Visualization method and related equipment

Abstract

The present application discloses a visualization method and related equipment, the method comprising: a user device sends a visualization request to a cloud server, the visualization request is used to request visualization of target training data of a target deep learning model; the user device receives a visualization response sent by the cloud server, the visualization response carries visualization information of the target training data; the user device displays the visualization information. The use of the embodiments of the present application can realize the visualization of the training data of the deep learning model, which helps to timely judge the feasibility of the current training strategy and provide a basis for decisions such as early stopping.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a visualization method and related equipment. Background Art

[0002] In recent years, deep learning models have been widely used in the field of image and video processing. Generally, deep learning models are obtained through training. However, the training process of deep learning models involves complex and profound calculations and requires multiple iterations. The training time is long, and R&D personnel can only evaluate the performance of the trained deep learning models after the training is completed. Summary of the invention

[0003] The embodiments of the present application provide a visualization method and related devices for realizing visualization of training data of a deep learning model.

[0004] In a first aspect, an embodiment of the present application provides a visualization method, which is applied to a user device, and the method includes:

[0005] Sending a visualization request to a cloud server, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0006] Receiving a visualization response sent by the cloud server, where the visualization response carries visualization information of the target training data;

[0007] The visualization information is displayed.

[0008] In a second aspect, an embodiment of the present application provides a visualization device, applied to a user equipment, the device comprising:

[0009] A sending unit, configured to send a visualization request to a cloud server, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0010] A receiving unit, configured to receive a visualization response sent by the cloud server, wherein the visualization response carries visualization information of the target training data;

[0011] A display unit is used to display the visual information.

[0012] In a third aspect, an embodiment of the present application provides a visualization method, which is applied to a cloud server, and the method includes:

[0013] Receiving a visualization request from a user device, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0014] Obtaining target training data corresponding to the visualization request;

[0015] Preprocessing the target training data to obtain visualization information;

[0016] A visualization response is sent to the user equipment, where the visualization response carries the visualization information.

[0017] In a fourth aspect, an embodiment of the present application provides a visualization device, applied to a cloud server, the device comprising:

[0018] A receiving unit, configured to receive a visualization request from a user device, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0019] An acquisition unit, configured to acquire target training data corresponding to the visualization request;

[0020] A determination unit, used for preprocessing the target training data to obtain visualization information;

[0021] A sending unit is used to send a visualization response to the user equipment, where the visualization response carries the visualization information.

[0022] In a fifth aspect, the present application provides a computer device, comprising a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the program comprises instructions for executing the steps of the method described in the first aspect or the third aspect of the embodiment of the present application.

[0023] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps described in the method described in the first aspect or the third aspect of the embodiment of the present application.

[0024] In a seventh aspect, an embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all of the steps described in the method described in the first aspect or the third aspect of the embodiment of the present application. The computer program product may be a software installation package.

[0025] It can be seen that in the embodiment of the present application, the user device first sends a visualization request to the cloud server, and then receives a visualization response sent by the cloud server. The visualization response carries visualization information of the target training data, and finally displays the visualization information. Through the visualization information of the target training data, the training status of the deep learning model can be intuitively understood, which helps to timely judge the feasibility of the current training strategy and provide a basis for decisions such as early stopping. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 It is a schematic diagram of the architecture of a visualization system provided in an embodiment of the present application;

[0028] Figure 2 It is a flowchart of a visualization method provided in an embodiment of the present application;

[0029] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application;

[0030] Figure 4 A schematic diagram of the structure of a visualization device provided in an embodiment of the present application;

[0031] Figure 5 A schematic diagram of the structure of another visualization device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0033] The following are detailed descriptions of each.

[0034] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present invention and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not limited to the listed steps or modules, but optionally includes steps or modules that are not listed, or optionally includes other steps or modules that are inherent to these processes, methods, products or devices.

[0035] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0036] See also Figure 1 , Figure 1 : is a schematic diagram of the architecture of a visualization system provided in an embodiment of the application, the visualization system includes a supercomputing cluster, a cloud server, a cloud storage and a user device. It should be noted that: Figure 1 The form and quantity of the supercomputing cluster, cloud server, cloud storage and user device shown in are for example only and do not constitute a limitation on the embodiments of the present application. The visualization system can realize the visualization of multi-dimensional and multi-class intermediate training data based on javascript and svg technology, and supports fast and efficient visualization rendering of a large amount of intermediate training data.

[0037] Among them, the visualization system is based on the kubernetes container orchestration system, providing stable production-level visualization system deployment and dynamic and controllable service capabilities. The kubernetes container is used to manage containerized applications on multiple hosts in cloud services.

[0038] Among them, the cloud server provides a visualization service, which is used to perform data operations such as preprocessing on the acquired intermediate training data, and to perform visualization rendering on the intermediate training data after the data operation to obtain visualization information.

[0039] Optionally, the cloud server may also provide at least one of the following services: service gateway, training management service, data storage service and permission service.

[0040] Among them, the service gateway can serve as the entrance to at least one of the following services: training management service, data storage service, data visualization service and permission service.

[0041] Alternatively, the service gateway can be an application.

[0042] Optionally, the service gateway may have a current limiting function.

[0043] Among them, the training management service can provide at least one of the following services: monitoring whether the training tasks of the deep learning model are completed, recording, indexing and searching for training tasks that generate exceptions, comparing the training processes of multiple training tasks of the same deep learning model, and sharing the training results of the deep learning model.

[0044] Among them, the data storage service can provide at least one of the following services: data persistence, using data threads to store intermediate training data generated by deep learning model training, and using cache to store intermediate training data generated by deep learning model training.

[0045] Among them, the permission management service is used to ensure the security of the visualization system. The visualization system can only be used after passing the permission management service. In addition, shared training is also authenticated through the permission management service. The permission management service supports two types of login methods: unified authentication account login and / or ordinary registration.

[0046] Among them, the cloud server can interact with the developer toolkit through the HyperText Transfer Protocol (HTTP), Google Remote Procedure Call (GRPC) protocol or other protocols.

[0047] Among them, the cloud server can directly store the training data in the cloud storage.

[0048] Among them, the supercomputing cluster provides development tools such as toolkits for programming languages ​​such as Python by running developer toolkits, which can be used by users together with tensorflow, pytorch, caffe frameworks, etc., and / or open new threads as dedicated threads for data upload, and upload intermediate training data generated by deep learning model training to the cloud server on the dedicated thread for data upload through HTTP communication protocol and GRPC protocol. The uploaded intermediate training data types may include vectors, scalars, images, videos, audios, etc. The uploaded intermediate training data are stored in the data cache of the developer toolkit.

[0049] Among them, starting a new thread as a dedicated thread for data uploading can ensure the calculation speed of the deep learning model.

[0050] Among them, cloud storage can optionally be distributed storage, which is used to uniformly manage the intermediate training data generated by the deep learning model, provide massive storage functions for the intermediate training data, and as the amount of intermediate training data increases, it can dynamically expand to meet the storage needs of the intermediate training data, providing subsequent guarantees for the growth of visualization business.

[0051] Optionally, the cloud storage may also be block-level cloud storage, file-level cloud storage, object-level cloud storage and / or other forms of cloud storage.

[0052] The user device may communicate with the cloud server through a preset application software, or may communicate with the cloud server through a preset network address.

[0053] Optionally, the user equipment may include various handheld devices with communication functions, vehicle-mounted devices, wearable devices, computing devices or other processing devices connected to a wireless modem, as well as various forms of user equipment (UE), mobile stations (MS), terminal devices, etc.

[0054] See also Figure 2 As shown, Figure 2 : is a flowchart of a visualization method provided in an embodiment of the present application, which is applied to the above-mentioned user device and cloud server, and specifically includes the following steps:

[0055] Step 201: The user device sends a visualization request to the cloud server, where the visualization request is used to request visualization of target training data of a target deep learning model.

[0056] Step 202: The cloud server receives a visualization request from a user device, where the visualization request is used to request visualization of target training data of a target deep learning model.

[0057] In one possible implementation, the visualization request carries the training task identifier of the target deep learning model. After receiving the visualization request, the cloud server can obtain the relevant training data of the training task based on the training task identifier of the target deep learning model. As an example, the relevant training data of the training tasks of each stored deep learning model can be associated with the training task identifier of the deep learning model. For example, the training data can be stored according to the training task, and the relevant training data of different training tasks are stored in different locations. For another example, the training data can be stored according to the deep learning model or data type.

[0058] Among them, the data type can be a two-dimensional heat map, a three-dimensional histogram, a feature map, a call timing diagram, a scalar line graph, a directed acyclic graph, and so on.

[0059] Different data types correspond to different visualization icons.

[0060] Among them, the training tasks are all located on the supercomputing cluster, and the training of deep learning models can be accelerated by utilizing the central processing unit of the supercomputing cluster.

[0061] At this time, optionally, the training data and the task identifier of the training task to which the training data belongs may be associated and stored, etc., which is not limited in the embodiments of the present disclosure.

[0062] In one possible implementation, the visualization request carries the training task identifier of the target deep learning model and the index information of the target training data, and the index information is used to find the target training data to determine the storage location of the target training data. The index information can be implemented in a variety of ways, for example, the index information can include any one or more of the data creation time, data type, data label, etc., and for another example, the index information includes the data type and the data label.

[0063] The same data type can have multiple data labels, and the data labels can be customized.

[0064] For example, if the data type is a scalar line chart, the data label can be the exact value, the loss value, and so on.

[0065] The visualization request may be a request to sample all data in the intermediate data set or randomly sample data in the intermediate data set corresponding to the index information according to the training task identifier and the index information.

[0066] The target deep learning model includes at least one deep neural network, and the target training data can be intermediate data generated by one of the deep neural networks in the target deep learning model, or can be intermediate data generated by multiple deep neural networks in the target deep learning model.

[0067] Optionally, the visualization request may be a request to visualize intermediate data generated by the target deep learning model within a first time period, wherein the duration of the first time period may be other values ​​such as 3 minutes, 5 minutes, 9 minutes and 15 minutes, and the end time of the first moment may be the current moment, or may be earlier than the current moment, or may be later than the current moment.

[0068] Among them, the target training data is stored in cloud storage.

[0069] In one possible implementation, the target training data includes at least one of the following: model performance change trend information, model loss trend information, model parameter distribution information, model processing intermediate results, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information sequence of multiple processes.

[0070] In one possible implementation, when the target training data is at least one of model performance change trend information, model loss trend information, current progress information of model training, and comparison information of different trainings for the same model, the category of visualization information is a scalar line chart.

[0071] Among them, when the target training data is the model performance change trend information, it can be determined based on the scalar line chart whether the performance change trend meets the first preset condition. If the first preset condition is met, the model training is stopped; if the first preset condition is not met, the model training is continued.

[0072] Among them, when the target training data is model loss trend information, it can be determined based on the scalar line graph whether the loss trend meets the second preset condition. If the second preset condition is met, the model training is stopped; if the preset condition is not met, the model training is continued.

[0073] Among them, when the target training data is the current progress information of the model training, it can be determined based on the scalar line chart whether the training progress meets the preset progress. If the preset progress is met, the model training will continue; if the preset conditions are not met, the model training will be stopped.

[0074] Among them, when the target training data is the comparison information of different trainings of the same model, it can be determined based on the scalar line graph whether the features extracted from different trainings are the same. If the features are the same, the model training continues; if the features are different, the model training is stopped.

[0075] Among them, when the target training data is the model parameter distribution information, the category of visualization information is a three-dimensional histogram. The three-dimensional histogram can be used to determine whether the model parameter distribution is abnormal. If the model parameter distribution is abnormal, the model training is stopped.

[0076] Among them, when the target training data is the intermediate result of model processing, the category of visualization information is feature map, through which it can be determined whether the features extracted by the model are correct. If the extracted features are wrong, the model is rebuilt.

[0077] Among them, when the target training data is model structure information, the category of visualization information is directed acyclic graph, through which it can be determined whether the structure of the model is correct. If the structure of the model is wrong, the structure of the model can be reconstructed.

[0078] Among them, the architecture of the deep learning model supported by the directed acyclic graph can be a standard architecture such as OpenNeural Network Exchange (ONNX) or other types of architectures. Among them, ONNX does not require data processing of intermediate training data to obtain a directed acyclic graph, while some architectures require preprocessing of intermediate training data to obtain a directed acyclic graph.

[0079] Among them, when the target training data is the scheduling time information of multiple processes, the category of visualization information is the call timing diagram of multiple processes. The time consumption of each operator can be determined through the call timing diagram. When the time consumption of the operator is greater than the preset duration, the operator whose time consumption is greater than the preset duration is optimized.

[0080] Among them, when the target training data is the intermediate result of model processing, the category of visualization information is a two-dimensional heat map. The heat map can be used to determine whether the features extracted by the model are correct. If the extracted features are wrong, the model can be rebuilt.

[0081] Among them, feature maps and heat maps are different forms of representation of the intermediate results of model processing using target training data.

[0082] The type of target training data includes at least one of the following: scalar, vector, picture, video, and audio.

[0083] Step 203: The cloud server obtains target training data corresponding to the visualization request.

[0084] In a possible implementation manner, the acquiring target training data corresponding to the visualization request includes:

[0085] Searching an index database based on the training task identifier and index information carried in the visualization request to obtain storage location information of the target training data;

[0086] Based on the storage location information of the target training data, the target training data stored in the cloud storage is acquired.

[0087] Wherein, a list of intermediate training data is obtained based on the training task identifier and the index information, and the intermediate training data is obtained in the cloud storage based on the list of intermediate training data.

[0088] Optionally, the visualization request carries identification information of target training data of the target deep learning model, and the identification information is used by the cloud server to obtain the target training data from the cloud storage; wherein the identification information is preset and the identification information is unique in the cloud storage.

[0089] For example, the training tasks of the target deep learning model include training task A and training task B, the training data generated by training task A include training data A1 and training data A2, the training data generated by training task B include training data B1 and training data B2, and the training data A1, training data A2, training data B1 and training data B2 are all stored in the cloud platform. The identification information of training data A1 is 1, the identification information of training data A2 is 2, the identification information of training data B1 is 3, and the identification information of training data B2 is 4. If the identification information is 1, the target training data obtained by the cloud server from the cloud storage is training data A1; if the identification information is 2, the target training data obtained by the cloud server from the cloud storage is training data A2; if the identification information is 3, the target training data obtained by the cloud server from the cloud storage is training data B1; if the identification information is 4, the target training data obtained by the cloud server from the cloud storage is training data B2.

[0090] Step 204: The cloud server pre-processes the target training data to obtain visualization information.

[0091] Step 205: The cloud server sends a visualization response to the user device, where the visualization response carries the visualization information.

[0092] Step 206: The user device receives a visualization response sent by the cloud server, where the visualization response carries visualization information of the target training data.

[0093] Step 207: The user equipment displays the visualization information.

[0094] In a possible implementation, the visualization information includes at least one of the following categories: a directed acyclic graph, a three-dimensional histogram, a call sequence diagram of multiple processes, a feature diagram, a two-dimensional heat map, and a scalar line graph.

[0095] Optionally, when the category of visualization information is a directed acyclic graph, the visualization information includes the following information: topological information of at least a part of the target deep learning model, wherein at least a part of the target deep learning model includes: multiple modules and / or multiple operators of the target deep learning model; resource occupancy information of each operator in at least one operator included in the target deep learning model.

[0096] Optionally, the topology information of the multiple modules includes at least one of the following: identification information of the multiple modules, dependencies between the multiple modules, data size of each module in the multiple modules, and information of at least one operator included in each of the modules.

[0097] The identification information of the module is unique and can be preset.

[0098] Optionally, the operator information includes at least one of the following: identification information of the operator, dependency between operators, and data volume of the operator.

[0099] Among them, the dependency relationship can be a sequential relationship or a parallel relationship.

[0100] Optionally, the resource occupancy information is determined by the cloud server based on at least one of the data type of the operator, the input data information of the operator, and the output data information of the operator.

[0101] The input of the operator can be pictures, videos, audio, scalars, vectors, etc. Different inputs correspond to different dimensions. For example, a picture is equivalent to a two-dimensional matrix, and audio is equivalent to a one-dimensional matrix.

[0102] Among them, after the operation of the operator, the dimension of the operator's input information will change.

[0103] The data type of the operator can be a double-precision type or a single-precision type.

[0104] The operator may be at least one of the following: convolution, batch normalization (BatchNorm, BN), full connection, pooling, matrix multiplication and division, dropout, activation, etc.

[0105] The input data information may be the size of the input data, and the output data information may be the size of the output data.

[0106] Optionally, when the category of visualization information is a scalar line chart, the visualization information includes at least one of target deep model performance change trend information, target deep model loss trend information, target deep learning model training progress information, and comparison information of different trainings in the target deep learning model.

[0107] Optionally, when the category of the visualization information is a feature map or a two-dimensional heat map, the visualization information includes features extracted by the target deep learning model.

[0108] Optionally, when the category of the visualization information is a three-dimensional histogram, the visualization information includes parameter distribution of each operator in the multiple operators.

[0109] Optionally, when the category of the visualization information is a call sequence diagram of multiple processes, the visualization information includes the running time of each operator in the multiple operators in each process in at least one process.

[0110] Optionally, the process may be at least one of an interpreted language process, a local process, and an AI chip process.

[0111] Among them, the interpreted language process, local process, and AI chip process execute the call timing diagram alternately in time.

[0112] It can be seen that in the embodiment of the present application, the user device first sends a visualization request to the cloud server, then receives a visualization response sent by the cloud server, the visualization response carries the visualization information of the target training data, and finally displays the visualization information. Since the visualization information of the target training data can be used to intuitively understand the training status of the deep learning model, it helps to improve the feasibility of the current training strategy in a timely manner and provide a basis for decisions such as early stopping.

[0113] In an implementation of the present application, displaying the visual information includes:

[0114] The visualization information is displayed in a folded display mode or a displayed display mode, wherein:

[0115] In the folded display mode, a module in the target deep learning model is used as a minimum display unit, wherein the module includes at least one operator; and / or

[0116] In the expanded display mode, the operator in the target deep learning model is used as the smallest display unit.

[0117] Among them, when the category of visualization information is a directed acyclic graph, the visualization information can be displayed in a collapsed display mode or in an expanded display mode. When the category of visualization information is not a directed acyclic graph, the visualization information is displayed in an expanded display mode.

[0118] The expanded display mode may be that the module expands and displays the operators in the module after receiving an operation instruction.

[0119] In the case where the category of the visualization information is a three-dimensional histogram, the expanded display mode may be to directly display the parameter distribution of each operator in each module.

[0120] In the case where the category of the visualization information is a call sequence diagram, the expanded display mode may be to directly display the running time of each operator in each module in at least one process.

[0121] For example, assuming that the category of visualization information is a directed acyclic graph, the target deep learning model includes two modules (module A and module B), module A includes three operators (A1, A2, and A3), and model B includes two operators (B1 and B2). If the visualization information is displayed in the collapsed display mode, module A and module B are displayed; if the visualization information is displayed in the expanded display mode, after module A receives the operation instruction, the operators A1, A2, and A3 in module A are displayed, and after module B receives the operation instruction, the operators B1 and B2 in module B are displayed.

[0122] For example, assuming that the category of visualization information is a three-dimensional histogram, the target deep learning model includes two modules (module A and module B), module A includes three operators (A1, A2, and A3), model B includes two operators (B1 and B2), and module A corresponds to a three-dimensional histogram. Figure 1 , module B corresponds to the three-dimensional histogram B, then the three-dimensional histogram Figure 1 It includes the parameter distribution of A1, the parameter distribution of A2 and the parameter distribution of A3, and the three-dimensional histogram Figure 2 It includes the parameter distribution of B1 and the parameter distribution of B2.

[0123] For example, suppose the category of visualization information is a call sequence diagram. The target deep learning model includes two modules (module A and module B) and two processes (processes C1 and C2). Module A includes two operators (A1 and A2), and model B includes two operators (B1 and B2). The corresponding call sequence of module A is Figure 1 , module B corresponds to calling sequence diagram B, then calling sequence Figure 1 The call sequence includes the running time of A1 in process C1 and the running time of A1 in process C2, including the running time of A2 in process C1 and the running time of A2 in process C2; Figure 2 The time period includes the time B1 spends running in process C1 and the time B1 spends running in process C2, and the time B2 spends running in process C1 and the time B2 spends running in process C2.

[0124] Optionally, the display mode of the visualization information may be preset, and the visualization information may be first displayed in a folded display mode, and then displayed in an expanded display mode after a first duration, and the first duration may be preset.

[0125] Optionally, in the case where the category of visualization information is a scalar line chart, the line chart can be multiple lines, which can be scaled or translated after receiving an operation instruction, and the expanded display mode of the scalar line chart can be to directly display at least one of the target deep model performance change trend, the target deep model loss trend, the training progress of the target deep learning model, and the features extracted from different training in the target deep learning model.

[0126] Optionally, when the category of the visualization information is a feature map or a two-dimensional heat map, the expandable display mode can be to directly display the features extracted by the target deep learning model.

[0127] Among them, in the two-dimensional heat map, the higher the heat, the greater the probability of being a feature.

[0128] It can be seen that in the embodiments of the present application, the user equipment displays visual information in different ways, which is conducive to improving the application scope of the user equipment.

[0129] In an implementation of the present application, before sending the visualization request to the cloud server, the method further includes:

[0130] The intermediate training data of the target deep learning model is sent to the cloud server through a data upload thread, and the intermediate training data is used by the cloud server to obtain index information and a training task identifier of the target training data, and store the index information in a database based on the training task identifier, and store the target training data in the cloud storage based on the index information.

[0131] Among them, the data upload thread is a newly opened thread.

[0132] The training task identifier is unique and may be determined after intermediate training data is generated or when training begins.

[0133] The intermediate training data includes target training data, training task identifier and index information.

[0134] Optionally, the intermediate training data of the target deep learning model is sent to the cloud server through the data upload thread, and the intermediate data is used by the cloud server to obtain identification information of the target training data, and store the identification information in a database, and store the target training data in the cloud storage based on the identification information.

[0135] Among them, the identification information is unique.

[0136] Optionally, before sending the intermediate training data of the target deep learning model to the cloud server through the data upload thread, the method further includes:

[0137] Determining the data amount of target training data in the intermediate training data;

[0138] In the case where the data amount is greater than a preset data amount, determining a training task identifier and index information of the target training data, and carrying the training task information and the index information of the target training data on the intermediate training data;

[0139] In a case where the data amount is less than a preset data amount, identification information of the target training data is determined, and the identification information is carried on the intermediate training data.

[0140] Optionally, when the cloud server receives the identification information, the identification information may be stored in a first buffer of a database, and the target training data may be stored in a second buffer of the database based on the identification information.

[0141] It can be seen that in the embodiment of the present application, storing the target data in the cloud storage through the cloud server is beneficial for the cloud server to quickly obtain the target training data after receiving the visualization request.

[0142] In an implementation of the present application, the preprocessing of the target training data to obtain visualization information includes:

[0143] Perform at least one preprocessing on the target training data to obtain preprocessed data;

[0144] The preprocessed data is rendered to obtain the visualization information.

[0145] Among them, if the data type of the target training data is picture or video, the preprocessing of the picture or video includes one or more of the following processing: insensitive area elimination processing, image accuracy enhancement processing, image noise reduction processing and image binarization processing.

[0146] Among them, if the data type of the target training data is text, the text preprocessing includes one or more of the following processes: document segmentation, text segmentation, stop word removal (including punctuation, numbers, single words and other meaningless words), text feature extraction, word frequency statistics and text vectorization.

[0147] The visual rendering is to assemble the pre-processed data into Hypertext Markup Language (HTML).

[0148] Among them, the type of visualization information obtained through preprocessing is not a directed acyclic graph.

[0149] It can be seen that in the embodiment of the present application, the target training data is preprocessed to obtain preprocessed data, and then the preprocessed data is rendered to obtain visualization information, which is conducive to analyzing the target deep learning model based on the visualization information.

[0150] In an implementation of the present application, the preprocessing of the target training data to obtain visualization information includes:

[0151] When the training task of the target deep learning model is executed under the target framework, the target training data is parsed and processed through a binary tree to obtain the visualization information.

[0152] Among them, the type of the target deep learning model that processes the target training data through the binary tree is parrots.

[0153] Among them, the type of visualization information obtained by parsing the target training data through the binary tree is a directed acyclic graph.

[0154] It can be seen that in the embodiment of the present application, the target training data is parsed and processed to obtain visualization information, which is conducive to analyzing the target deep learning model based on the visualization information.

[0155] See also Figure 3 , Figure 3 It is a structural diagram of a computer device provided in an embodiment of the present application. As shown in the figure, the computer device includes a processor, a memory, a communication interface and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor.

[0156] In one implementation, the computer device is a user device, and the program includes instructions for executing the following steps:

[0157] Sending a visualization request to a cloud server, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0158] Receiving a visualization response sent by the cloud server, where the visualization response carries visualization information of the target training data;

[0159] The visualization information is displayed.

[0160] Optionally, the visualization information includes at least one of the following categories: directed acyclic graph, three-dimensional histogram, call sequence diagram of multiple processes, feature graph, two-dimensional heat map and scalar line graph; and / or

[0161] The target training data includes at least one of the following: model performance change trend information, model loss trend information, model parameter distribution information, model processing intermediate results, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes.

[0162] Optionally, the visualization information includes at least one of the following information:

[0163] Topological information of at least a portion of the target deep learning model, wherein at least a portion of the target deep learning model includes: a plurality of modules and / or a plurality of operators of the target deep learning model;

[0164] Resource occupancy information of each operator in at least one operator included in the target deep learning model.

[0165] Optionally, the topology information of the multiple modules includes at least one of the following:

[0166] The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

[0167] Optionally, the resource occupancy information is determined by the cloud server based on at least one of a data type of the operator, input data information of the operator, and output data information of the operator.

[0168] Optionally, the visualization request carries a training task identifier of the target deep learning model and index information of the target training data, wherein the index information includes: a data type and a data label.

[0169] Optionally, in terms of displaying the visual information, the program includes instructions specifically for executing the following steps:

[0170] In the embodiments of the present disclosure, display of visual data in different modes is supported. In some embodiments, the visual information can be displayed in a folded display mode or a display display mode, or other display modes can be defined, but the embodiments of the present disclosure are not limited thereto.

[0171] In some embodiments, in the folded display mode, a module in the target deep learning model is used as the smallest display unit, wherein the module includes at least one operator. The modules here may be divided by a default division method, or may be set by a user, which is not limited in the embodiments of the present disclosure.

[0172] In some embodiments, in the expanded display mode, the operator in the target deep learning model is used as the smallest display unit. At this time, information of all operators in the model can be displayed.

[0173] In some embodiments, it can also be displayed in a mixed display mode, that is, some modules are displayed in a folded state, while other modules are displayed in an expanded state. It can be displayed optionally based on user settings, which is not limited in the embodiments of the present disclosure.

[0174] Optionally, before sending the visualization request to the cloud server, the program further includes instructions for executing the following steps:

[0175] The intermediate training data of the target deep learning model is sent to the cloud server through a data upload thread, and the intermediate training data is used by the cloud server to obtain index information and a training task identifier of the target training data, and store the index information in a database based on the training task identifier, and store the target training data in the cloud storage based on the index information.

[0176] In another implementation, the computer device is a cloud server, and the program includes instructions for performing the following steps:

[0177] Receiving a visualization request from a user device, wherein the visualization request is used to request visualization of target training data of a target deep learning model;

[0178] Obtaining target training data corresponding to the visualization request;

[0179] Preprocessing the target training data to obtain visualization information;

[0180] A visualization response is sent to the user equipment, where the visualization response carries the visualization information.

[0181] Optionally, the visualization information includes at least one of the following categories: directed acyclic graph, three-dimensional histogram, call sequence diagram of multiple processes, feature graph, two-dimensional heat map and scalar line graph; and / or

[0182] The target training data includes at least one of the following: model performance change trend information, model loss trend information, model parameter distribution information, model processing intermediate results, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes.

[0183] Optionally, the visualization information includes at least one of the following information:

[0184] Topological information of at least a portion of the target deep learning model, wherein at least a portion of the target deep learning model includes: a plurality of modules and / or a plurality of operators of the target deep learning model;

[0185] Resource occupancy information of each operator in at least one operator included in the target deep learning model.

[0186] Optionally, the topology information of the multiple modules includes at least one of the following:

[0187] The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

[0188] Optionally, in terms of preprocessing the target training data to obtain visualization information, the program includes instructions for executing the following steps:

[0189] Based on at least one of a data type of at least one operator in the target deep learning model, input data information of the operator, and output data information of the operator, determine resource occupancy information of each operator in the at least one operator.

[0190] Optionally, the visualization request carries a training task identifier of the target deep learning model and index information of the target training data, where the index information includes: a data type and a data label;

[0191] In terms of obtaining the target training data corresponding to the visualization request, the above-mentioned program includes instructions for executing the following steps: searching the index database based on the training task identifier and index information carried in the visualization request to obtain the storage location information of the target training data; based on the storage location information of the target training data, obtaining the target training data stored in the cloud storage.

[0192] Optionally, in terms of preprocessing the target training data to obtain visualization information, the program includes instructions for executing the following steps:

[0193] Perform at least one preprocessing on the target training data to obtain preprocessed data;

[0194] The preprocessed data is rendered to obtain the visualization information.

[0195] Optionally, in terms of preprocessing the target training data to obtain visualization information, the program includes instructions for executing the following steps:

[0196] When the training task of the target deep learning model is executed under the target framework, the target training data is parsed and processed through a binary tree to obtain the visualization information.

[0197] It should be noted that the specific implementation process of this embodiment can refer to the specific implementation process described in the above method embodiment, which will not be described here.

[0198] See also Figure 4 , Figure 4 A visualization device provided in an embodiment of the present application is applied to a user device, and the device includes:

[0199] A sending unit 401 is used to send a visualization request to a cloud server, where the visualization request is used to request visualization of target training data of a target deep learning model;

[0200] A receiving unit 402 is configured to receive a visualization response sent by the cloud server, wherein the visualization response carries visualization information of the target training data;

[0201] The display unit 403 is used to display the visual information.

[0202] In one implementation of the present application, the visualization information includes at least one of the following categories: a directed acyclic graph, a three-dimensional histogram, a call sequence diagram of multiple processes, a feature diagram, a two-dimensional heat map, and a scalar line graph; and / or

[0203] The target training data includes at least one of the following: model performance change trend information, model loss trend information, model parameter distribution information, model processing intermediate results, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes.

[0204] In an implementation of the present application, the visual information includes at least one of the following information:

[0205] Topological information of at least a portion of the target deep learning model, wherein at least a portion of the target deep learning model includes: a plurality of modules and / or a plurality of operators of the target deep learning model;

[0206] Resource occupancy information of each operator in at least one operator included in the target deep learning model.

[0207] In an implementation of the present application, the topology information of the multiple modules includes at least one of the following:

[0208] The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

[0209] In one implementation of the present application, the resource occupancy information is determined by the cloud server based on at least one of the data type of the operator, the input data information of the operator, and the output data information of the operator.

[0210] In one implementation of the present application, the visualization request carries the training task identifier of the target deep learning model and the index information of the target training data, wherein the index information includes: data type and data label.

[0211] In an implementation of the present application, in terms of displaying the visual information, the display unit 403 is specifically used to:

[0212] The visualization information is displayed in a folded display mode or a displayed display mode, wherein:

[0213] In the folded display mode, a module in the target deep learning model is used as a minimum display unit, wherein the module includes at least one operator; and / or

[0214] In the expanded display mode, the operator in the target deep learning model is used as the smallest display unit.

[0215] In an implementation of the present application, the sending unit 401 is specifically further used to send the intermediate training data of the target deep learning model to the cloud server through the data upload thread before sending the visualization request to the cloud server, and the intermediate training data is used by the cloud server to obtain the index information and training task identifier of the target training data;

[0216] The device also includes: a data storage unit 404, which is used to store the index information in a database based on the training task identifier, and store the target training data in a cloud storage based on the index information.

[0217] It should be noted that the sending unit 401 and the receiving unit 402 can be implemented through a communication interface, the display unit 403 can be implemented through a display screen, and the data storage unit 404 can be implemented through a processor screen.

[0218] See also Figure 5 , Figure 5 A visualization device provided in an embodiment of the present application is applied to a cloud server, and the device includes:

[0219] A receiving unit 501 is used to receive a visualization request from a user device, where the visualization request is used to request visualization of target training data of a target deep learning model;

[0220] An acquisition unit 502 is used to acquire target training data corresponding to the visualization request;

[0221] A determination unit 503 is used to pre-process the target training data to obtain visualization information;

[0222] The sending unit 504 is configured to send a visualization response to the user equipment, where the visualization response carries the visualization information.

[0223] In one implementation of the present application, the visualization information includes at least one of the following categories: a directed acyclic graph, a three-dimensional histogram, a call sequence diagram of multiple processes, a feature diagram, a two-dimensional heat map, and a scalar line graph; and / or

[0224] The target training data includes at least one of the following: model performance change trend information, model loss trend information, model parameter distribution information, model processing intermediate results, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes.

[0225] In an implementation of the present application, the visual information includes at least one of the following information:

[0226] Topological information of at least a portion of the target deep learning model, wherein at least a portion of the target deep learning model includes: a plurality of modules and / or a plurality of operators of the target deep learning model;

[0227] Resource occupancy information of each operator in at least one operator included in the target deep learning model.

[0228] In an implementation of the present application, the topology information of the multiple modules includes at least one of the following:

[0229] The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

[0230] In one implementation of the present application, in terms of preprocessing the target training data to obtain visualization information, the above program includes instructions for executing the following steps:

[0231] Based on at least one of a data type of at least one operator in the target deep learning model, input data information of the operator, and output data information of the operator, determine resource occupancy information of each operator in the at least one operator.

[0232] In an implementation of the present application, the visualization request carries the training task identifier of the target deep learning model and the index information of the target training data, and the index information includes: a data type and a data label;

[0233] In terms of acquiring the target training data corresponding to the visualization request, the acquisition unit 502 is specifically used to:

[0234] Searching an index database based on the training task identifier and index information carried in the visualization request to obtain storage location information of the target training data;

[0235] Based on the storage location information of the target training data, the target training data stored in the cloud storage is acquired.

[0236] In an implementation of the present application, in preprocessing the target training data to obtain visualization information, the determination unit 503 is specifically used to:

[0237] Perform at least one preprocessing on the target training data to obtain preprocessed data;

[0238] The preprocessed data is rendered to obtain the visualization information.

[0239] In an implementation of the present application, in preprocessing the target training data to obtain visualization information, the determination unit 503 is specifically used to:

[0240] When the training task of the target deep learning model is executed under the target framework, the target training data is parsed and processed through a binary tree to obtain the visualization information.

[0241] It should be noted that the receiving unit 501 and the sending unit 504 may be implemented through a communication interface, and the acquiring unit 502 and the determining unit 503 may be implemented through a processor.

[0242] An embodiment of the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute some or all of the steps described by the electronic device or cloud server in the above method embodiment.

[0243] The embodiment of the present application also provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all of the steps described by the electronic device or the cloud server in the above method. The computer program product can be a software installation package.

[0244] The steps of the method or algorithm described in the embodiment of the present application can be implemented in a hardware manner, or can be implemented by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a random access memory (Random Access Memory, RAM), a flash memory, a read-only memory (Read Only Memory, ROM), an erasable programmable read-only memory (Erasable Programmable ROM, EPROM), an electrically erasable programmable read-only memory (Electrically EPROM, EEPROM), a register, a hard disk, a mobile hard disk, a read-only compact disk (CD-ROM) or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in an access network device, a target network device, or a core network device. Of course, the processor and the storage medium can also be present in an access network device, a target network device, or a core network device as discrete components.

[0245] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the embodiments of the present application can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiments of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (Digital Subscriber Line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrations. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital video disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0246] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above description is only the specific implementation method of the embodiments of the present application and is not intended to limit the protection scope of the embodiments of the present application. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application should be included in the protection scope of the embodiments of the present application.

Claims

1. A visualization method, characterized in that: Applied to user equipment, the method comprises: Sending a visualization request to a cloud server, wherein the visualization request is used to request visualization of target training data of a target deep learning model, wherein the target training data includes at least one of the following: model performance change trend information, model parameter distribution information, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes; Receiving a visualization response sent by the cloud server, where the visualization response carries visualization information of the target training data; Displaying the visualization information, wherein the visualization information includes at least one of the following categories: directed acyclic graph, scalar line graph; the visualization information includes at least one of the following information: topological information of at least a part of the target deep learning model, wherein at least a part of the target deep learning model includes: multiple modules and / or multiple operators of the target deep learning model; resource occupancy information of each operator in at least one operator included in the target deep learning model, wherein the operator information includes at least one of the following: identification information of the operator, dependency relationship between operators, and data volume of the operator; In the case where the target training data is the model structure information, the category of the visualization information is a directed acyclic graph, and whether the structure of the target deep learning model is correct can be determined through the directed acyclic graph. In the case where the structure of the target deep learning model is wrong, the structure of the target deep learning model is reconstructed; When the target training data is the model performance change trend information, whether the performance change trend satisfies a first preset condition can be determined based on the scalar line chart. If the first preset condition is met, the training of the target deep learning model is stopped; if the first preset condition is not met, the training of the target deep learning model is continued.

2. The method according to claim 1, characterized in that The visualization information includes at least one of the following categories: a three-dimensional histogram, a call sequence diagram of multiple processes, a feature diagram, a two-dimensional heat map; and / or The target training data includes at least one of the following: model loss trend information and model processing intermediate results.

3. The method according to claim 1, characterized in that The topology information of the multiple modules includes at least one of the following: The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

4. The method according to claim 1 or 3, characterized in that: The resource occupancy information is determined by the cloud server based on at least one of a data type of the operator, input data information of the operator, and output data information of the operator.

5. The method according to any one of claims 1 to 3, characterized in that The visualization request carries the training task identifier of the target deep learning model and the index information of the target training data, wherein the index information includes: data type and data label.

6. The method according to any one of claims 1 to 3, characterized in that The displaying of the visual information comprises: The visualization information is displayed in a collapsed display mode or an expanded display mode, wherein: In the folded display mode, a module in the target deep learning model is used as a minimum display unit, wherein the module includes at least one operator; and / or In the expanded display mode, the operator in the target deep learning model is used as the smallest display unit.

7. The method according to claim 5, characterized in that Before sending the visualization request to the cloud server, the method further includes: The intermediate training data of the target deep learning model is sent to the cloud server through a data upload thread, and the intermediate training data is used by the cloud server to obtain index information and a training task identifier of the target training data, and store the index information in an index database based on the training task identifier, and store the target training data in the cloud storage based on the index information.

8. A visualization method, characterized in that: Applied to a cloud server, the method comprises: Receiving a visualization request from a user device, the visualization request is used to request visualization of target training data of a target deep learning model, the target training data including at least one of the following: model performance change trend information, model parameter distribution information, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes; Obtaining target training data corresponding to the visualization request; Preprocessing the target training data to obtain visualization information; Sending a visualization response to the user equipment, the visualization response carrying the visualization information, the visualization information including at least one of the following categories: a directed acyclic graph, a scalar line graph; the visualization information including at least one of the following information: topological information of at least a part of the target deep learning model, wherein at least a part of the target deep learning model includes: multiple modules and / or multiple operators of the target deep learning model; resource occupancy information of each operator in at least one operator included in the target deep learning model, the operator information including at least one of the following: identification information of the operator, dependency between operators, and data volume of the operator; In the case where the target training data is the model structure information, the category of the visualization information is a directed acyclic graph, and whether the structure of the target deep learning model is correct can be determined through the directed acyclic graph. In the case where the structure of the target deep learning model is wrong, the structure of the target deep learning model is reconstructed; When the target training data is the model performance change trend information, whether the performance change trend satisfies a first preset condition can be determined based on the scalar line chart. If the first preset condition is met, the training of the target deep learning model is stopped; if the first preset condition is not met, the training of the target deep learning model is continued.

9. The method according to claim 8, characterized in that The visualization information includes at least one of the following categories: a three-dimensional histogram, a call sequence diagram of multiple processes, a feature diagram, a two-dimensional heat map; and / or The target training data includes at least one of the following: model loss trend information and model processing intermediate results.

10. The method according to claim 8, characterized in that The topology information of the multiple modules includes at least one of the following: The identification information of the multiple modules, the dependency relationship between the multiple modules, the data size of each module in the multiple modules, and the information of at least one operator included in each module.

11. The method according to any one of claims 8 to 10, characterized in that The preprocessing of the target training data to obtain visualization information includes: Based on at least one of a data type of at least one operator in the target deep learning model, input data information of the operator, and output data information of the operator, determine resource occupancy information of each operator in the at least one operator.

12. The method according to any one of claims 8 to 10, characterized in that The visualization request carries the training task identifier of the target deep learning model and the index information of the target training data, where the index information includes: a data type and a data label; The obtaining target training data corresponding to the visualization request includes: Searching an index database based on the training task identifier and index information carried in the visualization request to obtain storage location information of the target training data; Based on the storage location information of the target training data, the target training data stored in the cloud storage is acquired.

13. The method according to any one of claims 8 to 10, characterized in that The preprocessing of the target training data to obtain visualization information includes: Perform at least one preprocessing on the target training data to obtain preprocessed data; The preprocessed data is rendered to obtain the visualization information.

14. The method according to any one of claims 8 to 10, characterized in that The preprocessing of the target training data to obtain visualization information includes: When the training task of the target deep learning model is executed under the target framework, the target training data is parsed and processed through a binary tree to obtain the visualization information.

15. A visualization device, characterized in that: Applied to user equipment, the device comprises: A sending unit, configured to send a visualization request to a cloud server, wherein the visualization request is used to request visualization of target training data of a target deep learning model, wherein the target training data includes at least one of the following: model performance change trend information, model parameter distribution information, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes; A receiving unit, configured to receive a visualization response sent by the cloud server, wherein the visualization response carries visualization information of the target training data; A display unit, used to display the visualization information, wherein the visualization information includes at least one of the following categories: a directed acyclic graph, a scalar line graph; the visualization information includes at least one of the following information: topological information of at least a part of the target deep learning model, wherein at least a part of the target deep learning model includes: multiple modules and / or multiple operators of the target deep learning model; resource occupancy information of each operator in at least one operator included in the target deep learning model, wherein the operator information includes at least one of the following: identification information of the operator, dependency between operators, and data volume of the operator; In the case where the target training data is the model structure information, the category of the visualization information is a directed acyclic graph, and whether the structure of the target deep learning model is correct can be determined through the directed acyclic graph. In the case where the structure of the target deep learning model is wrong, the structure of the target deep learning model is reconstructed; When the target training data is the model performance change trend information, whether the performance change trend satisfies a first preset condition can be determined based on the scalar line chart. If the first preset condition is met, the training of the target deep learning model is stopped; if the first preset condition is not met, the training of the target deep learning model is continued.

16. A visualization device, characterized in that: Applied to a cloud server, the device comprises: A receiving unit, configured to receive a visualization request from a user device, wherein the visualization request is used to request visualization of target training data of a target deep learning model, wherein the target training data includes at least one of the following: model performance change trend information, model parameter distribution information, model structure information, current progress information of model training, comparison information of different trainings for the same model, and scheduling time information of multiple processes; An acquisition unit, configured to acquire target training data corresponding to the visualization request; A determination unit, used for preprocessing the target training data to obtain visualization information; A sending unit, configured to send a visualization response to the user equipment, wherein the visualization response carries the visualization information, and the visualization information includes at least one of the following categories: a directed acyclic graph, a scalar line graph; the visualization information includes at least one of the following information: topological information of at least a part of the target deep learning model, wherein at least a part of the target deep learning model includes: multiple modules and / or multiple operators of the target deep learning model; resource occupancy information of each operator in at least one operator included in the target deep learning model, wherein the information of the operator includes at least one of the following: identification information of the operator, dependencies between operators, and data volume of the operator; In the case where the target training data is the model structure information, the category of the visualization information is a directed acyclic graph, and whether the structure of the target deep learning model is correct can be determined through the directed acyclic graph. In the case where the structure of the target deep learning model is wrong, the structure of the target deep learning model is reconstructed; When the target training data is the model performance change trend information, whether the performance change trend satisfies a first preset condition can be determined based on the scalar line chart. If the first preset condition is met, the training of the target deep learning model is stopped; if the first preset condition is not met, the training of the target deep learning model is continued.

17. A computer device, characterized in that: The user equipment includes a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for executing the steps in the method as described in any one of claims 1-7 or 8-14.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is processed to execute the method according to any one of claims 1-7 or 8-14.

Images