Model training method and device, equipment, medium and vehicle
The model training method and device enhance efficiency by providing a pre-configured development environment with multiple frameworks and dynamic resource allocation, simplifying user operations and optimizing resource utilization in model training processes.
Patent Information
- Application Number
- CN202310746391.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-25
- Publication Date
- 2024-12-27
AI Technical Summary
Existing model training processes based on AI algorithms are inefficient due to the need for manual tasks such as data management, environment configuration, model construction, and training framework setup, which leads to idle computing resources and reduced training efficiency.
A model training method and device that provides a pre-configured development environment with multiple training frameworks, allowing users to develop and train models without setting up environments, and utilizes reserved and elastic computing resources for efficient resource utilization.
Simplifies user operations, improves model development and training efficiency, and optimizes resource utilization through a one-stop service platform for algorithm engineers.
Smart Images
Figure CN119201657A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of computer technology, and in particular, relates to a model training method, device, equipment, medium and vehicle. Background Art
[0002] With the development of science and technology, artificial intelligence technology has been widely used in people's lives. Among them, the implementation basis of artificial intelligence technology is artificial intelligence (AI) algorithm.
[0003] At present, in order to perform model training based on AI algorithms, algorithm engineers usually need to perform many tasks such as data management, data processing, environment configuration, model construction, model training, model management, model testing, model reasoning, etc., which seriously affects the training efficiency of the model. Summary of the invention
[0004] The purpose of the embodiments of the present application is to provide a model training method, device, equipment, medium and vehicle that can improve the training efficiency of the model.
[0005] In a first aspect, an embodiment of the present application provides a model training method, the method comprising:
[0006] receiving a model development operation performed by a user in a target development space, wherein a model development environment is configured in the target development space;
[0007] In response to the model development operation, acquiring a target algorithm model developed in the model development environment;
[0008] Receiving a training request for the target algorithm model;
[0009] In response to the training request, the target algorithm model is trained using the training data under a pre-set plurality of training frameworks to obtain a trained target algorithm model, wherein the training data is data uploaded by the user.
[0010] In a possible implementation, the training of the target algorithm model using the training data under a plurality of pre-set training frameworks to obtain the trained target algorithm model includes:
[0011] Calling a training component, wherein the training component is pre-set with multiple deep learning frameworks;
[0012] Through the training component, the target algorithm model is trained using the training data under the multiple deep learning frameworks to obtain a trained target algorithm model.
[0013] In a possible implementation, the data uploaded by the user also includes test data;
[0014] After obtaining the trained target algorithm model, the method further includes:
[0015] Obtaining a test component for the target algorithm model;
[0016] Invoke the test component;
[0017] By means of the test component, the trained target algorithm model is evaluated using the test data to obtain an evaluation result;
[0018] The evaluation results are displayed.
[0019] In a possible implementation, before receiving the model development operation performed by the user in the target development space, the method further includes:
[0020] Determine the reserved computing resources allocated by the management user to the user;
[0021] According to the reserved computing resources, a target development space matching the reserved computing resources is allocated to the user.
[0022] In a possible implementation, after receiving the model development operation performed by the user in the target development space, the method further includes:
[0023] Determining target computing resources required to run the target algorithm model;
[0024] When the target computing resource is greater than the reserved computing resource, obtaining elastic computing resources;
[0025] Receiving an operation instruction for the target algorithm model;
[0026] In response to the running instruction, the target algorithm model is run using the reserved computing resources and the elastic computing resources.
[0027] In a possible implementation manner, after allocating a target development space matching the reserved computing resources to the user, the method further includes:
[0028] Receive a call request for a target image;
[0029] In response to the call request, a model operating environment is configured according to the target image, and the model operating environment includes at least one of the model development environment, the model training environment, and the model deployment environment.
[0030] In a possible implementation, after obtaining the trained target algorithm model, the method further includes:
[0031] Receiving a download request for the target algorithm model triggered by a user;
[0032] In response to the download request, a file corresponding to the target algorithm model is sent to a terminal device corresponding to the user.
[0033] In a second aspect, an embodiment of the present application provides a model training device, the device comprising:
[0034] A first receiving module is used to receive a model development operation performed by a user in a target development space, wherein a model development environment is configured in the target development space;
[0035] A first acquisition module, configured to acquire a target algorithm model developed in the model development environment in response to the model development operation;
[0036] A second receiving module is used to receive a training request for the target algorithm model;
[0037] The training module is used to respond to the training request and train the target algorithm model using training data under a pre-set multiple training frameworks to obtain a trained target algorithm model, wherein the training data is data uploaded by the user.
[0038] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising: a processor and a memory storing computer program instructions;
[0039] When the processor executes the computer program instructions, it implements the steps of the model training method in any one of the embodiments of the first aspect.
[0040] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the model training method provided by any one of the implementations in the first aspect is implemented.
[0041] In a fifth aspect, an embodiment of the present application provides a vehicle, the vehicle at least comprising:
[0042] As the model training device of the second aspect;
[0043] Such as the electronic device of the third aspect;
[0044] A computer-readable storage medium as in the fourth aspect.
[0045] In the model training method, device, equipment, medium and vehicle in the embodiment of the present application, since the target development space is configured with a model development environment, by receiving the development operation performed by the user in the target development space, the user does not need to deploy the model development environment by himself, and can directly carry out model development, thereby simplifying user operations and improving the development efficiency of the model. Since a variety of training frameworks have been pre-set, therefore, by responding to the model development operation, the target algorithm model developed in the model development environment is obtained, a training request for the target algorithm model is received, and in response to the training request, the target algorithm model is trained using training data under a variety of pre-set training frameworks, so that the user does not need to build a training framework by himself, and can directly train the target algorithm model, thereby simplifying user operations and improving the training efficiency of the model. In this way, through the embodiment of the present application, the training efficiency of the model can be improved from the whole process of model development and model training. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0047] Figure 1 It is a flowchart of a model training method provided in an embodiment of the present application;
[0048] Figure 2 It is a flowchart of another model training method provided in an embodiment of the present application;
[0049] Figure 3 It is a structural schematic diagram of a model training device provided in an embodiment of the present application;
[0050] Figure 4 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0051] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.
[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.
[0053] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprising a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0054] At present, in order to perform model training based on AI algorithms, algorithm engineers usually need to perform many tasks such as data management, data processing, environment configuration, model construction, model training, model management, model testing, model reasoning, etc., which seriously affects the training efficiency of the model.
[0055] In addition, traditional AI algorithms are usually trained on a fixed single machine or multiple machines, causing computing resources to be idle for a long time and not fully utilized.
[0056] Thus, in order to solve the problems of the prior art, the embodiments of the present application provide a model training method, device, equipment, medium and vehicle.
[0057] The model training method provided in the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.
[0058] Figure 1 It is a flow chart of a model training method provided in an embodiment of the present application. The model training method can be applied to an artificial intelligence platform (Platform of Artificial Intelligence, PAI) and executed by the artificial intelligence platform. The artificial intelligence platform can be a one-stop service platform built for algorithm engineers. The artificial intelligence platform can provide full-process support such as data preprocessing, data management, model construction, model training, model management, model testing, and model deployment.
[0059] like Figure 1 As shown, the model training method may specifically include the following steps:
[0060] S110, receiving a model development operation performed by a user in a target development space, where a model development environment is configured;
[0061] S120, in response to the model development operation, obtaining a target algorithm model developed in the model development environment;
[0062] S130, receiving a training request for a target algorithm model;
[0063] S140. In response to the training request, the target algorithm model is trained using the training data under a plurality of pre-set training frameworks to obtain a trained target algorithm model, wherein the training data is data uploaded by the user.
[0064] Thus, since the target development space is configured with a model development environment, by receiving the development operations performed by the user in the target development space, the user does not need to deploy the model development environment by himself, and can directly carry out model development, thereby simplifying user operations and improving the development efficiency of the model. Since a variety of training frameworks have been pre-set, therefore, by responding to the model development operation, obtaining the target algorithm model developed in the model development environment, receiving a training request for the target algorithm model, and responding to the training request, the target algorithm model is trained using training data under a variety of pre-set training frameworks, so that the user does not need to build a training framework by himself, and can directly train the target algorithm model, thereby simplifying user operations and improving the training efficiency of the model. In this way, through the embodiment of the present application, the training efficiency of the model can be improved from the whole process of model development and model training.
[0065] The specific implementation methods of the above steps are introduced below.
[0066] In some embodiments, in S110, the target development space may be a development space corresponding to the user. That is, the user may develop a model in the target development space corresponding to the user. The user may monopolize a container environment in the development space without being disturbed by others. Among them, the development space may provide a cloud development environment, including computing resources, cloud storage, preset images, etc. Multiple training frames may be pre-set in the development space. Based on this, the model development environment may include computing resources, cloud storage, preset images, training frameworks, etc.
[0067] As an example, the user can develop and debug the code in the target development space. The code that the user debugs in the target development space can be the code that the user develops (i.e., writes) in the target development space, or it can be the code that the user links to the target development space from the local integrated development environment (IDE) through configuration. Specifically, the user can run and debug the code linked from the local IDE in the target development space.
[0068] In some implementations, in S120, the target algorithm model may include a model imported from a local computer and a model developed by a user in a target development environment. The model may be a debugged model or an undebugged model, which is not limited here. The model may exist in the form of code.
[0069] In some embodiments, in S130, the training request may be used to request training of the target algorithm model. After obtaining the target algorithm model, the target algorithm model may be trained to obtain the trained target algorithm model. Before training the target algorithm model, a training request for the target algorithm model may be received first.
[0070] In some embodiments, in S140, the artificial intelligence platform may receive data uploaded by the user. The data may be all data used by the user in development, training, and the like. The data may specifically include training data and test data. The training data may be used to train the model, and the test data may be used to test the model.
[0071] As an example, after receiving the data uploaded by the user, the artificial intelligence platform can save the data uploaded by the user to provide a data basis for subsequent model development and model training. Specifically, the artificial intelligence platform may include a data management module. The data management module may provide two data storage methods: data volume and data set. Among them, the data stored in the data volume may be data that needs to be changed and processed. The data stored in the data set may be fixed data that does not need to be changed or processed. That is, if the data is stored in the data volume, the data can be processed. If the data is stored in the data set, the data may not be processed.
[0072] Specifically, when users use the development space for model development, if they need to process file-type data such as data sets and models, they can use data volumes for file management, and can directly hang file-type data on the development space for use. In addition, data sets can be used to define solidified training data. Data sets can support data storage in multiple formats. Multiple formats may include, for example, a small amount of file-type data, a large amount of display (manifest) data, and structured data. Among them, a small amount of file-type data can be data uploaded locally, or it can be data uploaded by a password system (perfect forwardsecrecy, pfs). Structured data can specifically be csv structured data.
[0073] In this way, the data management module can provide full-process data preparation and data processing capabilities, allowing users to store data in data volumes or data sets for data preparation during model development, model training and other stages.
[0074] In addition, the artificial intelligence platform may include a model development module, which may include a development space and a training component. A variety of training frameworks may be pre-set in the training component.
[0075] As an example, after receiving the training request, the model development module can respond to the training request and train the target algorithm model using the training data under a plurality of pre-set training frameworks to obtain a trained target algorithm model. The plurality of training frameworks may include a training framework in the target development space and a training framework in the training component. Specifically, the user may select one or more from the plurality of training frameworks so that the model development module can perform model training based on the training framework selected by the user.
[0076] Based on this, in order to improve model training efficiency and resource utilization, in some embodiments, the above S140 may specifically include:
[0077] Call the training component, which has multiple deep learning frameworks pre-set;
[0078] Through the training component, the target algorithm model is trained using the training data under various deep learning frameworks to obtain the trained target algorithm model.
[0079] Here, a variety of deep learning frameworks can be pre-set in the training component. A variety of deep learning frameworks may include, for example, the TensorFlow framework and the PyTorch framework. Since the training component includes a variety of deep learning frameworks, the training component can provide single-machine and multi-machine multi-card training methods. In addition, the training component can also support a variety of computing power models such as graphics processing units (GPUs) and central processing units (CPUs). Specifically, the training component can perform model training via the web and / or software development kit (SDK).
[0080] In this way, training the target algorithm model through training components can improve the model training efficiency and improve resource utilization.
[0081] After obtaining the trained target algorithm model, the target algorithm model can be evaluated to determine the performance of the target algorithm model. Based on this, the data uploaded by the user can also include test data. In order to determine the performance of the target algorithm model, in some embodiments, after the above S140, the following steps can also be included:
[0082] Obtain test components for the target algorithm model;
[0083] Call the test component;
[0084] Through the test component, the trained target algorithm model is evaluated using the test data to obtain the evaluation results;
[0085] Displays the evaluation results.
[0086] Here, the artificial intelligence platform may include a model evaluation module. Users can create test components in the model evaluation module. User creation of test components may specifically include user importing test code and user manually writing test code. After the user imports the test code or completes writing the test code, the evaluation module can obtain the test component. After obtaining the test component, the evaluation module can call the test component and use the test data to evaluate the trained target algorithm model to obtain the evaluation result.
[0087] As an example, the test component may include a display method of the evaluation result. Therefore, displaying the evaluation result may specifically include displaying the evaluation result according to the display method in the test component.
[0088] In this way, by evaluating the target algorithm model, the performance of the target algorithm model can be determined.
[0089] In addition, the model evaluation module can evaluate multiple algorithm models at the same time. That is, the model evaluation module can run multiple test components at the same time.
[0090] Based on this, in order to improve the user experience of the artificial intelligence platform, in some embodiments, after the above S140, the following steps may also be included:
[0091] Receiving a download request for a target algorithm model triggered by a user;
[0092] In response to the download request, the file corresponding to the target algorithm model is sent to the terminal device corresponding to the user.
[0093] Here, the artificial intelligence platform can include a model management module, which can provide users with unified model asset storage and management. Specifically, the model management module can be used to perform operations such as model upload, model download, model version management, model detail viewing, model comparison, and model evaluation.
[0094] As an example, the model management module may include a visualization interface, and the user may trigger a download request for the target algorithm model in the visualization interface. After receiving the download request, the model management module may respond to the download request and send a file corresponding to the target algorithm model to the terminal device corresponding to the user.
[0095] In this way, by sending the file corresponding to the target algorithm model to the user's corresponding terminal device, the user can save the target algorithm model locally and further deploy the target algorithm model, thereby improving the user's experience of the artificial intelligence platform.
[0096] In order to improve resource utilization, as another implementation method of the present application, the present application also provides another implementation method of the model training method, please refer to the following embodiments for details.
[0097] See also Figure 2 Another implementation of the model training method provided in the present application includes the following steps: S210-S260, which are explained in detail below.
[0098] S210: Determine the reserved computing resources allocated by the management user to the user.
[0099] Here, the artificial intelligence platform may include reserved computing resources and elastic computing resources. Among them, the reserved computing resources may be exclusive resources. That is, after the reserved computing resources are allocated to the user, even if the user does not use the reserved computing resources. The reserved computing resources still belong to the user and will not incur any fees. If the user uses the elastic computing resources, the elastic computing resources can be charged.
[0100] As an example, the artificial intelligence platform may include a space management module. The management user may allocate corresponding reserved computing resources to each user through the space management module.
[0101] In addition, management users can also view the basic information, resource quotas, member information, storage information, etc. of the space through the space management module. Specifically, management users can view resource statistics, freely filter time to view the average GPU utilization, average CPU utilization, average random access memory (RAM) utilization, etc. of the space, view the tasks running in the workspace (including task type and task creator), filter and view resource usage under current conditions, etc.
[0102] Administrators can also manage members through the space management module. Specifically, they can view members who have joined the space, as well as add and remove members.
[0103] Management users can also manage storage through the space management module. Specifically, the artificial intelligence platform can refine storage according to different user needs to meet the different users' needs for shared storage space, mainly including storage occupancy, shared storage and user storage occupancy. Among them, the storage occupancy can be to display the storage occupancy of the workspace, including storage quota and data volume quota. The storage quota displays the allocated storage and total storage. The allocated storage refers to the storage allocated to shared storage, the storage occupied when creating data volumes and personal initialization in the space. The total storage refers to the storage quota allocated when creating a workspace. The data volume quota displays the occupied data volumes and the total number of data volumes. Shared storage can be used to store code, models, data files, etc. shared in the space. The storage capacity is set and it will be automatically mounted when the development space is started. User storage occupancy can be to display the storage and data volume occupancy of each user in the space.
[0104] S220. According to the reserved computing resources, a target development space matching the reserved computing resources is allocated to the user.
[0105] After the reserved computing resources corresponding to the user are determined, a target development space matching the reserved computing resources can be allocated to the user.
[0106] S230: receiving a model development operation performed by a user in a target development space, wherein a model development environment is configured in the target development space.
[0107] Based on this, in order to improve the operating efficiency of the target algorithm model, in some embodiments, after the above S230, the following may also be included:
[0108] Determine the target computing resources required to run the target algorithm model;
[0109] When the target computing resources are greater than the reserved computing resources, elastic computing resources are obtained;
[0110] Receive operation instructions for the target algorithm model;
[0111] In response to the run instruction, the target algorithm model is run using the reserved computing resources and the elastic computing resources.
[0112] Here, the target computing resources may be resources required to run the target algorithm model. The target computing resources may be less than, equal to, or greater than the reserved computing resources. When the target computing resources are not greater than the reserved computing resources, the reserved computing resources may be directly used to run the target algorithm model. Running the target algorithm model may include debugging and training the target algorithm model. In addition, when the target computing resources are greater than the reserved computing resources, elastic computing resources may be obtained.
[0113] In this way, by utilizing reserved computing resources and elastic computing resources to run the target algorithm model, the operating efficiency of the target algorithm model can be improved.
[0114] S240. In response to the model development operation, obtain a target algorithm model developed in the model development environment.
[0115] S250: Receive a training request for a target algorithm model.
[0116] S260. In response to the training request, the target algorithm model is trained using the training data under a plurality of pre-set training frameworks to obtain a trained target algorithm model, wherein the data includes the training data.
[0117] In addition, in the embodiment of the present application, S230 to S260 are the same as S110 to S140 in the above embodiment, and for the sake of brevity, they are not described in detail here.
[0118] Therefore, by allocating the corresponding reserved computing resources to each user, resources can be scheduled uniformly, thereby improving resource utilization.
[0119] Based on this, in order to further improve the user experience of the artificial intelligence platform, in some embodiments, after the above S220, the following steps may also be included:
[0120] Receive a call request for a target image;
[0121] In response to the call request, the model operating environment is configured according to the target image, and the model operating environment includes at least one of a model development environment, a model training environment, and a model deployment environment.
[0122] Here, the target image can encapsulate the software dependencies required for the program to run. The target image can provide a running environment for model development, model training, and model deployment.
[0123] As an example, the artificial intelligence platform may include an image management module. The image management module may receive a call request for a target image, and in response to the call request, configure the model operating environment according to the target image. Specifically, configuring the model operating environment according to the target image may be initializing the model operating environment.
[0124] In this way, by configuring the model operating environment according to the target image, it can ensure the effectiveness of the model operating environment even if problems occur in the model operating environment, further improving the user experience of the artificial intelligence platform.
[0125] In addition, users can manage images through the image management module. Specifically, users can use the image management module to manage versions of images, view image details, and build custom images with different types of images as the base image. Users can also import custom development space images through the image management module and select their own uploaded images when creating a development environment. Based on this, users can use multiple types of images through the image management module, and create various workloads by selecting preset images or custom images on the console interface, or filling in the Uniform Resource Locator (URL) of the public network image.
[0126] Therefore, through the embodiment of the present application, a one-stop service platform (i.e., artificial intelligence platform) and unified resource scheduling can be provided to algorithm engineers. Among them, the one-stop service platform greatly improves development efficiency, and unified resource scheduling significantly improves resource utilization efficiency.
[0127] It should be noted that the application scenarios described in the above-mentioned embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.
[0128] Based on the same inventive concept, the present application also provides a model training device. Figure 3 Provide detailed explanation.
[0129] Figure 3 It is a structural schematic diagram of a model training device provided by an embodiment of the present application.
[0130] like Figure 3 As shown, the model training device 300 may include:
[0131] A first receiving module 310 is used to receive a model development operation performed by a user in a target development space, where a model development environment is configured;
[0132] A first acquisition module 320, for acquiring a target algorithm model developed in a model development environment in response to a model development operation;
[0133] A second receiving module 330 is used to receive a training request for a target algorithm model;
[0134] The training module 340 is used to respond to the training request and train the target algorithm model using the training data under a plurality of pre-set training frameworks to obtain the trained target algorithm model, wherein the training data is the data uploaded by the user.
[0135] The model training device 300 is described in detail below, as shown below:
[0136] In some embodiments, the training module 340 may specifically include:
[0137] The calling submodule is used to call the training component, in which a variety of deep learning frameworks are pre-set;
[0138] The training submodule is used to train the target algorithm model using training data under various deep learning frameworks through training components to obtain the trained target algorithm model.
[0139] In some embodiments, the data uploaded by the user may also include test data. Based on this, the model training device 300 may also include:
[0140] A second acquisition module is used to acquire a test component for the target algorithm model after obtaining the trained target algorithm model;
[0141] Calling module, used to call the test component;
[0142] The evaluation module is used to evaluate the trained target algorithm model using test data through the test component to obtain the evaluation results;
[0143] The display module is used to display the evaluation results.
[0144] In some embodiments, the model training device 300 may further include:
[0145] A first determination module is used to determine the reserved computing resources allocated to the user by the management user before receiving the model development operation performed by the user in the target development space;
[0146] The allocation module is used to allocate a target development space matching the reserved computing resources to the user according to the reserved computing resources.
[0147] In some embodiments, the model training device 300 may further include:
[0148] A second determination module is used to determine the target computing resources required to run the target algorithm model after receiving the model development operation performed by the user in the target development space;
[0149] A third acquisition module is used to acquire elastic computing resources when the target computing resources are greater than the reserved computing resources;
[0150] A third receiving module is used to receive an operation instruction for a target algorithm model;
[0151] The running module is used to respond to the running instruction and use the reserved computing resources and elastic computing resources to run the target algorithm model.
[0152] In some embodiments, the model training device 300 may further include:
[0153] A fourth receiving module, used to receive a call request for a target image;
[0154] The configuration module is used to configure the model operating environment according to the target image in response to the call request, and the model operating environment includes at least one of the model development environment, the model training environment and the model deployment environment.
[0155] In some embodiments, the model training device 300 may further include:
[0156] A fifth receiving module, used for receiving a download request for the target algorithm model triggered by a user after obtaining the trained target algorithm model;
[0157] The sending module is used to send the file corresponding to the target algorithm model to the terminal device corresponding to the user in response to the download request.
[0158] Thus, since the target development space is configured with a model development environment, by receiving the development operations performed by the user in the target development space, the user does not need to deploy the model development environment by himself, and can directly carry out model development, thereby simplifying user operations and improving the development efficiency of the model. Since a variety of training frameworks have been pre-set, therefore, by responding to the model development operation, obtaining the target algorithm model developed in the model development environment, receiving a training request for the target algorithm model, and responding to the training request, the target algorithm model is trained using training data under a variety of pre-set training frameworks, so that the user does not need to build a training framework by himself, and can directly train the target algorithm model, thereby simplifying user operations and improving the training efficiency of the model. In this way, through the embodiment of the present application, the training efficiency of the model can be improved from the whole process of model development and model training.
[0159] Figure 4 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application.
[0160] The electronic device 400 may include a processor 401 and a memory 402 storing computer program instructions.
[0161] Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
[0162] Memory 402 may include a large capacity memory for data or instructions. By way of example and not limitation, memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, memory 402 may include a removable or non-removable (or fixed) medium. In appropriate cases, memory 402 may be inside or outside of an integrated gateway disaster recovery device. In a specific embodiment, memory 402 is a non-volatile solid-state memory.
[0163] In certain embodiments, the memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of the present application.
[0164] The processor 401 implements any one of the model training methods in the above embodiments by reading and executing computer program instructions stored in the memory 402.
[0165] In some examples, the electronic device 400 may further include a communication interface 403 and a bus 410. Figure 4 As shown, the processor 401, the memory 402, and the communication interface 403 are connected via a bus 410 and communicate with each other.
[0166] The communication interface 403 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.
[0167] Bus 410 includes hardware, software or both, and the components of online data flow billing equipment are coupled to each other. For example, but not limitation, bus 410 may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-end bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 410 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.
[0168] Exemplarily, the electronic device 400 may be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA).
[0169] The electronic device 400 can execute the model training method in the embodiment of the present application, thereby realizing the combination Figures 1 to 3 Described model training method and device.
[0170] In addition, in combination with the model training method in the above embodiment, the embodiment of the present application may provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the model training methods in the above embodiment is implemented. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as portable disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, etc.
[0171] An embodiment of the present application also provides a computer program product. When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the model training method provided in the embodiment of the present application.
[0172] A vehicle is also provided in an embodiment of the present application, which may include at least one of a power battery charging device, an electronic device, and a computer storage medium. The specific model training device, electronic device, and computer storage medium involved are consistent with the above description and will not be repeated here.
[0173] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.
[0174] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.
[0175] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.
[0176] The above reference is according to the method of the embodiment of the present application, the flow chart of the device (system) and the computer program product and / or the block diagram described various aspects of the present application.It should be understood that each square box in the flow chart and / or the block diagram and the combination of each square box in the flow chart and / or the block diagram can be realized by computer program instructions.These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the realization of the function / action specified in one or more square boxes of the flow chart and / or the block diagram.Such a processor can be but is not limited to a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit.It can also be understood that each square box in the block diagram and / or the flow chart and the combination of the square boxes in the block diagram and / or the flow chart can also be realized by the dedicated hardware that performs the specified function or action, or can be realized by the combination of dedicated hardware and computer instructions.
[0177] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.
Claims
1. A model training method, characterized in that: include: receiving a model development operation performed by a user in a target development space, wherein a model development environment is configured in the target development space; In response to the model development operation, acquiring a target algorithm model developed in the model development environment; Receiving a training request for the target algorithm model; In response to the training request, the target algorithm model is trained using the training data under a pre-set plurality of training frameworks to obtain a trained target algorithm model, wherein the training data is data uploaded by the user.
2. The method according to claim 1, characterized in that The method of training the target algorithm model using the training data under a plurality of pre-set training frameworks to obtain the trained target algorithm model includes: Calling a training component, wherein the training component is pre-set with multiple deep learning frameworks; Through the training component, the target algorithm model is trained using the training data under the multiple deep learning frameworks to obtain a trained target algorithm model.
3. The method according to claim 1, characterized in that The data uploaded by the user also includes test data; After obtaining the trained target algorithm model, the method further includes: Obtaining a test component for the target algorithm model; Invoke the test component; By means of the test component, the trained target algorithm model is evaluated using the test data to obtain an evaluation result; The evaluation results are displayed.
4. The method according to claim 1, characterized in that Before receiving the model development operation performed by the user in the target development space, the method further includes: Determine the reserved computing resources allocated by the management user to the user; According to the reserved computing resources, a target development space matching the reserved computing resources is allocated to the user.
5. The method according to claim 4, characterized in that After receiving the model development operation performed by the user in the target development space, the method further includes: Determining target computing resources required to run the target algorithm model; When the target computing resource is greater than the reserved computing resource, obtaining elastic computing resources; Receiving an operation instruction for the target algorithm model; In response to the running instruction, the target algorithm model is run using the reserved computing resources and the elastic computing resources.
6. The method according to claim 4, characterized in that After allocating a target development space matching the reserved computing resources to the user, the method further includes: Receive a call request for a target image; In response to the call request, a model operating environment is configured according to the target image, and the model operating environment includes at least one of the model development environment, the model training environment, and the model deployment environment.
7. The method according to claim 1, characterized in that After obtaining the trained target algorithm model, the method further includes: Receiving a download request for the target algorithm model triggered by a user; In response to the download request, a file corresponding to the target algorithm model is sent to a terminal device corresponding to the user.
8. A model training device, characterized in that: include: A first receiving module is used to receive a model development operation performed by a user in a target development space, wherein a model development environment is configured in the target development space; A first acquisition module, configured to acquire a target algorithm model developed in the model development environment in response to the model development operation; A second receiving module is used to receive a training request for the target algorithm model; The training module is used to respond to the training request and train the target algorithm model using training data under a pre-set multiple training frameworks to obtain a trained target algorithm model, wherein the training data is data uploaded by the user.
9. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the steps of the model training method as described in any one of claims 1-7 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the model training method according to any one of claims 1 to 7.
11. A vehicle, characterized in that: Include at least one of the following: The model training device as claimed in claim 8; The electronic device as claimed in claim 9; The computer readable storage medium of claim 10.