CN114443098A

CN114443098A - Model deployment updating processing method and device, equipment, medium and product thereof

Info

Publication number: CN114443098A
Application number: CN202210105937.XA
Authority: CN
Inventors: 林剑周
Original assignee: Guangzhou Huaduo Network Technology Co Ltd
Current assignee: Guangzhou Huaduo Network Technology Co Ltd
Priority date: 2022-01-28
Filing date: 2022-01-28
Publication date: 2022-05-06

Abstract

The application discloses a model deployment updating processing method and a device, equipment, medium and product thereof, wherein the method comprises the following steps: responding to a model updating event acting on the target model online service, and acquiring a new neural network model and corresponding model configuration information thereof; pushing a new model storage instruction to the shared container service, and driving the shared container service to respond to the instruction so as to correspondingly store the new neural network model and the model version number to a target service storage space in a target model pool according to model configuration information; and when the target model online service meets a preset new model deployment strategy, pushing the shared storage path to the target model online service, and driving the target model online service to acquire a new neural network model from the shared container service according to the shared storage path for deployment. According to the method and the device, automatic management of the deployment and the updating of the new neural network model is realized, the deployment time of the new model is saved, the hot updating efficiency of the model is improved, and the data safety of the model is ensured by carrying out isolation management on the data of the model.

Description

Model deployment updating processing method and device, equipment, medium and product thereof

Technical Field

The present application relates to the field of neural network models, and in particular, to a model deployment update processing method, and further, to a corresponding apparatus, device, non-volatile storage medium, and computer program product of the method.

Background

Various types of online services built based on neural network models exist in the existing internet platform, for example, model services corresponding to the neural network models for semantic reasoning generally exist in the online services for providing intelligent customer service for users, model services corresponding to the neural network models for commodity classification generally exist in the online services for providing commodity classification for users, and with the technical development of the neural networks, each platform can update the neural network models for providing the reasoning function in the online services, so that the business processing efficiency of the online services built based on the neural network models is improved, and the use experience of the users is optimized.

However, when a developer in an existing platform deploys a new neural network model to a corresponding model online service, the developer is often required to manually deploy the new neural network model to the corresponding model online service for model update after completing development and training of the new neural network model, and a manual mode is slow in execution efficiency, and model deployment is prone to be unsmooth due to misoperation in the manual deployment process, which affects the stability of the model online service.

Secondly, although some platforms may also have corresponding model automation deployment systems, the operation modes of manual deployment are only written into corresponding execution scripts to implement automation deployment, and the neural network models of a large number of model online services in the platforms cannot be effectively managed.

In view of the problems with the deployment of neural network models in existing online services, the applicant has made a corresponding search for a solution to this problem.

Disclosure of Invention

The application aims to meet the requirements of users and provides a model deployment updating processing method, and further relates to a corresponding device, equipment, a non-volatile storage medium and a computer program product of the method.

In order to realize the purpose of the application, the following technical scheme is adopted:

the model deployment updating processing method adapted to the purpose of the application comprises the following steps:

responding to a model updating event acting on the target model online service, and acquiring a new neural network model and corresponding model configuration information thereof, wherein the new neural network model and the corresponding model configuration information comprise a model pool identifier, a model service identifier and a model version number;

a new model storage instruction is pushed to the shared container service, the shared container service is driven to respond to the instruction so as to correspondingly store the new neural network model and the model version number to a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier according to the model configuration information;

and when the target model online service meets a preset new model deployment strategy, pushing a shared storage path corresponding to the new neural network model to the target model online service, and driving the target model online service to acquire the new neural network model from the shared container service according to the shared storage path for deployment.

In a further embodiment, the step of obtaining the new neural network model and the corresponding model configuration information included in the event in response to the model update event acting on the target model online service includes:

receiving a new neural network model and corresponding model configuration information thereof, which are pushed by a model development terminal through a model uploading interface;

and correspondingly storing the new neural network model and the model configuration information into a model cloud storage space.

In a further embodiment, the step of driving the shared container service to respond to the instruction to store the new neural network model and the model version number in the target model pool corresponding to the model pool identifier and corresponding to the target service storage space corresponding to the service identifier in the target model pool corresponding to the model pool identifier includes the following steps executed by the shared container service:

responding to a new model storage instruction pushed by a model updating service, and acquiring a new neural network model corresponding to the instruction and model configuration information thereof from a model cloud storage space;

obtaining a model pool identifier and a model service identifier contained in the model configuration information, and inquiring a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier;

and acquiring a model version number contained in the model configuration information, and correspondingly storing the new neural network model and the model version number into the target service storage space.

In a further embodiment, when the target model online service meets a preset new model deployment policy, the step of pushing the shared storage path corresponding to the new neural network model to the target model online service includes:

obtaining model updating time preset for the target model online service, and monitoring whether the current time exceeds the model updating time;

and when the current time exceeds the model updating time, pushing a model updating instruction to the target model online service, wherein the model updating instruction comprises the shared storage path.

In a further embodiment, the step of acquiring, by the driving target model online service, a new neural network model from the shared container service according to the shared storage path for deployment includes the following steps performed by the target model online service:

responding to a model updating instruction pushed by a model updating service, and acquiring a shared storage path contained in the instruction;

acquiring the new neural network model from the target service storage space in the target model pool of the shared container service according to the shared storage path;

and after the deployment of the new neural network model is completed, the current old neural network model is off-line, and the new neural network model is on-line to provide a reasoning function for the current model service.

In a further embodiment, in the step of pushing the shared storage path corresponding to the new neural network model to the target model online service, the shared storage path is pushed to the current model update service by a shared container service, and the shared storage path points to the new neural network model in the target service storage space of the target model pool.

A model deployment update processing apparatus proposed adapted to an object of the present application includes:

the update event response module is used for responding to a model update event acting on the target model online service and acquiring a new neural network model and corresponding model configuration information thereof, wherein the new neural network model and the corresponding model configuration information comprise a model pool identifier, a model service identifier and a model version number;

the model sharing storage module is used for pushing a new model storage instruction to the sharing container service, and driving the sharing container service to respond to the instruction so as to correspondingly store the new neural network model and the model version number to a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier according to the model configuration information;

and the new model deployment module is used for pushing a shared storage path corresponding to the new neural network model to the target model online service when the target model online service meets a preset new model deployment strategy, and driving the target model online service to acquire the new neural network model from the shared container service according to the shared storage path for deployment.

In a further embodiment, the update event response module comprises:

the model acquisition submodule is used for receiving a new neural network model and corresponding model configuration information thereof which are pushed by the model development end through the model uploading interface;

and the model cloud storage submodule is used for correspondingly storing the new neural network model and the model configuration information into a model cloud storage space.

In a further embodiment, the model shared storage module comprises:

the storage instruction response submodule is used for responding to a new model storage instruction pushed by the model updating service and acquiring a new neural network model corresponding to the instruction and model configuration information thereof from the model cloud storage space;

a storage space query submodule, configured to obtain a model pool identifier and a model service identifier included in the model configuration information, and query a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier;

and the model storage submodule is used for acquiring a model version number contained in the model configuration information and correspondingly storing the new neural network model and the model version number into the target service storage space.

In a further embodiment, the new model deployment module comprises:

the model updating time monitoring submodule is used for acquiring model updating time preset for the target model online service and monitoring whether the current time exceeds the model updating time;

and the model updating instruction pushing submodule is used for pushing a model updating instruction to the target model online service when the current time is monitored to exceed the model updating time, and the model updating instruction comprises the shared storage path.

In a preferred embodiment, the new model deployment module further comprises:

the update instruction response submodule is used for responding to a model update instruction pushed by the model update service and acquiring a shared storage path contained in the instruction;

a new model obtaining sub-module, configured to obtain the new neural network model from the target service storage space in the target model pool of the shared container service according to the shared storage path;

and the new model online sub-module is used for offline the current and old neural network models after the deployment of the new neural network model is completed, and online the new neural network model to provide a reasoning function for the current model service.

In order to solve the above technical problem, an embodiment of the present invention further provides a computer device, including a memory and a processor, where the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, cause the processor to execute the steps of the model deployment update processing method.

In order to solve the above technical problem, an embodiment of the present invention further provides a storage medium storing computer readable instructions, which when executed by one or more processors, cause the one or more processors to execute the steps of the model deployment update processing method described above.

In order to solve the above technical problem, an embodiment of the present invention further provides a computer program product, which includes a computer program and computer instructions, and when the computer program and the computer instructions are executed by a processor, the processor executes the steps of the model deployment update processing method.

Compared with the prior art, the application has the following advantages:

the model deployment management system comprises a model updating service and a shared container service, the model updating service provides a deployment interface for a development end where a developer is located, so that the developer uploads a newly developed neural network model to the model updating service to perform deployment processing of the new model, manual deployment of the developer is not needed, time of the developer is saved, and model updating efficiency is improved.

Secondly, the shared container service in the application can be a neural network model of each version of model online service in a platform automation management platform, and each version of neural network model of each model online service is managed in a pool-divided grouping mode in the shared container service, so that data isolation of each neural network model is realized, the safety of model data is ensured, and the system stability of the model online service is improved.

In addition, the shared container service of the application manages the neural network models of different versions of the model online service, so that the platform can deploy multiple versions of models in the shared container service to realize the multi-version model operation of the model online service.

Drawings

The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

fig. 1 is a schematic diagram of a typical network deployment architecture related to implementing the technical solution of the present application;

FIG. 2 is a schematic flow chart diagram illustrating an exemplary embodiment of a model deployment update processing method according to the present application;

FIG. 3 is a schematic flow chart illustrating an embodiment of a model cloud storage space storing a new neural network model by a model update service according to the present application;

FIG. 4 is a schematic flowchart illustrating a specific embodiment of the present application for a shared container service to acquire a new neural network model from a model cloud storage space for storage processing;

FIG. 5 is a schematic flow chart illustrating a specific embodiment of notifying a target model online service of new model deployment when the model update service reaches the model update time according to the present application;

FIG. 6 is a schematic flow chart illustrating an embodiment of the present application for obtaining a new neural network model from a shared container service for hot-update when a target model online service is running;

FIG. 7 is a functional block diagram of an exemplary embodiment of a model deployment update processing apparatus of the present application;

fig. 8 is a block diagram of a basic structure of a computer device according to an embodiment of the present application.

Detailed Description

Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application.

As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.

It will be understood by those within the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

As will be appreciated by those skilled in the art, "client," "terminal," and "terminal device" as used herein include both devices that are wireless signal receivers, which are devices having only wireless signal receivers without transmit capability, and devices that are receive and transmit hardware, which have receive and transmit hardware capable of two-way communication over a two-way communication link. Such a device may include: cellular or other communication devices such as personal computers, tablets, etc. having single or multi-line displays or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service), which may combine voice, data processing, facsimile and/or data communication capabilities; a PDA (Personal Digital Assistant), which may include a radio frequency receiver, a pager, internet/intranet access, a web browser, a notepad, a calendar and/or a GPS (Global Positioning System) receiver; a conventional laptop and/or palmtop computer or other device having and/or including a radio frequency receiver. As used herein, a "client," "terminal device" can be portable, transportable, installed in a vehicle (aeronautical, maritime, and/or land-based), or situated and/or configured to operate locally and/or in a distributed fashion at any other location(s) on earth and/or in space. The "client", "terminal Device" used herein may also be a communication terminal, a web terminal, a music/video playing terminal, such as a PDA, an MID (Mobile Internet Device) and/or a Mobile phone with music/video playing function, and may also be a smart tv, a set-top box, and the like.

The hardware referred to by the names "server", "client", "work node", etc. is essentially an electronic device with the performance of a personal computer, and is a hardware device having necessary components disclosed by the von neumann principle such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, an output device, etc., a computer program is stored in the memory, and the central processing unit calls a program stored in an external memory into the internal memory to run, executes instructions in the program, and interacts with the input and output devices, thereby completing a specific function.

It should be noted that the concept of "server" as referred to in this application can be extended to the case of a server cluster. According to the network deployment principle understood by those skilled in the art, the servers should be logically divided, and in physical space, the servers may be independent from each other but can be called through an interface, or may be integrated into one physical computer or a set of computer clusters. Those skilled in the art will appreciate this variation and should not be so limited as to restrict the implementation of the network deployment of the present application.

Referring to fig. 1, the hardware basis required for implementing the related art embodiments of the present application may be deployed according to the architecture shown in the figure. The server 80 is deployed at the cloud end, and serves as an online server, which may be responsible for further connecting to related data servers and other servers providing related support, so as to form a logically associated server cluster to provide services for related terminal devices, such as the smart phone 81 and the personal computer 82 shown in the figure, or a third-party server (not shown). Both the smart phone and the personal computer can access the internet through a known network access mode, and establish a data communication link with the cloud server 80 so as to run a terminal application program related to the service provided by the server.

For the server, the application program is usually constructed as a service process, and a corresponding program interface is opened for remote call of the application program running on various terminal devices.

The application program refers to an application program running on a server or a terminal device, the application program implements the related technical scheme of the application in a programming mode, a program code of the application program can be saved in a nonvolatile storage medium which can be identified by a computer in a form of a computer executable instruction, and is called into a memory by a central processing unit to run, and the related device of the application is constructed by running the application program on the computer.

The person skilled in the art will know this: although the various methods of the present application are described based on the same concept so as to be common to each other, they may be independently performed unless otherwise specified. In the same way, for each embodiment disclosed in the present application, it is proposed based on the same inventive concept, and therefore, concepts of the same expression and concepts of which expressions are different but are appropriately changed only for convenience should be equally understood.

Referring to fig. 2, a model deployment update processing method according to the present application, in an exemplary embodiment, includes the following steps:

step S11, responding to a model update event acting on the target model online service, and acquiring a new neural network model and corresponding model configuration information contained in the event, wherein the model configuration information contains a model pool identifier, a model service identifier and a model version number:

the current model updating service responds to the model updating event acting on the target model online service to obtain a new updated neural network model contained in the model updating event and the model configuration information corresponding to the new neural network model.

The model updating event is generally generated by being triggered by a model development end, the model development end is a development end which establishes a data communication link with a current model updating service, after a development user at the model development end completes a new neural network model aiming at a target model online service and edits a model pool identifier, a model service identifier and a model version number corresponding to the new neural network model, the model development end encapsulates the model pool identifier, the model service identifier and the model version number to generate model configuration information, and then triggers the model updating event, so that the current model updating service responds to the model updating event to acquire the new neural network model and the model configuration information.

The model updating service is a service for providing model updating for model online service, a model development end and a shared container service which are associated with the model updating service, and specifically, the model updating service is responsible for storing a new neural network model newly developed or updated by the model development end into the shared container service and informing the model development end of acquiring a corresponding new neural network model from the shared container service for deployment; the model online service provides corresponding neural network model reasoning service through a neural network model deployed by the model online service; the shared container service is responsible for storing newly updated or developed neural network models of the model online services.

The model configuration information is used to represent the model version of the new neural network model corresponding to the model configuration information and the storage space corresponding to the new neural network model stored in the shared container service, for example, the model version included in the model configuration information is used to represent the model version of the new neural network model, the model pool identifier and the model service identifier included in the model configuration information are used to represent the storage space of the new neural network model in the shared container service, and the specific storage manner refers to the description of the related embodiments in the subsequent steps, which is not repeated in this step.

After the current model updating service acquires the new neural network model and the model configuration information, the neural network model and the model configuration information can be stored in a model cloud space to drive the shared container service to acquire the neural network model and the model configuration information from the model cloud space and store the neural network model and the model configuration information in a corresponding storage space, and the current model updating service is not required to push the new neural network model and the model configuration information to the updating container service, so that the data transmission pressure of the current model updating service is reduced.

Step S12, pushing a new model storage instruction to the shared container service, and driving the shared container service to respond to the instruction to store the new neural network model and the model version number in correspondence to the target service storage space corresponding to the service identifier in the target model pool corresponding to the model pool identifier according to the model configuration information:

and after the current model updating service acquires the new neural network model and the model configuration information, the new model storage instruction is served to the shared container so that the shared container service responds to the new model storage instruction and stores the new neural network model into a corresponding storage space according to the model configuration information.

The shared container service is used for sharing and storing the neural network models corresponding to the model online services, the shared container service is provided with a plurality of model pools, one or more service storage spaces are stored in the model pools, and each service storage space is used for storing each version of neural network model required by the operation of the model online service corresponding to the service storage space; the shared container service performs data isolation on the neural network model required by the separated storage operation of the online service of each model through the plurality of model pools and the service storage space, so that the mutual coverage of data is avoided, and the system stability of the online service of each model is improved.

And after responding to the new model storage instruction pushed by the current model updating service, the shared container service acquires the new neural network model and the corresponding configuration information thereof through a data communication link established with the current model updating service, or acquires the new neural network model and the corresponding configuration information pushed by the current model updating service, and in addition, when the current model updating service stores the new neural network model and the corresponding configuration information thereof into the model cloud space, after responding to the new model storage instruction, the shared container service correspondingly acquires the new neural network model and the corresponding configuration information thereof from the model cloud space.

Regarding the specific way of the shared container service for storing and processing the new neural network model according to the model configuration information, after the shared container service acquires the new neural network model and the model configuration information corresponding to the new neural network model, the shared container service queries a target model pool corresponding to a model pool identifier in a plurality of model pools according to the model pool identifier included in the model configuration information, and queries a target service storage space corresponding to the service identifier in a plurality of service storage spaces in the target model pool according to the service identifier included in the model configuration information, so as to store the new neural network model and the model version number as mapping relationship data into the target service storage space.

Step S13, when the target model online service meets the preset new model deployment strategy, pushing a shared storage path corresponding to the new neural network model to the target model online service, and driving the target model online service to acquire the new neural network model from the shared container service according to the shared storage path for deployment:

and when the shared container service finishes the storage processing of the new neural network model, the current model updating service monitors the target model online service, and when the monitored target model online service meets the preset new model deployment processing, the current model updating service drives the target model online service to acquire the new neural network model from the shared container service for deployment.

The new model deployment strategy is used by the current model updating service for judging whether to drive the target model online service to deploy a new neural network model, the current model updating service can judge whether the current time exceeds the preset model updating time for the target model online service, if so, the target model online service is driven to deploy the new neural network model to be online, or the current model updating service judges whether the workload of the target model online service needs to be online to the new neural network model so as to improve the service efficiency of the new neural network model, or the current updating service judges whether the storage capacity of the target model online service can meet the storage requirement of the new neural network model, and if so, the target model online service deploys the new neural network model to be online; the technical personnel in the field can flexibly design the new model deployment strategy according to the business scene of the model online service so as to deploy the developed new model deployment strategy to the current model updating service, and then the strategy judgment is carried out by the current model updating service.

When the current model updating service determines that the target model online service meets a preset new model deployment strategy, pushing the shared storage path which is pushed by the shared container service and points to the new neural network model to the target model online service so that the target model pointing service can acquire the new neural network model and the corresponding model version number thereof from the shared container service according to the shared storage path; and the shared container service correspondingly stores the new neural network model and the model version number corresponding to the new neural network model into a target service storage space, generates a shared storage path representing the new neural network model from a target model pool to the target service storage space, and pushes the shared storage path into the current model updating service.

After the target model online service acquires the shared storage path, the new neural network model and the corresponding model version number thereof are acquired from the target service storage space of the target model pool pointed by the shared storage path in the shared container service according to the shared storage path, and then the new neural network model is deployed so as to bring the current online old neural network model on line and bring the new neural network model on line to provide a model reasoning service based on the new neural network model for a client or a server associated with the target model online service.

The current model sharing service can push the shared storage path to a plurality of model online services applicable to the new neural network model, so that the model online services acquire the new neural network model from the shared container service for deployment.

According to the typical implementation mode of the method, the method can be used for automatically deploying and managing the neural network models to which the online services of the models belong for the platform, the model deployment management system comprises the model updating service and the shared container service, the model updating service provides a deployment interface for a development end where developers are located, so that the developers can upload the newly-developed neural network models to the model updating service to deploy the new models, manual deployment of the developers is not needed, the time of the developers is saved, and the updating efficiency of the models is improved; secondly, the shared container service in the method can be used for online service of each version of neural network model for the model in the platform automation management platform, and the neural network models of each version of online service of each model are grouped and managed in pools in the shared container service so as to realize data isolation of each neural network model, ensure the safety of model data and improve the system stability of the online service of the model; in addition, the neural network models of different versions of the model online service are managed in the shared container service of the method, so that a platform can deploy multiple versions of models in the shared container service to realize the running of the model online service multi-version models.

The above exemplary embodiments and variations thereof fully disclose embodiments of the model deployment update processing method of the present application, but many variations of the method can be deduced by transforming and augmenting some technical means, and other embodiments are summarized as follows:

in an embodiment, referring to fig. 3, the step of obtaining the new neural network model and the corresponding model configuration information included in the event in response to the model update event acting on the target model online service includes:

step S111, receiving a new neural network model and corresponding model configuration information thereof pushed by the model development terminal through the model upload interface:

the current model update service receives a new neural network model and corresponding model configuration information pushed by the model development terminal through a model upload interface, which is a data communication link interface provided by the current model update service to the model development terminal.

Step S112, correspondingly storing the new neural network model and the model configuration information into a model cloud storage space:

after the current model updating service acquires the new neural network model and the model configuration information, the new neural network model and the model configuration information are stored in the cloud storage space, so that the shared container service acquires the new neural network model and the model configuration information from the cloud storage space to perform storage processing on the new neural network model, and the cloud storage space stores a plurality of neural network models and corresponding model configuration information so that the shared container service associated with the neural network models acquires the corresponding new neural network model to perform storage processing.

In this embodiment, the current model update service stores the newly developed neural network model and the corresponding model configuration information in the cloud space to cope with model forwarding of the update container service, thereby effectively reducing data transmission pressure of the current model update service and saving local storage space.

In an embodiment, referring to fig. 4, the step of driving the shared container service to respond to the instruction to store the new neural network model corresponding to the model version number to the target service storage space corresponding to the service identifier in the target model pool corresponding to the model pool identifier includes the following steps executed by the shared container service:

step S121, responding to a new model storage instruction pushed by the model update service, and acquiring a new neural network model corresponding to the instruction and model configuration information thereof from the model cloud storage space:

and the shared container service responds to the new model storage instruction pushed by the model updating service, determines a new neural network model pointed by the new storage instruction, and further acquires the new neural network model and the corresponding model configuration information thereof through a data communication link established with the model cloud storage space.

Step S122, obtaining the model pool identifier and the model service identifier included in the model configuration information, and querying a target service storage space corresponding to the service identifier in the target model pool corresponding to the model pool identifier:

and the shared container service analyzes the model configuration information, acquires the model pool identification and the model service identification contained in the model configuration information, determines a target model pool corresponding to the model pool identification, and further queries a target service storage space corresponding to the model service identification in the target model pool.

Step S123, obtaining a model version number included in the model configuration information, and correspondingly storing the new neural network model and the model version number into the target service storage space:

and after the shared container service determines a target service storage space, correspondingly storing the new neural network model and the model version number contained in the model configuration information thereof into the target service storage space.

In this embodiment, the shared container service acquires the new neural network model and the model configuration information from the cloud storage space to perform storage processing of the new neural network model, performs model isolation of the new neural network model, prevents data pollution and data coverage of the new neural network model, ensures the security of model data, and is convenient for acquisition of the new model of the corresponding model online service.

In an embodiment, referring to fig. 5, when the target model online service satisfies a preset new model deployment policy, the step of pushing the shared storage path corresponding to the new neural network model to the target model online service includes:

step S131, obtaining the model update time preset for the target model online service, and monitoring whether the current time exceeds the model update time:

the current model updating service monitors the target model online service, acquires the model updating time preset for the target model online service, and judges whether the current time exceeds the model updating time.

Step S132, when it is monitored that the current time exceeds the model updating time, a model updating instruction is pushed to the target model online service, and the model updating instruction comprises the following shared storage path:

and when the current model updating service monitors that the current time exceeds the model updating time, pushing a model updating instruction containing the shared storage path to the target model online service so that the target online service corresponds to the instruction, and further acquiring a new neural network model and a model version number thereof from the shared container service according to the shared storage path to perform model deployment.

In this embodiment, the model sharing service drives the corresponding model online service to deploy the new neural network model only when the time reaches the model update time, so as to effectively perform new model deployment and update, and ensure the current operation stability of the model online service.

In an embodiment, referring to fig. 6, the step of the driving target model online service acquiring a new neural network model from the shared container service according to the shared storage path for deployment includes the following steps performed by the target model online service:

step S131', in response to the model update command pushed by the model update service, acquiring the shared storage path included in the command:

and when the target model online service receives the model updating instruction pushed by the model updating service, acquiring the shared storage path contained in the instruction.

Step S132', according to the shared storage path, obtaining the new neural network model from the target service storage space in the target model pool of the shared container service:

and after the target model online service acquires the shared storage path, acquiring the new neural network model and the model version number corresponding to the new neural network model from the target service storage space of the target model pool pointed by the shared storage path in the shared container service according to the shared storage path.

Step S133', after the deployment of the new neural network model is completed, the current old neural network model is offline, and the new neural network model is online to provide a reasoning function for the current model service:

and the target model online service acquires the new neural network model and the corresponding model version number thereof, deploys the new neural network model to enable the current online old neural network model to be online, and then the new neural network model is online to provide a model inference service based on the new neural network model for a client or a server associated with the target model online service.

In this embodiment, the model online service may obtain a new model from the shared container for deployment, and then perform model hot update to provide a new version of the model inference service, without obtaining the new model from the model update service, so as to save data transmission pressure of the model update service and improve the overall execution efficiency of the model hot update system.

Further, a model deployment update processing apparatus of the present application can be constructed by functionalizing the steps in the methods disclosed in the above embodiments, and according to this idea, please refer to fig. 7, wherein in an exemplary embodiment, the apparatus includes: the update event response module 11 is configured to respond to a model update event that acts on a target model online service, and acquire a new neural network model and model configuration information corresponding to the new neural network model, where the model configuration information includes a model pool identifier, a model service identifier, and a model version number; the model sharing storage module 12 is configured to push a new model storage instruction to a shared container service, and drive the shared container service to respond to the instruction so as to correspondingly store the new neural network model and the model version number to a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier according to the model configuration information; and the new model deployment module 13 is configured to, when the target model online service meets a preset new model deployment policy, push a shared storage path corresponding to the new neural network model to the target model online service, and drive the target model online service to acquire the new neural network model from the shared container service according to the shared storage path for deployment.

In one embodiment, the update event response module 11 comprises: the model acquisition submodule is used for receiving a new neural network model and corresponding model configuration information thereof which are pushed by the model development end through the model uploading interface; and the model cloud storage submodule is used for correspondingly storing the new neural network model and the model configuration information into a model cloud storage space.

In one embodiment, the model shared storage module 12 comprises: the storage instruction response submodule is used for responding to a new model storage instruction pushed by the model updating service and acquiring a new neural network model corresponding to the instruction and model configuration information thereof from the model cloud storage space; a storage space query submodule, configured to obtain a model pool identifier and a model service identifier included in the model configuration information, and query a target service storage space corresponding to the service identifier in a target model pool corresponding to the model pool identifier; and the model storage submodule is used for acquiring a model version number contained in the model configuration information and correspondingly storing the new neural network model and the model version number into the target service storage space.

In one embodiment, the new model deployment module 13 includes: the model updating time monitoring submodule is used for acquiring model updating time preset for the target model online service and monitoring whether the current time exceeds the model updating time; and the model updating instruction pushing submodule is used for pushing a model updating instruction to the target model online service when the current time is monitored to exceed the model updating time, and the model updating instruction comprises the shared storage path.

In another embodiment, the new model deployment module 13 further includes: the update instruction response submodule is used for responding to a model update instruction pushed by the model update service and acquiring a shared storage path contained in the instruction; a new model obtaining sub-module, configured to obtain the new neural network model from the target service storage space in the target model pool of the shared container service according to the shared storage path; and the new model online sub-module is used for offline the current and old neural network models after the deployment of the new neural network model is completed, and online the new neural network model to provide a reasoning function for the current model service.

In order to solve the foregoing technical problem, an embodiment of the present application further provides a computer device, configured to run a computer program implemented according to the model deployment update processing method. Referring to fig. 8, fig. 8 is a block diagram of a basic structure of a computer device according to the present embodiment.

As shown in fig. 8, the internal structure of the computer device is schematic. The computer device includes a processor, a non-volatile storage medium, a memory, and a network interface connected by a system bus. The non-volatile storage medium of the computer device stores an operating system, a database and computer readable instructions, the database can store control information sequences, and the computer readable instructions, when executed by the processor, can enable the processor to implement a model deployment update processing method. The processor of the computer device is used for providing calculation and control capability and supporting the operation of the whole computer device. The memory of the computer device may have stored therein computer readable instructions that, when executed by the processor, may cause the processor to perform a model deployment update processing method. The network interface of the computer device is used for connecting and communicating with the terminal. It will be appreciated by those skilled in the art that the configuration shown in fig. 8 is a block diagram of only a portion of the configuration associated with the present application, and is not intended to limit the computing device to which the present application may be applied, and that a particular computing device may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In this embodiment, the processor is configured to execute specific functions of each module/sub-module in the model deployment and update processing apparatus of the present application, and the memory stores program codes and various types of data required for executing the modules. The network interface is used for data transmission to and from a user terminal or a server. The memory in this embodiment stores program codes and data required for executing all modules/submodules in the model deployment and update processing apparatus, and the server can call the program codes and data of the server to execute the functions of all the submodules.

The present application also provides a non-volatile storage medium, where the model deployment update processing method is written as a computer program and stored in the storage medium in the form of computer readable instructions, and when the computer readable instructions are executed by one or more processors, the program is executed in a computer, so that the one or more processors execute the steps of any of the above-described model deployment update processing methods.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the computer program is executed. The storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM).

To sum up, the method and the device for managing the new neural network model deployment and updating automatically realize the automatic management of the new neural network model deployment and updating, save the deployment time of the new model, improve the hot updating efficiency of the model, and perform isolation management on model data to ensure the data safety of the model.

It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

Those of skill in the art will appreciate that the various operations, methods, steps in the processes, acts, or solutions discussed in this application can be interchanged, modified, combined, or eliminated. Further, other steps, measures, or schemes in various operations, methods, or flows that have been discussed in this application can be alternated, altered, rearranged, broken down, combined, or deleted. Further, the steps, measures, and schemes in the various operations, methods, and flows disclosed in the present application in the prior art can also be alternated, modified, rearranged, decomposed, combined, or deleted.

The foregoing is only a partial embodiment of the present application, and it should be noted that, for those skilled in the art, several modifications and decorations can be made without departing from the principle of the present application, and these modifications and decorations should also be regarded as the protection scope of the present application.

Claims

1. A model deployment updating processing method is characterized by comprising the following steps:

2. The method according to claim 1, wherein the step of obtaining the new neural network model and the corresponding model configuration information included in the event in response to the model update event acting on the target model online service comprises:

3. The method according to claim 2, wherein the step of driving the shared container service to respond to the instruction to store the new neural network model corresponding to the model version number in the target service storage space corresponding to the service identifier in the target model pool corresponding to the model pool identifier according to the model configuration information comprises the following steps performed by the shared container service:

4. The method according to claim 1, wherein the step of pushing the shared storage path corresponding to the new neural network model to the target model online service when the target model online service satisfies a preset new model deployment policy includes:

5. The method of claim 1, wherein the step of the target model online service acquiring a new neural network model from the shared container service for deployment according to the shared storage path comprises the following steps performed by the target model online service:

according to the shared storage path, acquiring the new neural network model from the target service storage space in the target model pool of the shared container service;

6. The method according to claim 1, wherein in the step of pushing the shared storage path corresponding to the new neural network model to the target model online service, the shared storage path is pushed by a shared container service to a current model update service, and the shared storage path points to the new neural network model in the target service storage space of the target model pool.

7. A model deployment update processing apparatus, comprising:

8. An electronic device comprising a central processor and a memory, characterized in that the central processor is configured to invoke execution of a computer program stored in the memory to perform the steps of the method according to any one of claims 1 to 6.

9. A non-volatile storage medium, characterized in that it stores, in the form of computer-readable instructions, a computer program implemented according to the method of any one of claims 1 to 6, which, when invoked by a computer, performs the steps comprised by the method.

10. A computer program product comprising computer program/instructions, characterized in that the computer program/instructions, when executed by a processor, implement the steps of the method of any one of claims 1 to 6.