CN115829060A - Machine learning model display and management platform system, method and electronic equipment - Google Patents
Machine learning model display and management platform system, method and electronic equipmentDownload PDFInfo
- Publication number
- CN115829060A CN115829060ACN202211701170.3ACN202211701170ACN115829060ACN 115829060 ACN115829060 ACN 115829060ACN 202211701170 ACN202211701170 ACN 202211701170ACN 115829060 ACN115829060 ACN 115829060A
- Authority
- CN
- China
- Prior art keywords
- model
- display
- machine learning
- unit
- snapshot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本发明涉及人工智能技术领域，尤其涉及一种机器学习模型展示与管理平台系统、方法及电子设备。The invention relates to the technical field of artificial intelligence, in particular to a machine learning model display and management platform system, method and electronic equipment.
背景技术Background technique
目前，随着计算机硬件技术的提升，大数据技术得到了迅速发展，导致了基于大数据的机器学习及人工智能技术也在快速发展，从而使得机器学习及人工智能模型更新的速度愈来越快。而无论是对于新兴的互联网企业，还是对于传统的制造业、金融业，机器学习及人工智能都有着举足轻重的地位，正因如此，面对机器学习及人工智能模型更新的速度愈来越快的问题，各个互联网企业都研发了各自的机器学习及人工智能开发平台。At present, with the improvement of computer hardware technology, big data technology has developed rapidly, leading to the rapid development of machine learning and artificial intelligence technology based on big data, which makes the update speed of machine learning and artificial intelligence models faster and faster. . Regardless of whether it is for emerging Internet companies or for traditional manufacturing and financial industries, machine learning and artificial intelligence play a pivotal role. Because of this, in the face of the faster and faster update of machine learning and artificial intelligence models The problem is that various Internet companies have developed their own machine learning and artificial intelligence development platforms.
现阶段的机器学习及人工智能开发平台，所解决的问题主要是机器学习及人工智能模型的快速开发、快速调优，所面向的用户主要是负责模型开发的专业研发人员，因此，无论是模型的开发、调优、验证，还是模型投产后的效果展示，都有赖于专业研发人员采用专业手段来进行，而对于产品经理、业务负责人员这些实际用户而言，现有平台存在的缺点包括：The current machine learning and artificial intelligence development platform mainly solves the problems of rapid development and rapid tuning of machine learning and artificial intelligence models. The development, tuning, and verification of the model, as well as the effect display after the model is put into production, all rely on professional R&D personnel to use professional methods. For actual users such as product managers and business managers, the shortcomings of the existing platform include:
1、无法直观地对模型进行效果展示，并难以自主地对模型进行效果验证与评价；1. It is impossible to intuitively display the effect of the model, and it is difficult to independently verify and evaluate the effect of the model;
2、模型的生命周期管理无法自动化，同一项目的多个模型迭代、效果对比、版本管理等都由专业研发人员或业务负责人员手动操作，工作量大且时间成本高；2. The life cycle management of the model cannot be automated. Multiple model iterations, effect comparisons, version management, etc. of the same project are manually operated by professional R&D personnel or business responsible personnel, with heavy workload and high time cost;
3、无法直观地对不同项目之间的模型进行展示，易导致因建模目标相同而出现重复的工作内容。3. It is impossible to intuitively display the models between different projects, which may easily lead to repeated work content due to the same modeling goals.
发明内容Contents of the invention
本发明的主要目的在于提供一种机器学习模型展示与管理平台系统、方法及电子设备，旨在对于机器学习及人工智能的模型算法，可以直观地进行展示，并自主地进行效果验证与评价，同时实现自动化的生命周期管理。The main purpose of the present invention is to provide a machine learning model display and management platform system, method and electronic equipment, aiming at visually displaying machine learning and artificial intelligence model algorithms, and independently performing effect verification and evaluation, At the same time, automatic life cycle management is realized.
为实现上述目的，本发明提供一种机器学习模型展示与管理平台系统，包括：前端模块，用于响应于用户的控制指令，执行所述控制指令对应的功能服务；后台模块，用于基于所述功能服务进行后台数据处理；其中，In order to achieve the above object, the present invention provides a machine learning model display and management platform system, including: a front-end module, used to respond to the user's control instructions, and execute the functional services corresponding to the control instructions; The above functional services perform background data processing; among them,
所述前端模块包括：The front-end modules include:
模型开发单元，用于执行模型开发服务，并获得模型快照；A model development unit, configured to perform model development services and obtain model snapshots;
模型投产单元，用于获取目标模型快照，对所述目标模型快照进行模型投产的操作；A model production unit, configured to obtain a target model snapshot, and perform a model production operation on the target model snapshot;
模型市场单元，用于在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作；The model market unit is used to determine the model to be displayed among the models that have been successfully put into production, and perform model display operations on the model to be displayed;
所述后台模块包括：The background module includes:
模型计算单元，用于对模型的训练、评估以及预测进行计算，并输出结果文件；The model calculation unit is used to calculate the training, evaluation and prediction of the model, and output the result file;
数据展示单元，用于对所述结果文件中的指标数据进行展示的操作。The data display unit is used for displaying the indicator data in the result file.
优选地，所述执行模型开发服务，并获得模型快照，包括：Preferably, the executing model development service and obtaining a model snapshot includes:
基于所述用户的控制指令，确定开发模式；determining a development mode based on the user's control instruction;
获取所述用户的输入数据，基于所述开发模式以及所述输入数据进行模型开发的操作；Obtain the input data of the user, and perform model development operations based on the development mode and the input data;
保存开发完成的模型的快照，作为所述模型快照。A snapshot of the developed model is saved as the model snapshot.
优选地，所述获取目标模型快照，对所述目标模型快照进行模型投产的操作，包括：Preferably, said acquiring a snapshot of the target model, and performing a model production operation on the snapshot of the target model include:
基于所述用户的控制指令，在所述模型快照中确定所述目标模型快照；determining the target model snapshot in the model snapshot based on the user's control instruction;
控制所述目标模型快照运行，确定是否运行成功；Control the operation of the snapshot of the target model to determine whether the operation is successful;
若运行不成功，则将所述目标模型快照返回至所述模型开发单元进行重新开发的操作；或者，If the operation is unsuccessful, returning the snapshot of the target model to the model development unit for redevelopment; or,
若运行成功，则确定所述目标模型快照投产成功。If the operation is successful, it is determined that the target model snapshot is successfully put into production.
优选地，所述在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作，包括：Preferably, the determination of the model to be displayed in the model successfully put into production, and the operation of model display for the model to be displayed include:
基于所述用户的控制指令，在投产成功的模型中确定所述待展示模型；Based on the user's control instruction, determine the model to be displayed among the models successfully put into production;
在预设界面中展示所述待展示模型对应的基本信息，其中，所述基本信息包括数据集、模型开发代码、数据处理及建模过程、模型效果以及其他说明。The basic information corresponding to the model to be displayed is displayed on a preset interface, wherein the basic information includes data sets, model development codes, data processing and modeling processes, model effects, and other descriptions.
优选地，所述对模型的训练、评估以及预测进行计算，并输出结果文件，包括：Preferably, the training, evaluation and prediction of the model are calculated, and the output result file includes:
获取所述用户输入的代码；Obtain the code input by the user;
将所述代码提交至预设容器环境运行，并获得输出文件，作为所述结果文件。Submit the code to a preset container environment for operation, and obtain an output file as the result file.
优选地，所述对所述结果文件中的指标数据进行展示的操作，包括：Preferably, the operation of displaying the indicator data in the result file includes:
获取所述结果文件，其中，所述结果文件包括所述指标数据以及所述指标数据对应的展示格式；Obtaining the result file, wherein the result file includes the index data and a display format corresponding to the index data;
基于所述展示格式对所述指标数据进行展示的操作。An operation of displaying the indicator data based on the display format.
优选地，所述后端模块，还包括：Preferably, the back-end module also includes:
对象存储单元，用于存储所述用户上传的数据文件以及所述结果文件，其中，所述结果文件具有对应的对象存储空间。The object storage unit is configured to store the data file uploaded by the user and the result file, wherein the result file has a corresponding object storage space.
优选地，所述后端模块，还包括：Preferably, the back-end module also includes:
权限管理单元，用于基于所述用户的用户信息确定所述用户对应的角色，并基于所述角色确定所述用户的系统权限。The authority management unit is configured to determine a role corresponding to the user based on the user information of the user, and determine the system authority of the user based on the role.
此外，为实现上述目的，本发明还提供一种机器学习模型展示与管理方法，应用于机器学习模型展示与管理平台系统，所述机器学习模型展示与管理平台系统包括前端模块以及后台模块，所述前端模块包括模型开发单元、模型投产单元以及模型市场单元，所述后台模块包括模型计算单元以及数据展示单元，所述机器学习模型展示与管理方法包括：In addition, in order to achieve the above object, the present invention also provides a machine learning model display and management method, which is applied to the machine learning model display and management platform system, and the machine learning model display and management platform system includes a front-end module and a background module. The front-end module includes a model development unit, a model production unit, and a model market unit, the background module includes a model calculation unit and a data display unit, and the machine learning model display and management method includes:
所述前端模块响应于用户的控制指令，执行所述控制指令对应的功能服务；The front-end module executes the functional service corresponding to the control instruction in response to the user's control instruction;
所述后台模块基于所述功能服务进行后台数据处理；The background module performs background data processing based on the functional service;
所述模型开发单元执行模型开发服务，并获得模型快照；The model development unit executes a model development service and obtains a model snapshot;
所述模型投产单元获取目标模型快照，对所述目标模型快照进行模型投产的操作；The model production unit acquires a target model snapshot, and performs a model production operation on the target model snapshot;
所述模型市场单元在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作；The model market unit determines the model to be displayed among the models that have been successfully put into production, and performs a model display operation on the model to be displayed;
所述模型计算单元对模型的训练、评估以及预测进行计算，并输出结果文件；The model calculation unit calculates the training, evaluation and prediction of the model, and outputs the result file;
所述数据展示单元对所述结果文件中的指标数据进行展示的操作。The data display unit displays the indicator data in the result file.
此外，为实现上述目的，本发明还提供一种电子设备，所述电子设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的机器学习模型展示与管理程序，所述机器学习模型展示与管理程序被所述处理器执行时实现如上所述的机器学习模型展示与管理方法的步骤。In addition, in order to achieve the above object, the present invention also provides an electronic device, which includes: a memory, a processor, and a machine learning model display and management program stored in the memory and operable on the processor When the machine learning model display and management program is executed by the processor, the steps of the above machine learning model display and management method are realized.
与现有技术相比，本发明提出的机器学习模型展示与管理平台系统，有益效果为：Compared with the prior art, the machine learning model display and management platform system proposed by the present invention has beneficial effects as follows:
1、可以直观地对模型进行效果展示，并且可以自主地对模型进行效果验证与评价；1. It can intuitively display the effect of the model, and can independently verify and evaluate the effect of the model;
2、实现模型的生命周期管理自动化，可以对同一个项目的多个模型的迭代过程进行管理，并且可以方便地对不同版本的模型进行效果对比；2. Realize the automation of model life cycle management, manage the iterative process of multiple models of the same project, and compare the effects of different versions of models conveniently;
3、可以在一个界面中直观地对不同项目之间的模型进行展示，避免因建模目标相同而出现重复的工作内容。3. The models between different projects can be displayed intuitively in one interface, avoiding duplication of work content due to the same modeling goals.
附图说明Description of drawings
图1是本发明实施例方案涉及的硬件运行环境的设备结构示意图；Fig. 1 is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present invention;
图2为本发明机器学习模型展示与管理平台系统具体实施方式的模块示意图；Fig. 2 is the module schematic diagram of the embodiment of the machine learning model display and management platform system of the present invention;
图3为本发明机器学习模型展示与管理方法的流程示意图。Fig. 3 is a schematic flowchart of a method for displaying and managing a machine learning model of the present invention.
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。The realization of the purpose of the present invention, functional characteristics and advantages will be further described in conjunction with the embodiments and with reference to the accompanying drawings.
具体实施方式Detailed ways
应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.
具体地，参照图1，图1为本发明机器学习模型展示与管理平台系统涉及的硬件运行环境的设备结构示意图。Specifically, referring to FIG. 1 , FIG. 1 is a schematic diagram of a device structure of a hardware operating environment involved in a machine learning model display and management platform system of the present invention.
如图1所示，该设备可以包括：处理器1001，例如CPU，网络接口1004，用户接口1003，存储器1005，通信总线1002。其中，通信总线1002用于实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)，可选用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如WI-FI接口)。存储器1005可以是高速RAM存储器，也可以是稳定的存储器(non-volatile memory)，例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。As shown in FIG. 1 , the device may include: a
本领域技术人员可以理解，图1中示出的设备结构并不构成对该设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。Those skilled in the art can understand that the device structure shown in FIG. 1 does not constitute a limitation to the device, and may include more or less components than shown in the figure, or combine some components, or arrange different components.
如图1所示，作为一种计算机存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及机器学习模型展示与管理程序。其中，操作系统是管理和控制设备硬件和软件资源的程序，支持机器学习模型展示与管理程序以及其它软件或程序的运行；网络通信模块用于管理和控制网络接口1002；用户接口1003主要用于与客户端进行数据通信；网络接口1004主要用于与服务器建立通信连接；而处理器1001可以用于调用存储器1005中存储的机器学习模型展示与管理程序。As shown in FIG. 1 , the
本发明还提供一种机器学习模型展示与管理平台系统，参照图2，图2为本发明机器学习模型展示与管理平台系统的模块示意图。The present invention also provides a machine learning model display and management platform system. Referring to FIG. 2 , FIG. 2 is a schematic module diagram of the machine learning model display and management platform system of the present invention.
下面结合附图对本发明做进一步的详细描述：Below in conjunction with accompanying drawing, the present invention is described in further detail:
本发明提供的机器学习模型展示与管理平台系统，包括：前端模块10，用于响应于用户的控制指令，执行所述控制指令对应的功能服务；后台模块20，用于基于所述功能服务进行后台数据处理；其中，The machine learning model display and management platform system provided by the present invention includes: a front-end module 10, configured to respond to a user's control instruction, and execute the functional service corresponding to the control instruction; a background module 20, configured to perform Background data processing; where,
所述前端模块10包括：The front-end module 10 includes:
模型开发单元110，用于执行模型开发服务，并获得模型快照；A model development unit 110, configured to execute model development services and obtain model snapshots;
模型投产单元120，用于获取目标模型快照，对所述目标模型快照进行模型投产的操作；The model production unit 120 is configured to obtain a target model snapshot, and perform a model production operation on the target model snapshot;
模型市场单元130，用于在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作；The model market unit 130 is used to determine the model to be displayed among the models that have been successfully put into production, and perform model display operations on the model to be displayed;
所述后台模块20包括：Described background module 20 comprises:
模型计算单元210，用于对模型的训练、评估以及预测进行计算，并输出结果文件；Model calculation unit 210, used for calculating the training, evaluation and prediction of the model, and outputting the result file;
数据展示单元220，用于对所述结果文件中的指标数据进行展示的操作。The data display unit 220 is configured to display the indicator data in the result file.
需要说明的是，机器学习模型展示与管理平台系统由服务端以及用户端组成，服务端包括服务器主机以及Web应用程序集合，Web应用程序集合以及对应的数据库部署在服务器主机中，通过网络连接，为用户提供响应以及功能服务；用户端为具有网络权限的个人计算机，用户通过Web浏览器进行访问，使用系统提供的功能服务。It should be noted that the machine learning model display and management platform system consists of a server and a client. The server includes a server host and a collection of web applications. The collection of web applications and the corresponding database are deployed on the server host and connected through the network. Provide response and functional services for users; the client is a personal computer with network authority, and users access it through a Web browser to use the functional services provided by the system.
具体地，机器学习模型展示与管理平台系统包括前端模块10以及后台模块20，前端模块所包括的模型开发单元110、模型投产单元120以及模型市场单元130，在客户端中可以显示为模型开发小程序、模型投产小程序以及模型市场小程序，首先，通过模型开发小程序，用户可以选择不同的方式进行建模，并保存开完完成的模型的快照，作为模型快照，其次，在模型投产小程序中，用户可以选择一个模型快照作为目标模型快照进行投产，模型投产小程序控制该目标模型快照运行，并确定是否可以运行成功，若运行不成功，则将该目标模型快照返回至模型开发小程序，进行重新开发，若运行成功，则确定目标模型快照投产成功，再次，用户可以在投产成功的模型中，确定需要进行展示的待展示模型，模型市场小程序在预设界面中展示对待展示模型进行展示，包括待展示模型的数据集、模型开发代码、数据处理及建模过程、模型效果以及其他说明。而后台模块所包括的模型计算单元210以及数据展示单元220，可以对模型的训练、评估以及预测进行计算，并输出结果文件，还可以对结果文件中的指标数据进行展示。本发明提供的机器学习模型Specifically, the machine learning model display and management platform system includes a front-end module 10 and a back-end module 20. The model development unit 110, model production unit 120, and model market unit 130 included in the front-end module can be displayed as a model development widget in the client. program, model production applet, and model market applet, firstly, through the model development applet, users can choose different ways to model, and save the snapshot of the completed model as a model snapshot; In the program, the user can select a model snapshot as the target model snapshot to put into production. The model production applet controls the running of the target model snapshot and determines whether the running is successful. If the running is unsuccessful, the target model snapshot is returned to the model development applet Program, re-develop, if the operation is successful, it is determined that the snapshot of the target model has been successfully put into production, and again, the user can determine the model to be displayed that needs to be displayed in the model that has been successfully put into production, and the model market applet will be displayed in the preset interface The model is displayed, including the data set of the model to be displayed, model development code, data processing and modeling process, model effect and other descriptions. The model calculation unit 210 and data display unit 220 included in the background module can calculate the training, evaluation and prediction of the model, output the result file, and display the index data in the result file. The machine learning model provided by the present invention
展示与管理平台系统，可以直观地对模型进行效果展示，并且可以自主地对5模型进行效果验证与评价；还可以实现模型的生命周期管理自动化，对同一个项目的多个模型的迭代过程进行管理，并且方便地对不同版本的模型进行效果对比；此外，通过在一个界面中直观地对不同项目之间的模型进行展示，可以避免因建模目标相同而出现重复的工作内容。The display and management platform system can intuitively display the effect of the model, and can independently verify and evaluate the effect of the 5 models; it can also realize the automation of the life cycle management of the model, and carry out the iterative process of multiple models of the same project. Management, and conveniently compare the effects of different versions of the model; in addition, by visually displaying the models between different projects in one interface, it is possible to avoid duplication of work content due to the same modeling goals.
可选地，如图2所示，前端模块10还可以包括门户单元100，用户可以0通过该门户单元进入模型开发单元110、模型投产单元120以及模型市场单元130。Optionally, as shown in FIG. 2 , the front-end module 10 may also include a portal unit 100 through which users can enter the model development unit 110 , model production unit 120 and model marketing unit 130 .
可选地，如图2所示，后台模块20还可以包括数据库单元200，该数据库单元通过关系型数据库管理系统MySQL对系统运行所需要保存的所有数据进行存储。Optionally, as shown in FIG. 2 , the background module 20 may also include a database unit 200, which stores all data required for system operation through the relational database management system MySQL.
5进一步地，模型开发单元110，用于执行模型开发服务，并获得模型快照，5 Further, the model development unit 110 is configured to execute the model development service and obtain the model snapshot,
包括：include:
步骤S1101，基于所述用户的控制指令，确定开发模式；Step S1101, based on the user's control instruction, determine the development mode;
步骤S1102，获取所述用户的输入数据，基于所述开发模式以及所述输入数据进行模型开发的操作；Step S1102, acquiring the input data of the user, and performing model development based on the development mode and the input data;
0步骤S1103，保存开发完成的模型的快照，作为所述模型快照。0 Step S1103, saving the snapshot of the developed model as the model snapshot.
具体地，根据用户的控制指令，模型开发单元110可以获取用户在Web页面中编写的python代码，以及相应的shell脚本来进行数据处理、模型训练的过程；也可以复制其他开放的模型，在该模型的代码基础上进行开发。其Specifically, according to the user's control instruction, the model development unit 110 can obtain the python code written by the user in the Web page, and the corresponding shell script to perform data processing and model training; it can also copy other open models, in this The code base of the model is developed. That
中，用户输入的数据需要符合系统所要求的CSV格式规范。此外，模型开发5单元110还可以保存开发完成的模型的快照，作为模型快照，每个模型快照In , the data entered by the user must conform to the CSV format specification required by the system. In addition, the model development 5 unit 110 can also save a snapshot of the developed model as a model snapshot, each model snapshot
为模型的一个版本，可以记录该版本的模型效果指标。从而通过模型快照直观地对模型效果进行展示，并且还可以在模型开发时自主地对模型进行效果验证。is a version of the model, and the model performance index of this version can be recorded. In this way, the effect of the model can be intuitively displayed through the snapshot of the model, and the effect of the model can also be independently verified during model development.
可选地，模型开发单元110还可以对正在开发的模型进行保存，但未开0发完成的模型可以留待后续进行开发，不可以作为模型快照用于模型投产。Optionally, the model development unit 110 can also save the model being developed, but the undeveloped model can be reserved for subsequent development, and cannot be used as a model snapshot for model production.
进一步地，模型投产单元120，用于获取目标模型快照，对所述目标模型快照进行模型投产的操作，包括：Further, the model production unit 120 is configured to obtain a target model snapshot, and perform a model production operation on the target model snapshot, including:
步骤S1201，基于所述用户的控制指令，在所述模型快照中确定所述目标模型快照；Step S1201, based on the user's control instruction, determine the target model snapshot in the model snapshot;
步骤S1202，控制所述目标模型快照运行，确定是否运行成功；Step S1202, control the running of the snapshot of the target model, and determine whether the running is successful;
步骤S1203，若运行不成功，则将所述目标模型快照返回至所述模型开发单元进行重新开发的操作；或者，Step S1203, if the operation is unsuccessful, return the snapshot of the target model to the model development unit for redevelopment; or,
步骤S1204，若运行成功，则确定所述目标模型快照投产成功。Step S1204, if the operation is successful, it is determined that the snapshot of the target model is put into production successfully.
具体地，模型开发单元110开发并调试完成后，用户可以在保存的模型快照中选择一个模型快照进行投产，模型投产单元120控制该模型快照运行，以再次确定该模型快照是否可以运行成功，若运行不成功，则确定投产失败，需要用户重新回到模型开发单元110对模型进行开发，并重新保存快照快照，若运行成功，则确定可以投产，而后可以发起投产审核，将模型上交至具有审核权限的审核用户，待审核通过后，确定模型投产成功。从而通过模型投产的方式，验证开发完成的模型是否可以运行，确保后续进行展示的模型的完整性。Specifically, after the development and debugging of the model development unit 110 is completed, the user can select a model snapshot from the saved model snapshots for production, and the model production unit 120 controls the operation of the model snapshot to determine whether the model snapshot can run successfully. If the operation is unsuccessful, it is determined that the production has failed, and the user needs to return to the model development unit 110 to develop the model and save the snapshot snapshot again. Audit users with audit authority, after the audit is passed, it is determined that the model has been successfully put into production. In this way, by putting the model into production, it is verified whether the developed model can run, ensuring the integrity of the model to be displayed later.
可选地，模型投产成功后，进行开发的用户与审核用户都有权限确定模型后续是否需要在模型市场单元130进行展示，若确定需要进行展示，则需要上传一张封面图片，用于后续的展示。Optionally, after the model is successfully put into production, both the developing user and the auditing user have the authority to determine whether the model needs to be displayed in the model market unit 130. If it is determined that the model needs to be displayed, a cover image needs to be uploaded for subsequent exhibit.
进一步地，模型市场单元130，用于在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作，具体步骤包括：Further, the model market unit 130 is used to determine the model to be displayed among the models that have been successfully put into production, and perform model display operations on the model to be displayed, and the specific steps include:
步骤S1301，基于所述用户的控制指令，在投产成功的模型中确定所述待展示模型；Step S1301, based on the user's control instruction, determine the model to be displayed among the models successfully put into production;
步骤S1302，在预设界面中展示所述待展示模型对应的基本信息，其中，所述基本信息包括数据集、模型开发代码、数据处理及建模过程、模型效果以及其他说明。Step S1302, displaying the basic information corresponding to the model to be displayed on the preset interface, wherein the basic information includes data set, model development code, data processing and modeling process, model effect and other descriptions.
具体地，用户可以在投产成功的模型中选择用于在模型市场单元130中展示的模型，作为待展示模型，模型市场单元130可以按模型组、业务场景、模型算法这三个维度对待展示模型进行分类与筛选，而后在模型市场单元130的预设界面如模型市场小程序中对待展示模型进行展示，展示内容包括待展示模型对应的基本信息，其中，基本信息包括数据集、模型开发代码、数据处理及建模过程、模型效果以及其他说明。通过这种方式直观地对模型进行多维度的展示，使得平台系统所展示的模型不只是面向专业研发人员，对于实际用户而言可以更易理解，提升了用户体验。Specifically, the user can select a model to be displayed in the model market unit 130 among the models that have been successfully put into production. As a model to be displayed, the model market unit 130 can treat the display model according to the three dimensions of model group, business scenario, and model algorithm. Carry out classification and screening, and then display the model to be displayed on the preset interface of the model market unit 130, such as the model market applet. The displayed content includes the basic information corresponding to the model to be displayed, wherein the basic information includes data sets, model development codes, Data processing and modeling process, model effect and other instructions. In this way, the model is intuitively displayed in multiple dimensions, so that the model displayed by the platform system is not only for professional R&D personnel, but also easier for actual users to understand, which improves the user experience.
可选地，待展示模型对应的基本信息可以归纳为对应的详情界面，预设界面中展示的模型附带各自对应的详情界面的进入按钮，用户浏览预设界面时，可以在选择感兴趣的模型后，通过点击该模型的进入按钮进入该模型的详情界面，从而查看基本信息。Optionally, the basic information corresponding to the model to be displayed can be summarized into the corresponding detail interface, and the models displayed in the preset interface are accompanied by the entry buttons of the corresponding detail interface. When the user browses the preset interface, he can select the model he is interested in After that, click the enter button of the model to enter the details interface of the model to view the basic information.
可选地，在模型市场单元130中，用户可以上传数据文件，对模型进行试运行，用户所上传的数据文件的内容、格式必须符合模型开发者对于数据文件的要求。Optionally, in the model market unit 130, users can upload data files to test run the model. The content and format of the data files uploaded by users must meet the requirements of model developers for data files.
进一步地，模型计算单元210，用于对模型的训练、评估以及预测进行计算，并输出结果文件，包括：Further, the model calculation unit 210 is used to calculate the training, evaluation and prediction of the model, and output the result file, including:
步骤S501，获取所述用户输入的代码；Step S501, acquiring the code input by the user;
步骤S502，将所述代码提交至预设容器环境运行，并获得输出文件，作为所述结果文件。Step S502, submitting the code to a preset container environment for running, and obtaining an output file as the result file.
具体地，模型计算单元210主要负责模型的训练、评估、预测的计算，当用户提交python代码后，模型计算单元210会将用户提交的代码提交至虚拟化的docker容器环境中运行，返回运行后的输出文件，将输出文件作为结果文件并保存。Specifically, the model calculation unit 210 is mainly responsible for model training, evaluation, and prediction calculation. When the user submits the python code, the model calculation unit 210 will submit the code submitted by the user to the virtualized docker container environment to run, and return to run , save the output file as a result file.
进一步地，数据展示单元220，用于对所述结果文件中的指标数据进行展示的操作，包括：Further, the data display unit 220 is used to display the indicator data in the result file, including:
步骤S601，获取所述结果文件，其中，所述结果文件包括所述指标数据以及所述指标数据对应的展示格式；Step S601, obtaining the result file, wherein the result file includes the index data and a display format corresponding to the index data;
步骤S602，基于所述展示格式对所述指标数据进行展示的操作。Step S602, displaying the indicator data based on the display format.
具体地，模型计算单元210输出的结果文件，包括了数据展示单元220所需要展示的指标数据，其中，指标数据的文件需要符合数据展示单元220的要求，包括图像类型(散点图、曲线图、柱状图)、横纵坐标轴的标题与刻度、图像中的数据点，上述内容按照特定的展示格式保存在结果文件中，数据展示单元220获得结果文件后，根据指标数据对应的展示格式对指标数据进行展示的操作。从而通过指标数据的展示对用户提交至计算单元210中的模型的模型效果进行直观的展示。Specifically, the result file output by the model calculation unit 210 includes the index data that the data display unit 220 needs to display, wherein, the file of the index data needs to meet the requirements of the data display unit 220, including the image type (scatter plot, graph , histogram), the title and scale of the horizontal and vertical axes, and the data points in the image. The above contents are stored in the result file according to a specific display format. The operation of displaying indicator data. Therefore, the model effect of the model submitted by the user to the calculation unit 210 is intuitively displayed through the display of the index data.
进一步地，所述后端模块20，还包括：Further, the back-end module 20 also includes:
对象存储单元230，用于存储所述用户上传的数据文件以及所述结果文件，其中，所述结果文件具有对应的对象存储空间。The object storage unit 230 is configured to store the data file uploaded by the user and the result file, wherein the result file has a corresponding object storage space.
具体地，对象存储单元230负责保存用户上传的数据文件，以及模型计算单元210运行后输出的结果文件，为每个模型分配对应的对象存储空间，其中，对象存储空间的大小具有一定限制。从而通过分开存储模型可以避免模型之间的干扰。Specifically, the object storage unit 230 is responsible for saving the data files uploaded by users and the output result files after the model calculation unit 210 runs, and allocates corresponding object storage space for each model, wherein the size of the object storage space has a certain limit. Interference between models can thus be avoided by storing the models separately.
进一步地，所述后端模块20，还包括：Further, the back-end module 20 also includes:
权限管理单元240，用于基于所述用户的用户信息确定所述用户对应的角色，并基于所述角色确定所述用户的系统权限。The authority management unit 240 is configured to determine the role corresponding to the user based on the user information of the user, and determine the system authority of the user based on the role.
具体地，权限管理单元240负责整个机器学习模型展示与管理平台系统的权限管理，其中，该权限管理包括用户、角色与权限三层体系，为每个角色赋予不同的权限后再将角色赋予用户，角色包括：管理员角色，负责整个平台系统的管理，拥有最高的权限；开发者角色：负责模型的开发，拥有模型开发单元110以及模型市场单元130的权限；审核者角色：负责审核模型的投产和展示，避免展示完成度低的模型；浏览者角色：只拥有模型展示单元130的权限，可以浏览该模块中展示的模型。通过不同角色对应不同权限，可以确保机器学习模型展示与管理平台系统管理的有序性，提升用户体验。Specifically, the authority management unit 240 is responsible for the authority management of the entire machine learning model display and management platform system, wherein the authority management includes a three-layer system of users, roles and permissions, and assigns different permissions to each role before assigning the role to the user , roles include: administrator role, responsible for the management of the entire platform system, has the highest authority; developer role: responsible for model development, has the authority of model development unit 110 and model market unit 130; reviewer role: responsible for reviewing model Put into production and display, avoid displaying models with a low degree of completion; viewer role: only has the authority of the model display unit 130, and can browse the models displayed in this module. Different roles correspond to different permissions, which can ensure the orderliness of machine learning model display and management platform system management, and improve user experience.
此外，本发明还提出一种机器学习模型展示与管理方法，应用于机器学习模型展示与管理平台系统，所述机器学习模型展示与管理平台系统包括前端模块以及后台模块，所述前端模块包括模型开发单元、模型投产单元以及模型市场单元，所述后台模块包括模型计算单元以及数据展示单元，参照图3，所述机器学习模型展示与管理方法包括：In addition, the present invention also proposes a machine learning model display and management method, which is applied to the machine learning model display and management platform system. The machine learning model display and management platform system includes a front-end module and a background module. The front-end module includes a model Development unit, model production unit and model market unit, the background module includes a model calculation unit and a data display unit, referring to Figure 3, the machine learning model display and management method includes:
所述前端模块响应于用户的控制指令，执行所述控制指令对应的功能服务；The front-end module executes the functional service corresponding to the control instruction in response to the user's control instruction;
所述后台模块基于所述功能服务进行后台数据处理；The background module performs background data processing based on the functional service;
所述模型开发单元执行模型开发服务，并获得模型快照；The model development unit executes a model development service and obtains a model snapshot;
所述模型投产单元获取目标模型快照，对所述目标模型快照进行模型投产的操作；The model production unit acquires a target model snapshot, and performs a model production operation on the target model snapshot;
所述模型市场单元在投产成功的模型中确定待展示模型，对所述待展示模型进行模型展示的操作；The model market unit determines the model to be displayed among the models that have been successfully put into production, and performs a model display operation on the model to be displayed;
所述模型计算单元对模型的训练、评估以及预测进行计算，并输出结果文件；The model calculation unit calculates the training, evaluation and prediction of the model, and outputs the result file;
所述数据展示单元对所述结果文件中的指标数据进行展示的操作。The data display unit displays the indicator data in the result file.
上述机器学习模型展示与管理方法可参照本发明机器学习模型展示与管理平台系统的各个模块与单元，此处不再赘述。For the above machine learning model display and management method, reference may be made to each module and unit of the machine learning model display and management platform system of the present invention, which will not be repeated here.
此外，本发明还提出一种电子设备，所述电子设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的机器学习模型展示与管理程序，所述机器学习模型展示与管理程序被所述处理器执行时实现如上所述的机器学习模型展示与管理方法的步骤。In addition, the present invention also proposes an electronic device, which includes: a memory, a processor, and a machine learning model display and management program stored on the memory and operable on the processor, the machine learning When the model display and management program is executed by the processor, the steps of the above machine learning model display and management method are realized.
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者系统不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者系统所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者系统中还存在另外的相同要素。It should be noted that, as used herein, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or system comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or system. Without further limitations, an element defined by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system comprising that element.
上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。The serial numbers of the above embodiments of the present invention are for description only, and do not represent the advantages and disadvantages of the embodiments.
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本发明各个实施例所述的方法。Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product in essence or in other words, the part that contributes to the prior art, and the computer software product is stored in a storage medium (such as ROM/RAM) as described above. , magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present invention.
以上仅为本发明的优选实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process conversion made by using the description of the present invention and the contents of the accompanying drawings, or directly or indirectly used in other related technical fields , are all included in the scope of patent protection of the present invention in the same way.
Claims (10)
Translated fromChinesePriority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211701170.3ACN115829060A (en) | 2022-12-22 | 2022-12-22 | Machine learning model display and management platform system, method and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211701170.3ACN115829060A (en) | 2022-12-22 | 2022-12-22 | Machine learning model display and management platform system, method and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115829060Atrue CN115829060A (en) | 2023-03-21 |
Family
ID=85519074
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211701170.3APendingCN115829060A (en) | 2022-12-22 | 2022-12-22 | Machine learning model display and management platform system, method and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115829060A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118153713A (en)* | 2024-03-21 | 2024-06-07 | 深圳数阔信息技术有限公司 | A management method, system, device and storage medium for machine learning models |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104360846A (en)* | 2014-10-27 | 2015-02-18 | 江西博微新技术有限公司 | OSGi-based adaptive design method for authority management of Web application development platform based on OSGi |
| CN111309378A (en)* | 2020-02-25 | 2020-06-19 | 电子科技大学 | Machine learning model life cycle management system and method |
| CN112256537A (en)* | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model running state display method and device, computer equipment and storage medium |
| CN115438113A (en)* | 2022-09-16 | 2022-12-06 | 中国建设银行股份有限公司 | Parameter management method and device, electronic equipment and storage medium |
- 2022
- 2022-12-22CNCN202211701170.3Apatent/CN115829060A/enactivePending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104360846A (en)* | 2014-10-27 | 2015-02-18 | 江西博微新技术有限公司 | OSGi-based adaptive design method for authority management of Web application development platform based on OSGi |
| CN111309378A (en)* | 2020-02-25 | 2020-06-19 | 电子科技大学 | Machine learning model life cycle management system and method |
| CN112256537A (en)* | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model running state display method and device, computer equipment and storage medium |
| CN115438113A (en)* | 2022-09-16 | 2022-12-06 | 中国建设银行股份有限公司 | Parameter management method and device, electronic equipment and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118153713A (en)* | 2024-03-21 | 2024-06-07 | 深圳数阔信息技术有限公司 | A management method, system, device and storage medium for machine learning models |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11934301B2 (en) | System and method for automated software testing | |
| US20240037020A1 (en) | System and Method for Automated Software Testing | |
| EP3757779B1 (en) | Application assessment system to achieve interface design consistency across micro services | |
| CN108170611B (en) | Automated testing method and device, storage medium, electronic equipment | |
| WO2021184725A1 (en) | User interface test method and apparatus, storage medium, and computer device | |
| US10515001B2 (en) | System and method for automating testing of software applications | |
| US10942840B2 (en) | System and method for managing a code repository | |
| CN109388376B (en) | Software development risk assessment method, device, equipment and readable storage medium | |
| CN110941467A (en) | Data processing method, device and system | |
| US9092572B2 (en) | Development life cycle management tool for set-top box widgets | |
| US12314689B2 (en) | Application development involving instant prototyping | |
| WO2020180734A1 (en) | Systems and methods that facilitate hiring and recruitment | |
| CN113505895A (en) | Machine learning engine service system, model training method and configuration method | |
| WO2021023171A1 (en) | Method and apparatus for test script recording time selection | |
| CN113296912A (en) | Task processing method, device and system, storage medium and electronic equipment | |
| CN115829060A (en) | Machine learning model display and management platform system, method and electronic equipment | |
| CN111008140B (en) | Cross-platform UI (user interface) automatic testing method and device | |
| CN112540747A (en) | Platform development method, device and computer readable storage medium | |
| TW202422418A (en) | Method and apparatus for managing an artificial intelligence model related service | |
| CN114663568B (en) | Material conversion method, device, computer equipment and medium | |
| CN111444170B (en) | Automatic machine learning method and equipment based on predictive business scene | |
| US20250258812A1 (en) | Rapid interactive iteration for prompt design | |
| CN120386717A (en) | Test case generation method, device, storage medium and program product | |
| CN118502724A (en) | BPM-based low code arrangement method, equipment, medium and product | |
| CN115543807A (en) | Automatic regression testing method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |