Operator online execution method, device and system and readable storage medium

Abstract

The invention provides an operator online execution method, device and system and a readable storage medium, and the method comprises the steps: executing a process pipeline of a target object, so as to sequentially load model objects corresponding to operator nodes in the process pipeline; executing a hook function of the process pipeline to load an online service object of the target object; the online service object of the target object comprises an operator code of the operator node and an online service framework; and in response to an execution request for a target operator node, executing the operator code of the target operator node and the online service framework according to a model object corresponding to the target operator node to obtain an execution result of the target operator node, the target operator node being any one or more operator nodes in a process pipeline.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and more specifically, to an operator online execution method, an operator online execution device, a system including at least one computing device and at least one storage device, and a computer-readable storage medium. Background Art

[0002] With the rapid development and widespread application of technologies such as artificial intelligence and big data, the application scenarios of artificial intelligence algorithms are becoming more and more extensive.

[0003] In the existing process of executing operators through wrappers, the process pipeline is usually constructed; a data set is constructed; input data of the operator nodes are read; and the execution process pipeline is executed to sequentially execute the operator nodes in the process pipeline according to the input data.

[0004] However, this operator execution mode can only cover offline batch processing and is suitable for computationally intensive operator nodes, that is, it usually takes a long time to execute, and most of the time is spent on data processing and calculation, while the time for building the operating environment and loading data only accounts for a small part. Summary of the invention

[0005] One objective of the present disclosure is to provide a new technical solution for online execution of operators.

[0006] According to a first aspect of the present disclosure, there is provided an operator online execution method, comprising:

[0007] Execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline;

[0008] Executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework;

[0009] In response to an execution request for a target operator node, the operator code of the target operator node and the online service framework are executed according to the model object corresponding to the target operator node to obtain an execution result of the target operator node, wherein the target operator node is any one or more operator nodes in the process pipeline.

[0010] Optionally, executing the hook function of the process pipeline to load the online service object of the target object includes:

[0011] Obtaining the operator code of the operator node in the process pipeline and the online service framework;

[0012] Adding the operator code of the operator node to the online service framework to obtain the online service code of the target object;

[0013] The online service code is loaded and registered.

[0014] Optionally, executing the process pipeline to sequentially load model objects corresponding to operator nodes in the process pipeline includes:

[0015] Acquire a first data storage address of a model parameter of the operator node, a first data format that conforms to the target object, and a second data format that conforms to the operator node;

[0016] Reading first model parameters in the first data format from the first data storage address, and converting the first model parameters into second model parameters in the second data format;

[0017] The model objects corresponding to the operator nodes in the process pipeline are loaded in sequence according to the second model parameters.

[0018] Optionally, in response to the execution request for the target operator node, executing the operator code of the target operator node and the online service framework according to the model object corresponding to the target operator node to obtain the execution result of the target operator node includes:

[0019] In response to an execution request for a target operator node, obtaining a second data storage address of input data of the target operator node, a third data format that conforms to the target object, and a fourth data format that conforms to the operator node;

[0020] According to the third data format and the fourth data format, a first data format converter is constructed; wherein the first data format converter is used to read the data in the third data format and convert the read data in the third data format into data in the fourth data format;

[0021] Reading the first data in the third data format from the second data storage address through the first data format converter, and converting the first data into second data in the fourth data format;

[0022] According to the second data and the model object corresponding to the target operator node, the operator code of the target operator node and the online service framework are executed to obtain the execution result of the target operator node.

[0023] Optionally, the first data format converter is further used to convert the data in the fourth data format into data in the third data format, and write the converted data in the third data format into a corresponding storage medium;

[0024] The method further comprises:

[0025] Obtain a third data storage address of the output data of the target operator node;

[0026] The first data format converter converts the third data in the fourth data format output by the target operator node into fourth data in the third data format, and stores the fourth data in the third data storage address.

[0027] Optionally, the execution request is an HTTP request, and the method further includes:

[0028] Obtaining execution parameters in the request body of the HTTP request;

[0029] The execution parameter is parsed to obtain the second data storage address and the third data format.

[0030] Optionally, the execution request is an HTTP request, and the method further includes:

[0031] The execution result is returned to the requester who issued the execution request through the response body of the HTTP request.

[0032] Optionally, the method further includes:

[0033] Obtaining the algorithm code and configuration information of the operator node;

[0034] Get the wrapper's algorithm code;

[0035] The algorithm code of the wrapper, the algorithm code and configuration information of the operator node, and the algorithm code of the hook function are packaged through a packaging script to obtain the target object.

[0036] Optionally, the method further includes:

[0037] In response to a trigger operation on the target object, providing a selection interface, wherein the selection interface includes options for the operator node;

[0038] Acquire the operator node selected through the selection interface as the target operator node;

[0039] According to the operation of executing the target operator node, an execution request for the target operator node is triggered.

[0040] Optionally, the operator node includes any one or more of the following: an operator node for speech classification, an operator node for image classification, an operator node for picture drawing, an operator node for parts search, an operator node for poster generation, and an operator node for production scheduling decisions.

[0041] Optionally, the method further includes:

[0042] The execution result is visually displayed.

[0043] According to a second aspect of the present disclosure, there is provided an operator online execution device, comprising:

[0044] An object loading module, used to execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline;

[0045] A service loading module, used for executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework;

[0046] A response execution module is used to respond to an execution request for a target operator node, execute the operator code of the target operator node and the online service framework according to the model object corresponding to the target operator node, and obtain the execution result of the target operator node, wherein the target operator node is any one or more operator nodes in the process pipeline.

[0047] Optionally, the service loading module includes:

[0048] A code acquisition unit, used to acquire the operator code of the operator node in the process pipeline and the online service framework;

[0049] A code adding unit, used for adding the operator code of the operator node to the online service framework to obtain the online service code of the target object;

[0050] The code registration unit is used to load and register the online service code.

[0051] Optionally, the object loading module includes:

[0052] A first acquisition unit, used to acquire a first data storage address of a model parameter of the operator node, a first data format conforming to the target object, and a second data format conforming to the operator node;

[0053] A first reading and converting unit, configured to read first model parameters in the first data format from the first data storage address, and convert the first model parameters into second model parameters in the second data format;

[0054] An object loading unit is used to sequentially load the model objects corresponding to the operator nodes in the process pipeline according to the second model parameters.

[0055] Optionally, the response execution module includes:

[0056] a second acquisition unit, configured to acquire, in response to an execution request for a target operator node, a second data storage address of input data of the target operator node, a third data format conforming to the target object, and a fourth data format conforming to the operator node;

[0057] A construction unit, configured to construct a first data format converter according to the third data format and the fourth data format; wherein the first data format converter is configured to read data in the third data format and convert the read data in the third data format into data in the fourth data format;

[0058] A second reading and converting unit, configured to read the first data in the third data format from the second data storage address through the first data format converter, and convert the first data into second data in the fourth data format;

[0059] A code execution unit is used to execute the operator code of the target operator node and the online service framework according to the second data and the model object corresponding to the target operator node, so as to obtain the execution result of the target operator node.

[0060] Optionally, the first data format converter is further used to convert the data in the fourth data format into data in the third data format, and write the converted data in the third data format into a corresponding storage medium;

[0061] The device also includes:

[0062] An address acquisition module, used to acquire a third data storage address of the output data of the target operator node;

[0063] A reading conversion module is used to convert the third data in the fourth data format output by the target operator node into fourth data in the third data format through the first data format converter, and store the fourth data in the third data storage address.

[0064] Optionally, the execution request is an HTTP request, and the device further includes:

[0065] A parameter acquisition module, used to obtain execution parameters in the request body of the HTTP request;

[0066] A parameter parsing module is used to parse the execution parameter to obtain the second data storage address and the third data format.

[0067] Optionally, the execution request is an HTTP request, and the device further includes:

[0068] The result returning module is used to return the execution result to the requester who issued the execution request through the response body of the HTTP request.

[0069] Optionally, the device further comprises:

[0070] A configuration acquisition module, used to obtain the algorithm code and configuration information of the operator node;

[0071] Code acquisition module, used to obtain the algorithm code of the wrapper;

[0072] The packaging processing module is used to package the algorithm code of the wrapper, the algorithm code and configuration information of the operator node, and the algorithm code of the hook function through a packaging script to obtain the target object.

[0073] Optionally, the device further comprises:

[0074] An interface providing module, configured to provide a selection interface in response to a trigger operation on the target object, wherein the selection interface includes options for the operator node;

[0075] A node acquisition module, used to acquire the operator node selected through the selection interface as the target operator node;

[0076] The request triggering module is used to trigger an execution request for the target operator node according to the operation of executing the target operator node.

[0077] Optionally, the operator node includes any one or more of the following: an operator node for speech classification, an operator node for image classification, an operator node for picture drawing, an operator node for parts search, an operator node for poster generation, and an operator node for production scheduling decisions.

[0078] Optionally, the device further comprises:

[0079] The visualization module is used to visualize the execution results.

[0080] According to a third aspect of the present disclosure, a system comprising at least one computing device and at least one storage device is provided, wherein the at least one storage device is used to store instructions, and the instructions are used to control the at least one computing device to execute the method according to the first aspect of the present disclosure; or, the system implements the device according to the second aspect of the present disclosure through the computing device and the storage device.

[0081] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in the first aspect of the present disclosure is implemented.

[0082] Through the embodiments of the present disclosure, the model objects corresponding to the operator nodes in the process pipeline are loaded in sequence in advance, and the online service objects of the target objects are loaded to build an online operating environment for the operator nodes in the process pipeline; when an execution request for the target operator node in the process pipeline is received, the operator code of the target operator node can be executed through the online service framework, without the need to build the operating environment again, and without the need to load the operator code of the target operator node again, that is, the online execution of the target operator node can be achieved, thus reducing the execution time of the target operator node and accelerating the response speed of the execution request for the target operator node.

[0083] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0085] Figure 1 is a block diagram of an example of a hardware configuration of an electronic device that can be used to implement an embodiment of the present disclosure;

[0086] Figure 2 is a flowchart of an operator online execution method provided according to an embodiment of the present disclosure;

[0087] Figure 3 is a flowchart of an example of an operator online execution method provided according to an embodiment of the present disclosure;

[0088] Figure 4 is a block diagram of an operator online execution device provided according to an embodiment of the present disclosure;

[0089] Figure 5 It is a block diagram of a system provided according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0090] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of the parts and steps, the numerical expressions and the numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present disclosure and its application or use. The techniques, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but in appropriate cases, the techniques, methods and equipment should be considered as part of the specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so that once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0091] Hereinafter, various embodiments and examples according to embodiments of the present disclosure are described with reference to the drawings.

[0092] <Hardware Configuration>

[0093] Figure 1 1 is a block diagram showing a hardware configuration of an electronic device 1000 that can implement an embodiment of the present disclosure.

[0094] The electronic device 1000 may be a portable computer, a desktop computer, a mobile phone, a tablet computer, etc. Figure 1 As shown, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, and the like. Among them, the processor 1100 may be a central processing unit CPU, a microprocessor MCU, and the like. The memory 1200 includes, for example, a ROM (read-only memory), a RAM (random access memory), a non-volatile memory such as a hard disk, and the like. The interface device 1300 includes, for example, a USB interface, a headphone interface, and the like. The communication device 1400 is, for example, capable of wired or wireless communication, and may specifically include Wifi communication, Bluetooth communication, 2G / 3G / 4G / 5G communication, and the like. The display device 1500 is, for example, a liquid crystal display screen, a touch display screen, and the like. The input device 1600 may include, for example, a touch screen, a keyboard, a somatosensory input, and the like. The user may input / output voice information through the speaker 1700 and the microphone 1800.

[0095] Figure 1The electronic device shown is merely illustrative and does not in any way imply any limitation on the present disclosure, its application or use. In the embodiments of the present disclosure, the memory 1200 of the electronic device 1000 is used to store instructions, and the instructions are used to control the processor 1100 to operate to execute any one of the operator online execution methods provided in the embodiments of the present disclosure. It should be understood by those skilled in the art that although Figure 1 In the electronic device 1000, multiple devices are shown, but the present disclosure may only involve some of the devices, for example, the electronic device 1000 only involves the processor 1100 and the storage device 1200. A technician can design instructions according to the scheme disclosed in the present disclosure. How instructions control the processor to operate is well known in the art, so it will not be described in detail here.

[0096] <Application scenarios>

[0097] The operator online execution method of this embodiment can be applied to various application scenarios that require operator online execution, such as speech classification, image classification, picture drawing, parts search, poster generation, and production scheduling decision-making.

[0098] First, an online operating environment of the operator nodes in the process pipeline of the target object may be constructed. Specifically, the electronic device may pre-execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline; then execute the hook function of the process pipeline to load the online service object of the target object, and construct the online operating environment of each operator node.

[0099] On this basis, when an HTTP request to execute an operator is received, any operator node in the process pipeline can be executed individually or collectively and respond within seconds.

[0100] Secondly, the electronic device may provide an option selection interface including each operator node in the process pipeline. The user may select the operator node corresponding to the selected option as the target operator node by clicking the option corresponding to the operator node to be executed in the current scenario in the selection interface, and after the selection is completed, perform a click operation on the execution control in the selection interface to trigger an execution request for the target operator.

[0101] Finally, the electronic device responds to the execution request for the target operator, executes the operator code and online service framework of the target operator node according to the model object corresponding to the target operator node, obtains the execution result of the target operator node, stores the execution result, returns the execution result to the requester who sent the execution request, and visualizes the execution result.

[0102] Through the embodiments of the present disclosure, an online operating environment for operator nodes in a process pipeline is pre-built; when an execution request for a target operator node in the process pipeline is received, the operator code of the target operator node can be executed through the online service framework without having to build the operating environment again or load the operator code of the target operator node again, that is, the online execution of the target operator node can be achieved, thereby reducing the execution time of the target operator node and accelerating the response speed of the execution request for the target operator node.

[0103] <Method Example>

[0104] In this embodiment, an operator online execution method is provided. The operator online execution method may be implemented by an electronic device. The electronic device may be Figure 1 An electronic device 1000 is shown.

[0105] according to Figure 2 As shown, the operator online execution method of this embodiment may include the following steps S2100 to S2300:

[0106] Step S2100, executing the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline.

[0107] An operator node is an algorithm code for implementing a corresponding function, specifically an algorithm code for a calculation process that uses input data to calculate and obtain a calculation result.

[0108] The process pipeline is the orchestration of the execution process of operator nodes. In an algorithm, it can be divided into different modules, namely operator nodes, according to specific functions. Then, some or all of the operator nodes are selected according to the needs and combined in a certain order to form an artificial intelligence algorithm that can complete specific functions. A group of operator nodes with a certain sequence can form an execution process pipeline.

[0109] Step S2200, executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework.

[0110] In this embodiment, the hook function is a code that can increase the algorithm execution function at different stages of the algorithm execution.

[0111] The hook function of the online service object used to load the target object can be executed after the process pipeline is executed, that is, the model objects corresponding to the operator nodes in the process pipeline are all loaded, or it can be executed before the process pipeline is executed, or it can be executed during the execution of the process pipeline, that is, when the model object corresponding to any operator node in the process pipeline is loaded. There is no limitation here.

[0112] In one example, a hook function is added at the end of the process pipeline, that is, the hook function is executed when the model objects corresponding to the operator nodes in the process pipeline are all loaded, so that the completion capability of the process pipeline can be enhanced. In one embodiment of the present disclosure, the hook function of the process pipeline is executed to load the online service object of the target object, including: obtaining the operator code and online service framework of the operator node in the process pipeline; adding the operator code of the operator node to the online service framework to obtain the target object online service code; loading and registering the online service code.

[0113] In one embodiment of the present disclosure, codes for multiple online service frameworks may be provided in advance, for example, including but not limited to HTTP frameworks such as FLASK and DJANGO.

[0114] Furthermore, the type of the online service framework may be determined by parsing the execution parameters of the wrapper. When executing the hook function, the corresponding online service framework may be obtained according to the type of the online service framework to load the online service object of the target object.

[0115] In this embodiment, the online service code is loaded and registered in the wrapper, and the wrapper can identify the target object online service code and then execute the target object online service code.

[0116] This embodiment completes the registration of the online service framework by executing the hook function of the online service framework, so that the online operator node can provide services to the outside in the form of an API interface.

[0117] In an example, the FLASK framework can be developed through the jiinjia2 template, the identifier of the operator node can be parsed, and the operator code of the corresponding operator node can be added to the FLASK framework according to the identifier of the operator node. In addition, the code for building a data format converter, the code for loading data, and the code for storing result data can be supplemented in the FLASK framework to obtain the target object online service code, and then the online service code can be loaded and registered to realize the registration and use of the online service code.

[0118] Step S2300 , in response to an execution request for a target operator node, execute the operator code and online service framework of the target operator node according to the model object corresponding to the target operator node, and obtain the execution result of the target operator node.

[0119] The target operator node is any one or more operator nodes in the process pipeline. The target operator node can be any one or more operator nodes selected by the user from the operator nodes in the process pipeline according to the application scenario or specific requirements.

[0120] Through the embodiments of the present disclosure, the model objects corresponding to the operator nodes in the process pipeline are loaded in sequence in advance, and the online service objects of the target objects are loaded to build an online operating environment for the operator nodes in the process pipeline; when an execution request for the target operator node in the process pipeline is received, the operator code of the target operator node can be executed through the online service framework, without the need to build the operating environment again, and without the need to load the operator code of the target operator node again, that is, the online execution of the target operator node can be achieved, thus reducing the execution time of the target operator node and accelerating the response speed of the execution request for the target operator node.

[0121] In one embodiment of the present disclosure, the operator nodes include any one or more of the following: an operator node for speech classification, an operator node for image classification, an operator node for picture drawing, an operator node for parts search, an operator node for poster generation, and an operator node for production scheduling decisions.

[0122] In one embodiment of the present disclosure, before executing step S2100, the method may further include: obtaining the algorithm code and configuration information of the operator node; obtaining the algorithm code of the wrapper; and packaging the algorithm code of the wrapper, the algorithm code and configuration information of the operator node, and the algorithm code of the hook function through a packaging script to obtain a target object.

[0123] In this embodiment, the target object may be pre-built and deployed online for subsequent execution.

[0124] Through this embodiment, efficient version delivery of the target object can be achieved, and the advantages and conveniences of cutting-edge technologies such as cloud native and microservices can be enjoyed.

[0125] In one embodiment of the present disclosure, before executing step S2100, the method may further include: constructing a process pipeline of the target object.

[0126] Specifically, constructing a process pipeline of a target object may include: sequentially constructing operator nodes in the process pipeline.

[0127] In the execution process pipeline, each operator node is combined in a certain order to complete a specific function. Therefore, each operator node in the execution process pipeline can be constructed in sequence according to the combination order of the operator nodes.

[0128] When the operator nodes in the process pipeline are constructed, the process pipeline can be obtained, and the operator nodes in the execution process pipeline have a corresponding order.

[0129] Furthermore, before constructing any operator node, determine whether the number of operator parameters of the operator node is variable; if the number of operator parameters is variable, construct an operator node with a variable number of operator parameters when constructing the operator node; if the number of operator parameters is immutable, construct an operator node with a fixed number of operator parameters when constructing the operator node.

[0130] In this embodiment, whether the number of operator parameters of the operator node is variable may be determined according to the configuration information of the operator node.

[0131] For an operator node with a variable number of operator parameters, the number of operator parameters used during execution may be defined in the execution parameters, so that the operator node is constructed according to the number defined in the execution parameters.

[0132] In one embodiment of the present disclosure, when the process pipeline is constructed, the method may further include: registering the process pipeline in a wrapper.

[0133] In this way, the wrapper can recognize the process pipeline to execute the process pipeline through the wrapper.

[0134] In one embodiment of the present disclosure, the data format that conforms to the target object may be different from the data format that conforms to the operator node. Therefore, in order to enable the operator node to recognize and process the model parameters stored in the first data storage address, the model parameters stored in the first data storage address need to be format converted.

[0135] Specifically, executing the process pipeline to sequentially load the model objects corresponding to the operator nodes in the process pipeline may include steps S2110 to S2130 as shown below:

[0136] Step S2110 , obtaining a first data storage address of a model parameter of an operator node, a first data format that conforms to a target object, and a second data format that conforms to an operator node.

[0137] In one embodiment of the present disclosure, the method may further include:

[0138] Obtain execution parameters of the wrapper; parse the execution parameters to obtain a first data storage address and a first data format.

[0139] The execution parameters in this embodiment may be used to execute the target object and are set by an algorithm engineer according to actual needs when configuring the target object.

[0140] Furthermore, an execution page may be provided in advance, the execution page including at least one packaged object. When any object is selected, a configuration window for the execution parameters of the wrapper of the object may be provided, so that the algorithm engineer can configure the execution parameters of the wrapper of the selected target object through the configuration window.

[0141] Step S2120 , reading first model parameters in a first data format from a first data storage address, and converting the first model parameters into second model parameters in a second data format.

[0142] In one embodiment of the present disclosure, a second data format converter may be constructed based on the first data format and the second data format; wherein the second data format converter is used to read the model parameters of the first data format and convert the read model parameters of the first data format into the model parameters of the second data format; and then the first model parameters of the first data format are read from the first data storage address by the second data format converter, and the first data is converted into the second model parameters of the second data format.

[0143] In this embodiment, the second data formats of the model parameters that can be identified may be different for different operator nodes. Moreover, for an operator node, the second data formats of the model parameters that can be identified may be one or more, so at least one second data format converter may be constructed for the second data formats that conform to all operator nodes and at least one first data format that conforms to the wrapper.

[0144] Specifically, when the number of second data formats that conform to all operator nodes is n, and the number of first data formats that conform to wrappers is m, the number of constructed second data format converters can be m*n. For example, the second data formats that conform to all operator nodes include the first format, the second format, and the third format, and the first data formats that conform to wrappers include the fourth format and the fifth format, then 6 second data format converters can be obtained, including: a second data format converter corresponding to the first format and the fourth format, a second data format converter corresponding to the first format and the fifth format, a second data format converter corresponding to the second format and the fourth format, a second data format converter corresponding to the second format and the fifth format, a second data format converter corresponding to the third format and the fourth format, and a second data format converter corresponding to the third format and the fifth format.

[0145] The second data format converter corresponding to the first format and the fourth format can be used to convert data in the first format and the fourth format; the second data format converter corresponding to the first format and the fifth format can be used to convert data in the first format and the fifth format; the second data format converter corresponding to the second format and the fourth format can be used to convert data in the second format and the fourth format; the second data format converter corresponding to the second format and the fifth format can be used to convert data in the second format and the fifth format; the second data format converter corresponding to the third format and the fourth format can be used to convert data in the third format and the fourth format; the second data format converter corresponding to the third format and the fifth format can be used to convert data in the third format and the fifth format.

[0146] In this way, by constructing a second data format converter, reading data through the second data format converter, and performing format conversion on the read data, there is no need to modify the operator code of the operator node when the model parameters matched by the operator node change, thereby improving the adaptability of the operator code of the operator node and improving the computational efficiency of the operator code of the operator node.

[0147] In the case of obtaining the second data format converter, the method may further include: registering the second data format converter in the wrapper.

[0148] In this way, the wrapper can identify the second data format converter, so that the wrapper reads the first model parameters of the first format stored in the first data storage address through the second data format converter, converts the read first model parameters of the first data format into second model parameters of the second data format, and provides the second model parameters to the corresponding operator node.

[0149] Step S2130: load the model objects corresponding to the operator nodes in the process pipeline in sequence according to the second model parameters.

[0150] In one embodiment of the present disclosure, the model objects corresponding to the operator nodes in the process pipeline may be loaded in sequence according to the order of the operator nodes in the process pipeline.

[0151] Specifically, the operator nodes in the process pipeline can be traversed according to the order of the operator nodes in the process pipeline, and the model object corresponding to the currently traversed operator node can be loaded according to the corresponding second model parameter. After the model object corresponding to the currently traversed operator node is loaded, it is determined whether the currently traversed operator node is the last operator node in the process pipeline. If so, the process ends; if not, the traversal continues.

[0152] Through this embodiment, when the data format that conforms to the target object is different from the data format that conforms to the operator node, the model parameters stored at the first data storage address are format converted so that the operator node can recognize and process the model parameters stored at the first data storage address, thereby realizing the diversity of storage methods and storage formats of model objects.

[0153] In one embodiment of the present disclosure, the node identifier of the target operator node may be included in the execution request, and the operator code and online service framework of the operator node corresponding to the node identifier may be executed to obtain the execution result. The node identifier may be used to uniquely identify the corresponding operator node.

[0154] In one embodiment of the present disclosure, the method may further include:

[0155] In response to a trigger operation on a target object, a selection interface is provided, the selection interface including options for operator nodes; the operator node selected through the selection interface is obtained as a target operator node; and according to an operation of executing the target operator node, an execution request for the target operator node is triggered.

[0156] In one example, an execution control may be provided in the selection interface. When the user selects the target operator node, the operation performed on the execution control is the operation of executing the target operator node, triggering an execution request for the target operator node.

[0157] In one embodiment of the present disclosure, the method further includes: visually displaying the execution result.

[0158] In one embodiment of the present disclosure, the data format that conforms to the target object may be different from the data format that conforms to the operator node. Therefore, in order to enable the operator node to recognize and process the data stored in the second data storage address, the data stored in the second data storage address needs to be format converted.

[0159] Specifically, in response to an execution request for a target operator node, the operator code and online service framework of the target operator node are executed according to the model object corresponding to the target operator node to obtain the execution result of the target operator node, including steps S2310 to S2340 as shown below:

[0160] Step S2310 , in response to an execution request for a target operator node, obtaining a second data storage address of input data of the target operator node, a third data format conforming to the target object, and a fourth data format conforming to the operator node.

[0161] In one embodiment of the present disclosure, the execution request is an HTTP request, and the method may further include: obtaining execution parameters in the request body of the HTTP request; parsing the execution parameters to obtain the second data storage address and the third data format.

[0162] Step S2320: construct a first data format converter according to the third data format and the fourth data format; wherein the first data format converter is used to read data in the third data format and convert the read data in the third data format into data in the fourth data format.

[0163] In this embodiment, the fourth data format of the data that can be recognized may be different for different operator nodes. Moreover, for an operator node, the fourth data format of the data that can be recognized may be one or more, so at least one first data format converter may be constructed for the fourth data format that conforms to all operator nodes and at least one third data format that conforms to the target object.

[0164] Specifically, when the number of fourth data formats that conform to all operator nodes is n, and the number of third data formats that conform to the target object is m, the number of first data format converters constructed can be m*n. For example, the fourth data formats that conform to all operator nodes include the first format, the second format, and the third format, and the third data formats that conform to the target object include the fourth format and the fifth format, then 6 first data format converters can be obtained, including: a first data format converter corresponding to the first format and the fourth format, a first data format converter corresponding to the first format and the fifth format, a first data format converter corresponding to the second format and the fourth format, a first data format converter corresponding to the second format and the fifth format, a first data format converter corresponding to the third format and the fourth format, and a first data format converter corresponding to the third format and the fifth format.

[0165] The first data format converter corresponding to the first format and the fourth format can be used to convert data in the first format and the fourth format; the first data format converter corresponding to the first format and the fifth format can be used to convert data in the first format and the fifth format; the first data format converter corresponding to the second format and the fourth format can be used to convert data in the second format and the fourth format; the first data format converter corresponding to the second format and the fifth format can be used to convert data in the second format and the fifth format; the first data format converter corresponding to the third format and the fourth format can be used to convert data in the third format and the fourth format; the first data format converter corresponding to the third format and the fifth format can be used to convert data in the third format and the fifth format.

[0166] In this way, by constructing a first data format converter, then reading data through the first data format converter and performing format conversion on the read data, there is no need to modify the operator code of the operator node when the data source matched by the operator node changes, thereby improving the adaptability of the operator code of the operator node and improving the computational efficiency of the operator code of the operator node.

[0167] In the case of obtaining the first data format converter, the method may further include: registering the first data format converter in the wrapper.

[0168] In this way, the wrapper can identify the first data format converter, so that the wrapper reads the first data in the first format stored in the first data storage address through the first data format converter, converts the read first data in the first data format into second data in the fourth data format, and provides the second data to the corresponding operator node.

[0169] Step S2330: read first data in a third data format from a second data storage address through a first data format converter, and convert the first data into second data in a fourth data format.

[0170] In this embodiment, the wrapper may be a first data format converter that converts first data in a third data format stored in a second data storage address into second data in a fourth data format that can be recognized by an operator node, and provides the second data to a corresponding operator node.

[0171] Step S2340: Execute the operator code and online service framework of the target operator node according to the second data and the model object corresponding to the target operator node to obtain the execution result of the target operator node.

[0172] In one embodiment of the present disclosure, the target operator nodes in the process pipeline may be executed in sequence according to the order of the target operator nodes in the process pipeline.

[0173] Through this embodiment, when the data format that conforms to the target object is different from the data format that conforms to the operator node, the data stored at the second data storage address is format converted so that the operator node can identify and process the data stored at the second data storage address, and the diversity of the storage method and storage format of the operator data can be achieved. In addition, the online execution of the target operator node can be achieved without re-building the operating environment and re-loading the operator code of the target operator node, so that the execution time of the target operator node can be reduced and the response speed of the execution request for the target operator node can be accelerated.

[0174] In one embodiment of the present disclosure, the first data format converter is further used to convert data in the fourth data format into data in the third data format, and write the converted data in the third data format into a corresponding storage medium.

[0175] On this basis, the method may also include: obtaining a third data storage address of the output data of the target operator node; converting the third data in the fourth data format output by the target operator node into fourth data in the third data format through a first data format converter, and storing the fourth data to the third data storage address.

[0176] Through the first data format converter of this embodiment, the data output after the operator node is executed can be stored in the third data storage address.

[0177] In one embodiment of the present disclosure, the execution request is an HTTP request, and the method may further include: returning the execution result to the requester who issued the execution request through the response body of the HTTP request.

[0178] In this embodiment, when an HTTP request arrives, the request may be responded to through the online service framework. Specifically, the online service framework may parse the execution parameters in the request body of the HTTP request, obtain the second data address and the third data format of the target object, and construct a first data format converter according to the third data format and the fourth data format that conforms to the operator node; the first data in the third data format read from the second data storage address is converted into the memory object required by the operator - the second data in the fourth data format through the first data format converter, and handed over to the target operator node for calculation to obtain the corresponding execution result. The online service framework also stores the execution result of the target operator node through the corresponding first data format converter, and returns the execution result to the requester of the HTTP request through the response body of the HTTP request.

[0179] Figure 3 The flowchart is an example of an operator online execution method according to an embodiment of the present disclosure.

[0180] according to Figure 3 As shown, the operator online execution method may include steps S3001 to S3017 as shown below:

[0181] Step S3001, constructing a process pipeline of the target object.

[0182] Step S3002, obtaining execution parameters of the wrapper.

[0183] Step S3003, parsing the execution parameter to obtain a first data storage address of a model parameter of an operator node, a first data format that conforms to a target object, and a second data format that conforms to an operator node.

[0184] Step S3004: construct a second data format converter according to the first data format and the second data format.

[0185] Step S3005: read the first model parameters in the first data format from the first data storage address through the second data format converter, and convert the first data into the second model parameters in the second data format.

[0186] Step S3006, traverse the operator nodes in the execution process pipeline.

[0187] Step S3007, loading the model objects corresponding to the currently traversed operator nodes in sequence according to the second model parameters.

[0188] Step S3008, determine whether the currently traversed operator node is the last operator node in the process pipeline, if not, continue to execute step S3006; if yes, execute step S3009.

[0189] Step S3009, executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework.

[0190] Step S3010: In response to an execution request for a target operator node, obtain execution parameters in a request body of the HTTP request.

[0191] Step S3011, parsing the execution parameters, obtaining the second data storage address of the input data of the target operator node, the third data storage address of the output data of the target operator node, and the third data format that conforms to the target object.

[0192] Step S3012, obtaining a fourth data format that conforms to the operator node.

[0193] Step S3013: construct a first data format converter according to the third data format and the fourth data format.

[0194] Step S3014: read first data in the third data format from the second data storage address through the first data format converter, and convert the first data into second data in the fourth data format.

[0195] Step S3015: Execute the operator code and online service framework of the target operator node according to the second data and the model object corresponding to the target operator node to obtain the execution result of the target operator node.

[0196] Step S3016: convert the third data in the fourth data format output by the target operator node into fourth data in the third data format through the first data format converter, and store the fourth data in the third data storage address.

[0197] Step S3017: Return the execution result to the requester who issued the execution request through the response body of the HTTP request.

[0198] <Device Example>

[0199] In this embodiment, an operator online execution device 4000 is provided, such as Figure 4 As shown, it includes an object loading module 4100, a service loading module 4200, and a response execution module 4300.

[0200] The object loading module 4100 is used to execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline.

[0201] The service loading module 4200 is used to execute the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework.

[0202] The response execution module 4300 is used to respond to an execution request for a target operator node, execute the operator code of the target operator node and the online service framework according to the model object corresponding to the target operator node, and obtain the execution result of the target operator node, wherein the target operator node is any one or more operator nodes in the process pipeline.

[0203] In one embodiment of the present disclosure, the service loading module 4200 includes a code acquisition unit, a code adding unit, and a code registration unit.

[0204] A code acquisition unit is used to acquire the operator code of the operator node in the process pipeline and the online service framework.

[0205] A code adding unit is used to add the operator code of the operator node to the online service framework to obtain the online service code of the target object.

[0206] The code registration unit is used to load and register the online service code.

[0207] In one embodiment of the present disclosure, the object loading module 4100 includes a first acquisition unit, a first reading conversion unit, and an object loading unit.

[0208] The first acquisition unit is used to acquire a first data storage address of a model parameter of the operator node, a first data format that conforms to the target object, and a second data format that conforms to the operator node.

[0209] The first reading and converting unit is configured to read the first model parameters in the first data format from the first data storage address, and convert the first model parameters into the second model parameters in the second data format.

[0210] An object loading unit is used to sequentially load the model objects corresponding to the operator nodes in the process pipeline according to the second model parameters.

[0211] In one embodiment of the present disclosure, the response execution module 4300 includes a second acquisition unit, a construction unit, a second read conversion unit, and a code execution unit.

[0212] The second acquisition unit is used to acquire, in response to an execution request for a target operator node, a second data storage address of input data of the target operator node, a third data format that conforms to the target object, and a fourth data format that conforms to the operator node.

[0213] A construction unit is used to construct a first data format converter according to the third data format and the fourth data format; wherein the first data format converter is used to read data in the third data format and convert the read data in the third data format into data in the fourth data format.

[0214] The second reading and converting unit is configured to read the first data in the third data format from the second data storage address through the first data format converter, and convert the first data into second data in the fourth data format.

[0215] A code execution unit is used to execute the operator code of the target operator node and the online service framework according to the second data and the model object corresponding to the target operator node, so as to obtain the execution result of the target operator node.

[0216] In one embodiment of the present disclosure, the first data format converter is further used to convert the data in the fourth data format into data in the third data format, and write the converted data in the third data format into a corresponding storage medium;

[0217] The device 4000 further includes an address acquisition module and a read conversion module.

[0218] An address acquisition module is used to acquire a third data storage address of the output data of the target operator node.

[0219] A reading conversion module is used to convert the third data in the fourth data format output by the target operator node into fourth data in the third data format through the first data format converter, and store the fourth data in the third data storage address.

[0220] In one embodiment of the present disclosure, the execution request is an HTTP request, and the device 4000 further includes a parameter acquisition module and a parameter parsing module.

[0221] The parameter acquisition module is used to obtain the execution parameters in the request body of the HTTP request.

[0222] A parameter parsing module is used to parse the execution parameter to obtain the second data storage address and the third data format.

[0223] In one embodiment of the present disclosure, the execution request is an HTTP request, and the device 4000 further includes a result returning module, which is used to return the execution result to the requester who issued the execution request through the response body of the HTTP request.

[0224] In one embodiment of the present disclosure, the device 4000 further includes a configuration acquisition module, a code acquisition module, and a packaging processing module.

[0225] The configuration acquisition module is used to obtain the algorithm code and configuration information of the operator node.

[0226] The code acquisition module is used to obtain the algorithm code of the wrapper.

[0227] The packaging processing module is used to package the algorithm code of the wrapper, the algorithm code and configuration information of the operator node, and the algorithm code of the hook function through a packaging script to obtain the target object.

[0228] In one embodiment of the present disclosure, the device 4000 further includes an interface providing module, a node acquiring module, and a request triggering module.

[0229] The interface providing module is used to provide a selection interface in response to a trigger operation on the target object, wherein the selection interface includes options of the operator node.

[0230] The node acquisition module is used to acquire the operator node selected through the selection interface as the target operator node.

[0231] The request triggering module is used to trigger an execution request for the target operator node according to the operation of executing the target operator node.

[0232] In one embodiment of the present disclosure, the operator node includes any one or more of the following: an operator node for speech classification, an operator node for image classification, an operator node for picture drawing, an operator node for parts search, an operator node for poster generation, and an operator node for production scheduling decisions.

[0233] In one embodiment of the present disclosure, the apparatus 4000 further includes a visualization display module for visually displaying the execution result.

[0234] Those skilled in the art should understand that the operator online execution device 4000 can be implemented in various ways. For example, the operator online execution device 4000 can be implemented by configuring the processor with instructions. For example, the instructions can be stored in a ROM, and when the device is started, the instructions are read from the ROM into a programmable device to implement the operator online execution device 4000. For example, the operator online execution device 4000 can be solidified into a dedicated device (such as an ASIC). The operator online execution device 4000 can be divided into independent units, or they can be combined together for implementation. The operator online execution device 4000 can be implemented by one of the various implementations described above, or can be implemented by a combination of two or more of the various implementations described above.

[0235] In this embodiment, the operator online execution device 4000 can have multiple implementation forms. For example, the operator online execution device 4000 can be a functional module running in any software product or application that provides operator online execution services, or it can be a peripheral embedded component, plug-in, patch, etc. of these software products or applications, or it can also be these software products or applications themselves.

[0236] <System Embodiment>

[0237] In this embodiment, if Figure 5 As shown, a system 5000 including at least one computing device 5100 and at least one storage device 5200 is also provided.

[0238] On the one hand, the at least one storage device 5200 is used to store executable instructions; the instructions are used to control at least one computing device 5100 to execute the operator online execution method according to any embodiment of the present disclosure.

[0239] On the other hand, the system 5000 implements the operator online execution device 4000 according to any embodiment of the present disclosure through the computing device 5100 and the storage device 5200.

[0240] In this embodiment, the system 5000 can be a mobile phone, a tablet computer, a PDA, a desktop computer, a laptop computer, a workstation, a game console, etc., or it can be a distributed system composed of multiple devices.

[0241] <Computer Readable Storage Medium>

[0242] In this embodiment, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the operator online execution method of any embodiment of the present disclosure is implemented.

[0243] The present disclosure may be a device, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0244] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0245] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0246] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0247] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0248] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0249] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0250] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of an instruction, and the module, a program segment or a part of an instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0251] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the marketplace, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the present disclosure is defined by the appended claims.

Claims

1. An operator online execution method, characterized in that: include: Execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline; Executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework; In response to an execution request for a target operator node, the operator code of the target operator node and the online service framework are executed according to the model object corresponding to the target operator node to obtain an execution result of the target operator node, wherein the target operator node is any one or more operator nodes in the process pipeline.

2. The method according to claim 1, characterized in that The executing the hook function of the process pipeline to load the online service object of the target object includes: Obtaining the operator code of the operator node in the process pipeline and the online service framework; Adding the operator code of the operator node to the online service framework to obtain the online service code of the target object; The online service code is loaded and registered.

3. The method according to claim 1, characterized in that The executing the process pipeline to sequentially load the model objects corresponding to the operator nodes in the process pipeline includes: Acquire a first data storage address of a model parameter of the operator node, a first data format that conforms to the target object, and a second data format that conforms to the operator node; Reading first model parameters in the first data format from the first data storage address, and converting the first model parameters into second model parameters in the second data format; The model objects corresponding to the operator nodes in the process pipeline are loaded in sequence according to the second model parameters.

4. The method according to claim 1, characterized in that: The step of responding to the execution request for the target operator node and executing the operator code of the target operator node and the online service framework according to the model object corresponding to the target operator node to obtain the execution result of the target operator node includes: In response to an execution request for a target operator node, obtaining a second data storage address of input data of the target operator node, a third data format that conforms to the target object, and a fourth data format that conforms to the operator node; According to the third data format and the fourth data format, a first data format converter is constructed; wherein the first data format converter is used to read the data in the third data format and convert the read data in the third data format into data in the fourth data format; Reading the first data in the third data format from the second data storage address through the first data format converter, and converting the first data into second data in the fourth data format; According to the second data and the model object corresponding to the target operator node, the operator code of the target operator node and the online service framework are executed to obtain the execution result of the target operator node.

5. The method according to claim 4, characterized in that The first data format converter is further used to convert the data in the fourth data format into data in the third data format, and write the converted data in the third data format into a corresponding storage medium; The method further comprises: Acquire a third data storage address of output data of the target operator node; The first data format converter converts the third data in the fourth data format output by the target operator node into fourth data in the third data format, and stores the fourth data in the third data storage address.

6. The method according to claim 4, characterized in that The execution request is an HTTP request, and the method further includes: Obtaining execution parameters in the request body of the HTTP request; The execution parameter is parsed to obtain the second data storage address and the third data format.

7. The method according to claim 1, characterized in that The execution request is an HTTP request, and the method further includes: The execution result is returned to the requester who issued the execution request through the response body of the HTTP request.

8. An operator online execution device, characterized in that: include: An object loading module, used to execute the process pipeline of the target object to sequentially load the model objects corresponding to the operator nodes in the process pipeline; A service loading module, used for executing the hook function of the process pipeline to load the online service object of the target object; the online service object of the target object includes the operator code of the operator node and the online service framework; A response execution module is used to respond to an execution request for a target operator node, execute the operator code of the target operator node and the online service framework according to the model object corresponding to the target operator node, and obtain the execution result of the target operator node, wherein the target operator node is any one or more operator nodes in the process pipeline.

9. A system comprising at least one computing device and at least one storage device, wherein: The at least one storage device is used to store instructions, and the instructions are used to control the at least one computing device to execute the method according to any one of claims 1 to 7; or, the system implements the device according to claim 8 through the computing device and the storage device.

10. A computer-readable storage medium, wherein: A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images