data based on the model processing task identifier or workflow task identifier to generate machine learning results.

[0088] In the disclosed embodiment, the data pipeline service of the RIC platform can send preprocessed data (or the address identifier where the data is stored) and the model processing task ID or workflow task ID used for model processing to the model processing service. Therefore, the model processing service can perform model processing on the preprocessed data based on the model processing task ID or workflow task ID to generate machine learning results.

[0089] That is, in some embodiments, the data pipeline service of the RIC platform can send preprocessed data (or the address identifier where the data is stored) and the model processing task ID used for model processing to the model processing service. Therefore, the model processing service can perform model processing on the preprocessed data based on the model processing task ID to generate machine learning results.

[0090] In some embodiments, the data pipeline service of the RIC platform can send preprocessed data (or the address identifier where the data is stored) and the workflow task ID used for model processing to the model processing service. Therefore, the model processing service can perform model processing on the preprocessed data based on the workflow task ID to generate machine learning results.

[0091] Method 4

[0092] Step z1: Receive the fifth request sent by the first application through the model processing service. The fifth request carries the model processing task identifier. The fifth request is used to request the machine learning result corresponding to the model processing task identifier from the model processing service.

[0093] In an embodiment of the present disclosure, the model processing service of the RIC platform can receive the fifth request sent by xApp, wherein the fifth request carries the model processing task ID. Therefore, xApp can use the fifth request to request the model processing service for the machine learning result corresponding to the model processing task ID.

[0094] Step z2: Send the data pipeline task identifier to the data pipeline service through the model processing service.

[0095] In an embodiment of the present disclosure, the model processing service of the RIC platform can send a data pipeline task ID to the data pipeline service, so that the data pipeline service can subsequently pre-process the collected data based on the data pipeline task ID.

[0096] Step z3: Send the subscription identifier to the data subscription service through the data pipeline service.

[0097] In an embodiment of the present disclosure, the data pipeline service of the RIC platform can send the E2 subscription ID / A1 subscription ID to the data subscription service, so that the data subscription service can subsequently request to collect data corresponding to the E2 subscription ID / A1 subscription ID.

[0098] Step z4: Send the collected data to the data pipeline service based on the subscription identifier through the data subscription service.

[0099] In an embodiment of the present disclosure, in response to the data subscription service receiving a subscription identifier sent by the data pipeline service, such as E2 subscription ID / A1 subscription ID, the data subscription service can collect data corresponding to the E2 subscription ID / A1 subscription ID and return the E2 / A1 data corresponding to the E2 subscription ID / A1 subscription ID to the data pipeline service.

[0100] Step z5: The data pipeline service preprocesses the collected data based on the data pipeline task identifier, and sends the preprocessed data to the model processing service.

[0101] In the disclosed embodiment, the data pipeline service of the RIC platform can pre-process the E2 / A1 data according to the data pipeline task ID obtained in step z2, and return the pre-processed data to the model processing service.

[0102] Step z6: Based on the model processing task identifier, the model processing service performs model processing on the preprocessed data to generate machine learning results.

[0103] In the disclosed embodiment, the model processing service of the RIC platform can perform model processing (model inference / training) on ​​the preprocessed data according to the model processing task ID obtained in step z1, and generate machine learning results.

[0104] Step 403: Send the machine learning result to the first application.

[0105] In the embodiment of the present disclosure, after xApp requests multiple services (data subscription service, data pipeline service, and model processing service) of the RIC platform to create tasks related to the machine learning workflow (data subscription, data pipeline tasks, and model processing tasks), the RIC platform can call these multiple services to perform tasks related to the machine learning workflow and obtain machine learning results, and then send the machine learning results to xApp through the model processing service.

[0106] The technical solutions of the embodiments of the present disclosure are illustrated below with reference to specific application examples.

[0107] Application Example 1

[0108] Figure 5 shows a schematic diagram of the "data collection - data preprocessing - model inference / training" workflow for upstream services to call downstream services. The task configuration and call processing processes in this solution can be implemented using the steps of Methods 1 and 3 above. As shown in Figure 5, the process includes the following steps:

[0109] Step 501: xAPP deploys the model to the RIC platform.

[0110] Step 502: xApp initiates a request to the model reasoning service of the RIC platform to request the creation of a model reasoning task. The model reasoning task is uniquely identified by a model reasoning task ID, which corresponds to a set of reasoning configuration information.

[0111] The inference configuration information includes, but is not limited to, the Model ID of the model used, the address for pushing or notifying the inference result (e.g., a combination of an IP address and a port number), and may also include the version number of the model used.

[0112] Step 503: xApp initiates a request to the data pipeline service of the RIC platform to request the creation of a data pipeline task. The data pipeline task is uniquely identified by a data pipeline task ID and corresponds to a set of data pipeline configuration information.

[0113] The data pipeline configuration information includes but is not limited to: operation instructions for preprocessing data (such as normalization, deduplication, and multi-source data association) and the model inference task ID used when calling the model inference service.

[0114] Step 504: xApp initiates a request to the E2 subscription service / A1 subscription service of the RIC platform to request the creation of one or more E2 data subscriptions / A1 data subscriptions. The subscription is uniquely identified by an E2 subscription ID / A1 subscription ID and corresponds to a set of data collection configuration information.

[0115] The data collection configuration information includes but is not limited to: the type and reporting cycle / condition of data collected by the E2 / A1 interface, and the data pipeline task ID used when calling the data pipeline service.

[0116] Step 505: The E2 subscription service / A1 subscription service initiates a request to the data pipeline service, carrying the data collected by the E2 / A1 interface (or the address identifier where the data is stored) and the data pipeline task ID used to process the data; the data pipeline service pre-processes the data based on the configuration information corresponding to the data pipeline task ID.

[0117] Step 506: The data pipeline service initiates a request to the model inference service, carrying the preprocessed data (or the address identifier where the data is stored) and the model inference task ID used for model inference; the model inference service infers the preprocessed data based on the configuration information corresponding to the model inference task ID and generates an inference result.

[0118] Step 507: The model inference service notifies the xApp and carries the inference result (or the address identifier of the data storage).

[0119] Application Example 2

[0120] Figure 6 shows a schematic diagram of the "data collection - data preprocessing - model inference / training" workflow for downstream services to call upstream services. The task configuration and call processing processes in this solution can be implemented using the steps of Methods 1 and 4 above. As shown in Figure 6, the process includes the following steps:

[0121] Step 601: xAPP deploys the model to the RIC platform.

[0122] Step 602: xApp initiates a request to the E2 subscription service / A1 subscription service of the RIC platform to request the creation of one or more E2 data subscriptions / A1 data subscriptions. The subscription is uniquely identified by an E2 subscription ID / A1 subscription ID and corresponds to a set of data collection configuration information.

[0123] The data collection configuration information includes but is not limited to: the type and reporting period / condition of data collected by the E2 / A1 interface, and the cache address of the E2 / A1 data (eg, a specific space in the RIC platform database).

[0124] Step 603: xApp initiates a request to the data pipeline service of the RIC platform to request the creation of a data pipeline task. The data pipeline task is uniquely identified by a data pipeline task ID and corresponds to a set of data pipeline configuration information.

[0125] The data pipeline configuration information includes but is not limited to: operation instructions for pre-processing data (such as normalization, deduplication, and multi-source data association), and the E2 subscription ID / A1 subscription ID used when obtaining E2 / A1 data from the E2 subscription service / A1 subscription service.

[0126] Step 604: xApp initiates a request to the model reasoning service of the RIC platform to request the creation of a model reasoning task. The model reasoning task is uniquely identified by a model reasoning task ID, which corresponds to a set of reasoning configuration information.

[0127] The inference configuration information includes, but is not limited to, the Model ID of the model used and the data pipeline task ID used when calling the data pipeline task; in addition, it may also include the version number of the model used.

[0128] Step 605: xApp initiates an inference request to the model inference service, carrying the model inference task ID.

[0129] Step 606: The model inference service initiates a data preprocessing request to the data pipeline service, carrying the data pipeline task ID.

[0130] Step 607: The data pipeline service initiates an E2 / A1 data request to the E2 subscription service / A1 subscription service, carrying the E2 subscription ID / A1 subscription ID.

[0131] Step 608: The E2 subscription service / A1 subscription service returns the E2 / A1 data corresponding to the E2 subscription ID / A1 subscription ID to the data pipeline service.

[0132] Step 609: The data pipeline service preprocesses the E2 data / A1 data according to the data pipeline task ID obtained in step 606, and returns the preprocessed data to the model inference service.

[0133] Step 610: The model inference service performs model inference on the preprocessed data according to the model inference task ID obtained in step 605, and returns the inference result to xApp.

[0134] Application Example 3

[0135] For example, Figure 7 shows a schematic diagram of a workflow that implements multiple services in series through a single service entry. The task configuration process and call processing process in this solution can be implemented by the steps of the above-mentioned methods 2 and 3. As shown in Figure 7, the process includes the following steps:

[0136] Step 701: xAPP deploys the model to the RIC platform.

[0137] Step 702: The xApp initiates a request to the model reasoning service of the RIC platform to establish a workflow task corresponding to a set of configuration parameters.

[0138] The configuration parameters include but are not limited to: the identifier Model ID of the model used.

[0139] Additionally, this may include:

[0140] The version number of the model used;

[0141] The address where the inference results are pushed;

[0142] Parameters required by the data pipeline service: for example, instructions for preprocessing data (such as normalization, deduplication, and multi-source data association);

[0143] Parameters required for the E2 subscription service / A1 subscription service, such as the type of data collected by the E2 / A1 interface and the reporting period / conditions.

[0144] Step 703: The model inference service assigns a workflow task ID.

[0145] Step 704: The model inference service interacts with the data pipeline service and configures the data pipeline task based on the configuration parameters. The parameters used to configure the data pipeline task include: workflow task ID and operation instructions for pre-processing data.

[0146] Step 705: The model inference service interacts with the E2 subscription service / A1 subscription service and configures data subscription based on configuration parameters. The parameters used in configuring data subscription include: workflow task ID, required E2 / A1 data type, and reporting period / conditions.

[0147] Step 706: Return the workflow ID to the xApp.

[0148] Step 707: The E2 subscription service / A1 subscription service calls the data pipeline service to push the E2 / A1 data and carries the workflow task ID.

[0149] Step 708: After preprocessing the E2 / A1 data, the data pipeline service calls the model inference service, pushes the preprocessed data, and carries the workflow task ID.

[0150] Step 709: The model inference service performs model inference on the pre-processed data according to the workflow task ID obtained in step 708, and returns the inference result to the xApp along with the workflow task ID.

[0151] In addition, Method 2 and Method 4 can also be combined to form a new solution to implement the "data collection-data preprocessing-AI / ML model inference / training" workflow. For the sake of brevity, we will not go into details here.

[0152] It should be noted that after creating a machine learning workflow, that is, creating the "data acquisition-data preprocessing-model inference / training" workflow, the model processing service or the data pipeline service can respectively receive requests sent by xApp to request changes to the configuration information of the corresponding model processing task or data pipeline task (indicated by the model processing task ID or the data pipeline task ID).

[0153] Furthermore, at the end of the machine learning workflow, the model processing service or the data pipeline service may also receive a request from the xApp to delete the corresponding model processing task or data pipeline task (indicated by the model processing task ID or the data pipeline task ID).

[0154] In the technical solution of the embodiment of the present disclosure, the RIC platform receives configuration information sent by the first application. This configuration information is used to create tasks related to the machine learning workflow on multiple services supported by the RIC platform. Then, the RIC platform calls multiple services to perform tasks related to the machine learning workflow, obtains machine learning results, and then sends the machine learning results to the first application. In this way, in the process of implementing the machine learning workflow, the first application pre-configures the relevant services, enabling continuous calls between multiple services without requiring the participation of the first application at every step. This greatly reduces the signaling overhead of the first application and the transmission pressure on the first application-side interface when forwarding data.

[0155] The present disclosure also provides a service invocation method for a first application. FIG8 is a flow chart of the service invocation method according to the present disclosure. As shown in FIG8 , the method may include the following steps:

[0156] Step 801: Send configuration information to the RIC platform.

[0157] Among them, the RIC platform supports multiple services, multiple services are used for the first application to perform machine learning workflows, and configuration information is used to create tasks related to machine learning workflows on multiple services.

[0158] In the disclosed embodiments, the xApp sends configuration information for creating tasks related to machine learning workflows to the RIC platform, enabling multiple services within the RIC platform to create corresponding tasks based on the configuration information. These services include a model processing service, a data pipeline service, and a data subscription service. Specifically, the model processing service can be a model inference service or a model training service.

[0159] In some embodiments, the multiple services also include a model management service. The xApp can send a model file to the RIC platform's model management service and deploy the model on the RIC platform based on the model file. The model is uniquely identified by a model ID. In some embodiments, the xApp can send a model file to the RIC platform's model management service before sending configuration information to the RIC platform and deploy the model on the RIC platform based on the model file.

[0160] In some implementations, step 801 may be implemented in two different ways (way 5 and way 6), where way 5 is implemented through steps x1-x3, and way 6 is implemented through step y1, specifically including:

[0161] Method 5

[0162] Step x1: Send a first request to the model processing service, where the first request carries first configuration information. The first request is used to request the model processing service to create a model processing task based on the first configuration information.

[0163] In the disclosed embodiment, the xApp sends a first request to the RIC platform's model processing service to create a model processing task. The model processing task is uniquely identified by a model processing task ID and corresponds to a set of configuration information (first configuration information). The first configuration information includes, but is not limited to, the model ID used, the address for pushing or notifying processing results (e.g., an IP address + port number combination), and the model version number. That is, the first request is used to request the model processing service to create a model processing task based on the first configuration information. The model processing task can be a model inference task or a model training task.

[0164] Step x2: Send a second request to the data pipeline service, where the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information.

[0165] In the disclosed embodiment, the xApp sends a second request to the data pipeline service of the RIC platform to request the creation of a data pipeline task. The data pipeline task is uniquely identified by a data pipeline task ID and corresponds to a set of configuration information (second configuration information). The second configuration information includes, but is not limited to, instructions for preprocessing data (e.g., normalization, deduplication, and multi-source data association), the model processing task ID used when invoking the model processing service, and the like. In other words, the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information.

[0166] Step x3: Send a third request to the data subscription service, where the third request carries third configuration information. The third request is used to request the data subscription service to create a data subscription based on the third configuration information.

[0167] In the disclosed embodiment, the xApp sends a third request to the data subscription service of the RIC platform to request the creation of one or more E2 data subscriptions / A1 data subscriptions. The data subscription is uniquely identified by an E2 subscription ID / A1 subscription ID and corresponds to a set of configuration information (third configuration information). This third configuration information includes, but is not limited to, the type and reporting period / conditions of data collected by the E2 / A1 interface, and the data pipeline task ID used when invoking the data pipeline service. In other words, the third request is used to request the data subscription service to create one or more E2 data subscriptions / A1 data subscriptions based on the third configuration information.

[0168] Method 6

[0169] Step y1: Send a fourth request to the model processing service, where the fourth request carries configuration information and is used to request the model processing service to create a workflow task based on the configuration information.

[0170] In the disclosed embodiment, the xApp sends a fourth request to the RIC platform's model processing service to request the creation of a workflow task. The workflow task corresponds to a set of configuration information, including, but not limited to, the model ID used, the model version used, the address to which the processing results are pushed, parameters required by the data pipeline service (such as the operation command for preprocessing data), and parameters required by the E2 subscription service / A1 subscription service (the type of data collected by the E2 / A1 interface and the reporting period / conditions). In other words, the fourth request is used to request the model processing service to create a workflow task based on the configuration information.

[0171] In some implementations, workflow tasks include: model processing tasks, data pipeline tasks, and data subscriptions. The model processing tasks can be model inference tasks or model training tasks. The configuration information includes: first configuration information, second configuration information, and third configuration information. The model processing tasks are created by the model processing service based on the first configuration information; the data pipeline tasks are created by the data pipeline service based on the second configuration information; and the data subscriptions are created by the data subscription service based on the third configuration information.

[0172] Step 802: Receive the machine learning results sent by the RIC platform. The machine learning results are obtained by the RIC platform calling multiple services to execute tasks.

[0173] In an embodiment of the present disclosure, xApp sends configuration information for creating tasks related to machine learning workflows to the RIC platform, so that multiple services in the RIC platform create their own corresponding tasks based on the configuration information, so that the RIC platform can call these multiple services to perform tasks related to the machine learning workflow and obtain machine learning results, so that xApp can receive the machine learning results sent by the model processing service in the RIC platform.

[0174] It should be noted that after creating a machine learning workflow, that is, creating the "data acquisition-data preprocessing-model inference / training" workflow, xApp can initiate a request to the model processing service or the data pipeline service respectively to request to change the configuration information of the corresponding model processing task or data pipeline task (indicated by the model processing task ID or the data pipeline task ID).

[0175] Furthermore, at the end of the machine learning workflow, the xApp can initiate a request to the model processing service or the data pipeline service to request the deletion of the corresponding model processing task or data pipeline task (indicated by the model processing task ID or the data pipeline task ID).

[0176] In the technical solution of the embodiment of the present disclosure, the RIC platform receives configuration information sent by the first application. This configuration information is used to create tasks related to the machine learning workflow on multiple services supported by the RIC platform. Then, the RIC platform calls multiple services to perform tasks related to the machine learning workflow, obtains machine learning results, and then sends the machine learning results to the first application. In this way, in the process of implementing the machine learning workflow, the first application pre-configures the relevant services, enabling continuous calls between multiple services without requiring the participation of the first application at every step. This greatly reduces the signaling overhead of the first application and the transmission pressure on the first application-side interface when forwarding data.

[0177] The present disclosure also proposes a service invocation device for a RIC platform, wherein the RIC platform supports multiple services, and the multiple services are used by a first application to perform a machine learning workflow. FIG9 is a schematic diagram of the structure of the service invocation device of the present disclosure embodiment. As shown in FIG9, the device includes:

[0178] The first receiving unit 901 is used to receive configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on multiple services.

[0179] The first calling unit 902 is used to call multiple services to perform tasks and obtain machine learning results.

[0180] The first sending unit 903 is used to send the machine learning result to the first application.

[0181] In some embodiments, the first receiving unit 901 is specifically used to receive a first request sent by a first application through a model processing service, the first request carries first configuration information, and the first request is used to request the model processing service to create a model processing task based on the first configuration information; receive a second request sent by the first application through a data pipeline service, the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information; receive a third request sent by the first application through a data subscription service, the third request carries third configuration information, and the third request is used to request the data subscription service to create a data subscription based on the third configuration information; wherein the multiple services include: a model processing service, a data pipeline service, and a data subscription service, the model processing service is a model inference service or a model training service, and the model processing task is a model inference task or a model training task.

[0182] In some embodiments, the first receiving unit 901 is further specifically used to receive a fourth request sent by the first application through the model processing service, where the fourth request carries configuration information, and the fourth request is used to request the model processing service to create a workflow task based on the configuration information; wherein the multiple services include: a model processing service, a data pipeline service, and a data subscription service, and the model processing service is a model inference service or a model training service.

[0183] In some embodiments, the first receiving unit 901 is further specifically used to assign a workflow task identifier through the model processing service, and create a model processing task based on the first configuration information; send second configuration information to the data pipeline service through the model processing service, and the second configuration information is used for the data pipeline service to create a data pipeline task; send third configuration information to the data subscription service through the model processing service, and the third configuration information is used for the data subscription service to create a data subscription; wherein, the workflow task includes: model processing task, data pipeline task and data subscription; the model processing task is a model inference task or a model training task; the configuration information includes: first configuration information, second configuration information and third configuration information.

[0184] In some embodiments, the first calling unit 902 is specifically used to subscribe to data through a data subscription service to obtain collected data; send the collected data and a data pipeline task identifier or a workflow task identifier to a data pipeline service through the data subscription service, and preprocess the collected data based on the data pipeline task identifier or the workflow task identifier through the data pipeline service; send the preprocessed data and a model processing task identifier or a workflow task identifier to a model processing service through the data pipeline service, and perform model processing on the preprocessed data based on the model processing task identifier or the workflow task identifier through the model processing service to generate a machine learning result.

[0185] In some embodiments, the first calling unit 902 is further specifically used to receive a fifth request sent by the first application through the model processing service, the fifth request carries a model processing task identifier, and the fifth request is used to request the model processing service for a machine learning result corresponding to the model processing task identifier; send the data pipeline task identifier to the data pipeline service through the model processing service; send the subscription identifier to the data subscription service through the data pipeline service; send the collected data to the data pipeline service based on the subscription identifier through the data subscription service; pre-process the collected data based on the data pipeline task identifier through the data pipeline service, and send the pre-processed data to the model processing service; perform model processing on the pre-processed data based on the model processing task identifier through the model processing service to generate a machine learning result.

[0186] In some embodiments, the first sending unit 903 is specifically configured to send the machine learning result to the first application through the model processing service.

[0187] In some embodiments, the device further comprises:

[0188] The first deployment unit is used to receive the model file sent by the first application through the model management service, so that the first application can deploy the model on the RIC platform based on the model file; wherein the multiple services also include: model management service.

[0189] In the technical solution of the embodiment of the present disclosure, the RIC platform receives configuration information sent by the first application. This configuration information is used to create tasks related to the machine learning workflow on multiple services supported by the RIC platform. Then, the RIC platform calls multiple services to perform tasks related to the machine learning workflow, obtains machine learning results, and then sends the machine learning results to the first application. In this way, in the process of implementing the machine learning workflow, the first application pre-configures the relevant services, enabling continuous calls between multiple services without requiring the participation of the first application at every step. This greatly reduces the signaling overhead of the first application and the transmission pressure on the first application-side interface when forwarding data.

[0190] The present disclosure also provides a service invocation device for a first application. FIG10 is a schematic diagram of the structure of the service invocation device according to the present disclosure. As shown in FIG10 , the device includes:

[0191] The second sending unit 1001 is used to send configuration information to the RIC platform. The RIC platform supports multiple services. The multiple services are used for the first application to perform machine learning workflows. The configuration information is used to create tasks related to the machine learning workflows on the multiple services.

[0192] The second receiving unit 1002 is used to receive the machine learning results sent by the RIC platform. The machine learning results are obtained by the RIC platform calling multiple services to execute tasks.

[0193] In some embodiments, the second sending unit 1001 is specifically used to send a first request to the model processing service, the first request carries first configuration information, and the first request is used to request the model processing service to create a model processing task based on the first configuration information; send a second request to the data pipeline service, the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information; send a third request to the data subscription service, the third request carries third configuration information, and the third request is used to request the data subscription service to create a data subscription based on the third configuration information; wherein, the multiple services include: model processing service, data pipeline service and data subscription service, the model processing service is a model inference service or a model training service, and the model processing task is a model inference task or a model training task.

[0194] In some embodiments, the second sending unit 1001 is further specifically used to send a fourth request to the model processing service, the fourth request carries configuration information, and the fourth request is used to request the model processing service to create a workflow task based on the configuration information; wherein, the multiple services include: model processing service, data pipeline service and data subscription service, the model processing service is a model inference service or a model training service, the workflow tasks include: model processing tasks, data pipeline tasks and data subscription, and the model processing tasks are model inference tasks or model training tasks; the configuration information includes: first configuration information, second configuration information and third configuration information; wherein, the model processing task is created by the model processing service based on the first configuration information; the data pipeline task is created by the data pipeline service based on the second configuration information; and the data subscription is created by the data subscription service based on the third configuration information.

[0195] In some embodiments, the second receiving unit 1002 is specifically used to receive the machine learning results sent by the model processing service.

[0196] In some embodiments, the device further comprises:

[0197] The second deployment unit is used to send the model file to the model management service and deploy the model on the RIC platform based on the model file; wherein the multiple services also include: model management service.

[0198] In the technical solution of the embodiment of the present disclosure, the RIC platform receives configuration information sent by the first application. This configuration information is used to create tasks related to the machine learning workflow on multiple services supported by the RIC platform. Then, the RIC platform calls multiple services to perform tasks related to the machine learning workflow, obtains machine learning results, and then sends the machine learning results to the first application. In this way, in the process of implementing the machine learning workflow, the first application pre-configures the relevant services, enabling continuous calls between multiple services without requiring the participation of the first application at every step. This greatly reduces the signaling overhead of the first application and the transmission pressure on the first application-side interface when forwarding data.

[0199] Those skilled in the art will appreciate that the functions implemented by each unit in the service invocation apparatus shown in Figures 9 and 10 can be understood by referring to the relevant description of the aforementioned method. The functions of each unit in the service invocation apparatus shown in Figures 9 and 10 can be implemented by a program running on a processor or by a specific logic circuit.

[0200] Figure 11 is a schematic diagram of the structure of a processing device provided in an embodiment of the present disclosure. The processing device can be a terminal device or a network device. The processing device shown in Figure 11 includes a processor 1101, which can call and run a computer program from a memory to implement the method in the embodiment of the present disclosure.

[0201] In some embodiments, as shown in FIG11 , the processing device may further include a memory 1102. The processor 1101 may call and execute a computer program from the memory 1102 to implement the method in the embodiment of the present disclosure.

[0202] The memory 1102 may be a separate device independent of the processor 1101 , or may be integrated into the processor 1101 .

[0203] In some embodiments, as shown in FIG11 , the processing device may further include a transceiver 1103 , and the processor 1101 may control the transceiver 1103 to communicate with other devices. Specifically, the transceiver 1103 may send information or data to other devices, or receive information or data sent by other devices.

[0204] The transceiver 1103 may include a transmitter and a receiver. The transceiver 1103 may further include an antenna, and the number of antennas may be one or more.

[0205] In some embodiments, the processing device may specifically be the RIC platform of the embodiment of the present disclosure, and the processing device may implement the corresponding processes implemented by the RIC platform in each method of the embodiment of the present disclosure, which will not be described here for the sake of brevity.

[0206] In some embodiments, the processing device may specifically be the first application / xApp of the embodiments of the present disclosure, and the processing device may implement the corresponding processes implemented by the first application / xApp in each method of the embodiments of the present disclosure. For the sake of brevity, they are not described here.

[0207] It is understood that the memory in the embodiments of the present disclosure may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0208] It should be understood that the above-mentioned memories are exemplary but not restrictive. For example, the memories in the embodiments of the present disclosure may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM RAM (DR RAM), etc. In other words, the memories in the embodiments of the present disclosure are intended to include, but are not limited to, these and any other suitable types of memories.

[0209] The embodiment of the present disclosure also provides a computer-readable storage medium for storing a computer program.

[0210] In some embodiments, the computer-readable storage medium can be applied to the RIC platform in the embodiments of the present disclosure, and the computer program enables the computer to execute the corresponding processes implemented by the RIC platform in the various methods of the embodiments of the present disclosure. For the sake of brevity, they will not be repeated here.

[0211] In some embodiments, the computer-readable storage medium can be applied to the first application / xApp in the embodiments of the present disclosure, and the computer program enables the computer to execute the corresponding processes implemented by the first application / xApp in the various methods of the embodiments of the present disclosure. For the sake of brevity, they are not repeated here.

[0212] An embodiment of the present disclosure also provides a computer program product, including computer program instructions.

[0213] In some embodiments, the computer program product can be applied to the RIC platform in the embodiments of the present disclosure, and the computer program instructions enable the computer to execute the corresponding processes implemented by the RIC platform in the various methods of the embodiments of the present disclosure. For the sake of brevity, they will not be repeated here.

[0214] In some embodiments, the computer program product can be applied to the first application / xApp in the embodiments of the present disclosure, and the computer program instructions enable the computer to execute the corresponding processes implemented by the first application / xApp in the various methods of the embodiments of the present disclosure. For the sake of brevity, they are not repeated here.

[0215] The embodiments of the present disclosure also provide a computer program.

[0216] In some embodiments, the computer program can be applied to the RIC platform in the embodiments of the present disclosure. When the computer program runs on a computer, it enables the computer to execute the corresponding processes implemented by the RIC platform in the various methods of the embodiments of the present disclosure. For the sake of brevity, they will not be repeated here.

[0217] In some embodiments, the computer program can be applied to the first application / xApp in the embodiments of the present disclosure. When the computer program runs on a computer, it enables the computer to execute the corresponding processes implemented by the first application / xApp in the various methods of the embodiments of the present disclosure. For the sake of brevity, they are not further described here.

[0218] All embodiments of the present disclosure may be implemented individually or in combination with other embodiments, and are all considered to be within the scope of protection claimed by the present disclosure.

[0219] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0220] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0221] In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0222] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0223] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0224] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0225] The above description is merely a specific embodiment of the present disclosure, but the scope of protection of the present disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this disclosure should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims.

Claims

1. A service invocation method, applied to a radio access network intelligent controller (RIC) platform, wherein the RIC platform supports multiple services, wherein the multiple services are used by a first application to perform a machine learning workflow; the method comprising: receiving configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on the multiple services; Calling the multiple services to perform the task and obtain machine learning results; Send the machine learning result to the first application.

2. The method according to claim 1, wherein The multiple services include: model processing service, data pipeline service and data subscription service; the model processing service is a model inference service or a model training service; The receiving the configuration information sent by the first application includes: receiving, through the model processing service, a first request sent by the first application, the first request carrying first configuration information, the first request being used to request the model processing service to create a model processing task based on the first configuration information; the model processing task being a model inference task or a model training task; receiving, through the data pipeline service, a second request sent by the first application, where the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information; A third request sent by the first application is received through the data subscription service, where the third request carries third configuration information, and the third request is used to request the data subscription service to create a data subscription based on the third configuration information.

3. The method according to claim 1, wherein The multiple services include: model processing service, data pipeline service and data subscription service; the model processing service is a model inference service or a model training service; The receiving the configuration information sent by the first application includes: A fourth request sent by the first application is received through the model processing service, where the fourth request carries the configuration information, and the fourth request is used to request the model processing service to create a workflow task based on the configuration information.

4. The method according to claim 3, wherein: The workflow tasks include: model processing tasks, data pipeline tasks, and data subscriptions; the model processing tasks are model inference tasks or model training tasks; the configuration information includes: first configuration information, second configuration information, and third configuration information; the method further includes: allocating a workflow task identifier through the model processing service, and creating a model processing task based on the first configuration information; Sending the second configuration information to the data pipeline service through the model processing service, where the second configuration information is used by the data pipeline service to create a data pipeline task; The third configuration information is sent to the data subscription service through the model processing service, and the third configuration information is used by the data subscription service to create a data subscription.

5. The method according to any one of claims 2 to 4, wherein The calling of the multiple services to perform the task and obtain the machine learning result includes: Subscribe to data through the data subscription service to obtain the collected data; sending the collected data and a data pipeline task identifier or a workflow task identifier to the data pipeline service through the data subscription service, and preprocessing the collected data based on the data pipeline task identifier or the workflow task identifier through the data pipeline service; The preprocessed data and the model processing task identifier or the workflow task identifier are sent to the model processing service through the data pipeline service, and the model processing service performs model processing on the preprocessed data based on the model processing task identifier or the workflow task identifier to generate the machine learning result.

6. The method according to any one of claims 2 to 4, wherein: The calling of the multiple services to perform the task and obtain the machine learning result includes: receiving, through the model processing service, a fifth request sent by the first application, the fifth request carrying a model processing task identifier, and the fifth request being used to request the model processing service for a machine learning result corresponding to the model processing task identifier; Sending a data pipeline task identifier to the data pipeline service through the model processing service; Sending a subscription identifier to the data subscription service through the data pipeline service; Sending the collected data to the data pipeline service based on the subscription identifier through the data subscription service; preprocessing the collected data based on the data pipeline task identifier by the data pipeline service, and sending the preprocessed data to the model processing service; The model processing service performs model processing on the preprocessed data based on the model processing task identifier to generate the machine learning result.

7. The method according to any one of claims 2 to 6, wherein The sending the machine learning result to the first application includes: The machine learning result is sent to the first application through the model processing service.

8. The method according to any one of claims 2 to 7, wherein The plurality of services further include: a model management service; the method further includes: The model file sent by the first application is received through the model management service, so that the first application deploys the model on the RIC platform based on the model file.

9. A service invocation method, using a first application, comprising: Sending configuration information to a RIC platform, where the RIC platform supports multiple services, the multiple services are used for the first application to perform a machine learning workflow, and the configuration information is used to create tasks related to the machine learning workflow on the multiple services; Receive the machine learning results sent by the RIC platform, where the machine learning results are obtained by the RIC platform calling the multiple services to perform the task.

10. The method according to claim 9, wherein: The multiple services include: model processing service, data pipeline service and data subscription service; the model processing service is a model inference service or a model training service; The sending of configuration information to the RIC platform includes: Sending a first request to the model processing service, where the first request carries first configuration information, and the first request is used to request the model processing service to create a model processing task based on the first configuration information; the model processing task is a model inference task or a model training task; Sending a second request to the data pipeline service, where the second request carries second configuration information, and the second request is used to request the data pipeline service to create a data pipeline task based on the second configuration information; A third request is sent to the data subscription service, where the third request carries third configuration information, and the third request is used to request the data subscription service to create a data subscription based on the third configuration information.

11. The method according to claim 9, wherein The multiple services include: model processing service, data pipeline service and data subscription service; the model processing service is a model inference service or a model training service; The sending of configuration information to the RIC platform includes: A fourth request is sent to the model processing service, where the fourth request carries configuration information and is used to request the model processing service to create a workflow task based on the configuration information.

12. The method according to claim 11, wherein The workflow tasks include: model processing tasks, data pipeline tasks and data subscription; the model processing tasks are model inference tasks or model training tasks; the configuration information includes: first configuration information, second configuration information and third configuration information; wherein, The model processing task is created by the model processing service based on the first configuration information; The data pipeline task is created by the data pipeline service based on the second configuration information; The data subscription is created by the data subscription service based on the third configuration information.

13. The method according to any one of claims 10 to 12, wherein The receiving of the machine learning results sent by the RIC platform includes: Receive the machine learning result sent by the model processing service.

14. The method according to any one of claims 10 to 13, wherein The plurality of services further include: a model management service; the method further includes: The model file is sent to the model management service, and the model is deployed on the RIC platform based on the model file.

15. A service invocation device, applied to a RIC platform, wherein the RIC platform supports multiple services, wherein the multiple services are used by a first application to perform a machine learning workflow; The device comprises: a first receiving unit, configured to receive configuration information sent by the first application, where the configuration information is used to create tasks related to the machine learning workflow on the multiple services; A first calling unit is configured to call the multiple services to execute the task and obtain a machine learning result; A first sending unit is used to send the machine learning result to the first application.

16. A service invoking device, applied to a first application, comprising: a second sending unit, configured to send configuration information to a RIC platform, wherein the RIC platform supports multiple services, the multiple services are used by the first application to perform a machine learning workflow, and the configuration information is used to create tasks related to the machine learning workflow on the multiple services; The second receiving unit is used to receive the machine learning results sent by the RIC platform, where the machine learning results are obtained by the RIC platform calling the multiple services to perform the task.

17. A processing device comprising: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 14.

18. A computer-readable storage medium for storing a computer program, wherein the computer program causes a computer to execute the method according to any one of claims 1 to 14.

19. A computer program product comprising computer program instructions, the computer program instructions causing a computer to execute the method according to any one of claims 1 to 14.

Images