CN110991649A - Deep learning model building method, device, equipment and storage medium - Google Patents
Deep learning model building method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110991649A CN110991649A CN201911030254.7A CN201911030254A CN110991649A CN 110991649 A CN110991649 A CN 110991649A CN 201911030254 A CN201911030254 A CN 201911030254A CN 110991649 A CN110991649 A CN 110991649A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- learning model
- target
- data set
- operated
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Electrically Operated Instructional Devices (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The application relates to a deep learning model building method, a deep learning model building device, deep learning model building equipment and a storage medium. The method comprises the following steps: the server displays the function components required when the target deep learning model corresponding to the application type is built by acquiring the application type selected by the user based on the visual interface according to the corresponding relation between the application type and the deep learning model, acquires the target function components and parameters selected by the user from the function components required by the target deep learning model, and constructs and stores the target deep learning model according to the parameters of the target function components and the target function components. The server displays the component information and the parameter information required by the model during construction to the user according to the deep learning models of different application types selected by the user, so that the user can set and debug the deep learning model more intuitively and clearly in the process of constructing the personalized deep learning model, and the degree of freedom of the user in constructing the deep learning model is improved.
Description
Technical Field
The application relates to the technical field of artificial intelligence, in particular to a deep learning model building method, device, equipment and storage medium.
Background
Deep neural networks are currently the basis for many human intelligence applications. Due to the breakthrough application of deep neural networks in speech recognition, image recognition and natural language processing, the application amount of deep neural networks is explosively increased. At present, a deep neural network model needs to have certain machine learning knowledge and programming capability, and training of the deep neural network model depends on some software and hardware environments, for example, in the aspect of software environments, a common open source architecture such as tenarslow, pitorch, paddlefold and the like needs to be built, and in the aspect of hardware environments, a Graphics Processing Unit (GPU) display card needs to be purchased to accelerate the model training speed. These factors undoubtedly raise the threshold of user deployment and application of the deep neural network model, and prevent further application and popularization of the deep neural network.
The prior art provides a cloud platform, which provides one-stop service from data processing, model training, service deployment to prediction for traditional machine learning and deep learning. Different deep learning models are packaged in the cloud platform, and a user can directly embed the needed deep learning models into own engineering without building the models.
However, in the cloud platform in the prior art, a user cannot modify a deep learning model in an algorithm according to the service requirement of the user, and the degree of freedom is poor.
Disclosure of Invention
In view of the above, it is necessary to provide a deep learning model building method, apparatus, device and storage medium for solving the above technical problems.
In a first aspect, the present application provides a deep learning model building method, including:
acquiring an application type selected by a user based on a visual interface;
displaying a function component required when a target deep learning model corresponding to the application type is built on a visual interface according to the corresponding relation between the application type and the deep learning model;
acquiring a target function component selected by a user from function components required by a target deep learning model based on a visual interface, and acquiring parameters of the target function component;
and constructing and storing a target deep learning model according to the target function component and the parameters of the target function component.
In one embodiment, the method further includes:
receiving a data set to be operated and a label corresponding to the data set to be operated, which are input by a user based on a visual interface;
testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain a quantification result; the quantification result comprises at least one of a calculation force test result, a data test result, a model parameter test result and a performance test result.
In one embodiment, the method further includes:
and displaying the test quantization process and the quantization result on a visual interface.
In one embodiment, before the testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain the quantification result, the method further includes:
receiving data service operation triggered by a user based on a visual interface; the data service operation comprises a label verification operation and a data enhancement operation; the tag verification operation is used for verifying whether a tag corresponding to the data set to be operated is accurate, and the data enhancement operation is used for expanding the data set to be operated;
processing the data set to be operated and the label corresponding to the data set to be operated according to the data service operation to obtain a processed data set and a processed label;
according to the data set to be operated and the label corresponding to the data set to be operated, the target deep learning model is tested and quantized to obtain a quantization result, and the method comprises the following steps:
and testing and quantifying the target deep learning model according to the processed data set and the processed label to obtain a quantification result.
In one embodiment, the data enhancement operation includes:
and expanding the data set to be operated by adopting at least one mode of up-down overturning, left-right overturning, diagonal overturning, noise adding and shifting.
In one embodiment, the method further includes:
if the quantization result does not accord with the preset model index, returning to execute the step of obtaining the target function component selected by the user from the function components required by the target deep learning model based on the visual interface and obtaining the parameter of the target function component; the model index includes at least one of a calculation power index, a data index, a model parameter index, and a performance index.
In one embodiment, before obtaining the parameters of the objective function component, the method further includes:
displaying parameter information of the objective function component on a visual interface;
receiving a modification instruction input by a user based on a visual interface;
and modifying the parameter information of the target function component according to the modification instruction.
In a second aspect, the present application provides a model building apparatus, the apparatus comprising:
the application type obtaining module is used for obtaining the application type selected by a user based on the visual interface;
the display module is used for displaying a function component required when the target deep learning model corresponding to the application type is built on a visual interface according to the corresponding relation between the application type and the deep learning model;
the acquisition component module is used for acquiring a target function component selected by a user from function components required by the target deep learning model based on a visual interface and acquiring parameters of the target function component;
and the model building module is used for building and storing the target deep learning model according to the target function component and the parameters of the target function component.
In a third aspect, the present application provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the deep learning model building method provided in any embodiment of the first aspect when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the deep learning model building method provided in any one of the embodiments of the first aspect.
According to the deep learning model building method, the device, the equipment and the storage medium, the server displays the function components required by building the target deep learning model corresponding to the application type on the visual interface by acquiring the application type selected by the user based on the visual interface according to the corresponding relation between the application type and the deep learning model, acquires the target function components selected by the user based on the visual interface from the function components required by the target deep learning model, simultaneously acquires the parameters of the target function components, and builds and stores the target deep learning model according to the parameters of the target function components and the target function components. According to the method and the device, the server displays the component information required by the model when the model is built and simultaneously displays the parameter information included by each component according to the deep learning models of different application types selected by the user, so that the user can set and debug the deep learning model more intuitively and clearly in the process of building the personalized deep learning model, and the degree of freedom of building the deep learning model by the user is improved.
Drawings
FIG. 1 is a diagram of an application environment of a deep learning model building method according to an embodiment;
FIG. 2 is a schematic flow chart of a deep learning model building method in one embodiment;
FIG. 3 is a schematic flow chart of a deep learning model building method in another embodiment;
FIG. 4 is a schematic flow chart of a deep learning model building method in another embodiment;
FIG. 5 is a schematic flow chart of a deep learning model building method in another embodiment;
FIG. 6 is a system architecture diagram of a deep learning model building system in another embodiment;
FIG. 7 is a block diagram showing the structure of a deep learning model building apparatus according to an embodiment;
FIG. 8 is a block diagram showing the structure of a deep learning model building apparatus according to another embodiment;
FIG. 9 is a block diagram showing the structure of a deep learning model building apparatus according to another embodiment;
FIG. 10 is a block diagram showing the structure of a deep learning model building apparatus according to another embodiment;
FIG. 11 is a block diagram showing the structure of a deep learning model building apparatus according to another embodiment;
fig. 12 is a block diagram showing the structure of a deep learning model building apparatus according to another embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The deep learning model building method provided by the application can be applied to the application environment shown in FIG. 1. Fig. 1 provides a computer device, which may be a server, and its internal structure diagram may be as shown in fig. 1. The computer device comprises a processor, a memory, a network interface, a database, a display screen and an input device which are connected through a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing deep learning model building data. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a deep learning model building method.
Deep neural networks are currently the basis for many human intelligence applications. Due to the breakthrough application of deep neural networks in speech recognition, image recognition and natural language processing, the application amount of deep neural networks is explosively increased. These deep neural networks are deployed in a variety of applications ranging from autonomous cars, cancer detection to complex games, and so on. In many of these areas, deep neural network models can exceed human accuracy.
In 2017, an artificial intelligence-oriented public data resource library and a cloud service platform are established, a test model and an evaluation model of an artificial intelligence algorithm are established, an artificial intelligence test and evaluation tool set is researched and developed, and application and popularization of the artificial intelligence algorithm are accelerated. In the same year, the construction of a batch of deep learning calculation service platforms is accelerated, and strong calculation support is provided for deep neural network model training. Meanwhile, the construction of an industrial public service platform is promoted, and the output of a standard data set and the generalization application of an artificial intelligence algorithm are constructed.
At present, a deep neural network model needs to have certain machine learning knowledge and programming capability. In addition, deep learning model training depends on some specific software and hardware environments, for example, in the aspect of software environments, a common open source architecture, such as tenserflow, pitorch, paddleadd and the like, needs to be built, and in the aspect of hardware environments, a GPU display card needs to be purchased to accelerate the model training speed.
A plurality of deep learning model building platforms in the prior art are proposed, for example, machine learning platforms PAI, easy DL and the like all provide an open source platform for users, so that the users can call the packaged deep learning models on the platform, but the common defects of the platforms in the prior art are that the packaged deep learning models are all black boxes, the users cannot see the components in the deep learning models, the users cannot modify the parameter settings in the model component functions according to the business demands of the users, and the model building freedom is poor.
Therefore, the technical solutions of the present application and how to solve the above technical problems will be described in detail below by way of embodiments and with reference to the accompanying drawings. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. It should be noted that, in the deep learning model building method provided in the embodiments of fig. 2 to fig. 6 of the present application, the execution main body may be a server, for example, a deep learning model building server, or may also be a deep learning model building device, and the deep learning model building device may be a part or all of the server through software, hardware, or a combination of software and hardware. In the following method embodiments, the following method embodiments are all described by taking the example where the execution subject is a server.
In one embodiment, as shown in fig. 2, a deep learning model building method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:
s201, acquiring the application type selected by the user based on the visual interface.
Wherein, the visual interface refers to an operable interface which shows the application type, all components, component parameters and other functional options to the user; the application type refers to an application type of the deep learning model, and the application type at least comprises different functional types such as image recognition, image classification, image positioning, semantic segmentation and the like.
In this embodiment, a user may click or input a desired application type on a visualization interface, for example, the user may select the desired application type through an application type list in the visualization interface, the application type list may include image recognition, image classification, image positioning, semantic segmentation, and the like, the user may select one of the application types according to a requirement of the user, for example, the image recognition, and the server acquires the application type as the image recognition according to the selection of the user, which is not limited in this embodiment.
S202, according to the corresponding relation between the application type and the deep learning model, displaying a function component required when the target deep learning model corresponding to the application type is built on a visual interface.
The corresponding relation between the application type and the deep learning model refers to that different model types of different applications correspond to different components required for building the deep learning model; exemplarily, the application type is an image recognition corresponding to the image recognition model, components required for building the image recognition model are obtained according to the corresponding relation, and the required components are displayed for a user; the target deep learning model refers to a deep learning model to be built according to the self requirements of the user.
In this embodiment, the server sets a corresponding relationship between an application type and a deep learning model in advance, configures a function component required for building different deep learning models, and configures an image recognition model corresponding to image recognition, an image classification model corresponding to image classification, an image positioning model corresponding to image positioning, a semantic segmentation model corresponding to semantic segmentation, and the like; according to the corresponding relation, the server obtains the application type selected by the user, the components required when the corresponding model is built are displayed for the user, and the server can display the components in a list form and can also display the components in a graph form so that the user can select the target components. For example, the components may include basic components such as contribution, position, batchnormation, and composite components such as contribution block, Resnetblock, inclusion _ respet _ stem, which is not limited in this embodiment.
S203, acquiring a target function component selected by a user from the function components required by the target deep learning model based on the visual interface, and acquiring parameters of the target function component.
The target function component refers to a target deep learning model determined according to the application type selected by the user and corresponds to a required function component; the parameters of the objective function component refer to the inputs, outputs, adjustment values and other related parameters included in the component.
In this embodiment, a user may select a component required for building a target deep learning model by clicking in a component list of a visual interface, or may select a component required for building a target deep learning model by clicking in a component graph of the visual interface, which is not limited in this embodiment, the server displays a setting list of all parameters included in the component to the user in the visual interface while acquiring the component selected by the user, the user may set or modify the parameters by inputting or selecting, and after the setting or modifying is completed, the server acquires corresponding parameters for building the model.
And S204, constructing and storing a target deep learning model according to the target function component and the parameters of the target function component.
The step of saving the target deep learning model refers to saving all function components included in the built target deep learning model, parameters set in the components and the structure of the built target deep learning model, namely the sequential relation among the formed function components.
In this embodiment, the server constructs a deep learning model required by the user according to an objective function component selected by the user and all parameters included in each objective function component, and stores function components, parameters in the function components, and a sequential structure between the function components, which are formed by the deep learning model, for example, the server may store all components, parameters, and a sequential structure included in the deep learning model in a model database, or may export and store all components, parameters, and a sequential structure of the deep learning model in an agreed format; the agreed format may be a YAML file format, which is not limited in this embodiment.
According to the deep learning model building method, the server obtains the application type selected by the user based on the visual interface, displays the function component required by building the target deep learning model corresponding to the application type on the visual interface according to the corresponding relation between the application type and the deep learning model, obtains the target function component selected by the user from the function components required by the target deep learning model based on the visual interface, obtains the parameters of the target function component at the same time, and builds and stores the target deep learning model according to the parameters of the target function component and the target function component. According to the method and the device, the server displays the component information required by the model when the model is built and simultaneously displays the parameter information included by each component according to the deep learning models of different application types selected by the user, so that the user can set and debug the deep learning model more intuitively and clearly in the process of building the personalized deep learning model, and the degree of freedom of building the deep learning model by the user is improved.
In an embodiment, after the target deep learning model is constructed, the target deep learning model may be further quantitatively evaluated to test performance indicators of the target deep learning model, and optionally, as shown in fig. 3, the method further includes:
s301, receiving a data set to be operated and a label corresponding to the data set to be operated, which are input by a user based on a visual interface.
The data set to be operated can be a training set, a verification set, a test set and the like preset by a user, or can be a data set which is not divided; the label corresponding to the data set to be operated refers to a feature label set by a user, for example, if the data set to be operated is subjected to image recognition, whether each sample in the data set to be operated is a cat is identified, at this time, the cat is the label corresponding to the data set to be operated, which is not limited in this embodiment.
In this embodiment, the server may obtain a data set to be operated and a tag corresponding to the data set to be operated, which are directly input by the user based on the visual interface, and may also obtain a data set file to be operated and a tag corresponding to the data set to be operated, which are imported by the user based on a selection toolbar on the visual interface; if the data set to be operated acquired by the server is a data set which is not divided, the server divides the data set according to a division rule preset by a user; if the server acquires the partitioned data set, the next operation is directly performed according to the data set, which is not limited in this embodiment.
S302, testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain a quantification result; the quantification result comprises at least one of a calculation force test result, a data test result, a model parameter test result and a performance test result.
The test quantification refers to that the server trains and evaluates the current deep learning model according to the data set to be operated and the corresponding label; the calculation force test result refers to a hard configuration environment test result during the training of the target deep learning model, the data test result refers to related test results such as data set distribution and training sample quantity during the training of the target deep learning model, the model parameter test result refers to related test results such as component quantity and parameter variable quantity used during the training of the target deep learning model, and the performance test result refers to related test results such as accuracy and loss value of the training set, the verification set or the test set during the training of the target deep learning model. After the server acquires the data set to be operated and the corresponding label thereof, the server trains and evaluates the target deep learning model, the training and evaluation of the server to the target learning model can be performed in a background, and can also be performed by displaying to the user through a skip interface, which is not limited in this embodiment.
Further, if the quantization result does not accord with the preset model index, returning to the step of executing the steps of obtaining the target function component selected by the user from the function components required by the target deep learning model based on the visual interface and obtaining the parameter of the target function component; the model index includes at least one of a calculation power index, a data index, a model parameter index, and a performance index.
In this embodiment, the server obtains the standard value set by the user for the model index, for example, the user may set the standard value of the accuracy of the test set in the performance index to be 80%, and may also set the standard value of the loss value in the performance index to be 5%, which is not limited in this embodiment. The server obtains a test quantization result of the training of the target deep learning model, and determines a model index in the result, for example, the server obtains a performance index in the model index for determination, and if the accuracy of the obtained test set is 78% and the loss value is 10%, it is determined that the quantization result does not meet the preset model index, and the process returns to step 203.
In this embodiment, if the server determines that the test quantization result of the target deep learning model does not conform to the preset model index, the server may return to the visual interface for selecting the target function and the parameter through jumping, so that the user performs optimization adjustment on the target deep learning model according to the test quantization result, which is not limited in this embodiment.
In this embodiment, the server trains and evaluates the target deep learning model according to the data set to be operated and the corresponding label thereof to obtain a quantization result, the quantization result comprises a plurality of test results, and the user can clearly obtain the training result of the target deep learning model through the quantization result, so that whether the test quantization result meets the preset model index is judged, if not, the step of selecting the target function component and the parameter is returned, so that the user can continuously optimize the target deep learning model according to the test quantization result in the process of building the deep learning model.
The model can also be optimally designed according to the result.
In an embodiment, after the server performs training on the deep learning model, the server may display a training result to the user, and the method further includes: and displaying the test quantization process and the quantization result on a visual interface.
Wherein, the measurement quantization process refers to a training process of a target deep learning model; the quantification result comprises at least one of the calculation force test result, the data test result, the model parameter test result and the performance test result.
In this embodiment, the server may show the test quantization process and the quantization result of the target deep learning model to the user in a form of jumping out of a frame, or may show the test quantization process and the quantization result of the target deep learning model to the user in a form of jumping to an interface, which is not limited in this embodiment.
In this embodiment, the server may display the test quantization result of the target deep learning model to the user in multiple ways, so that the user may observe the change of the performance index of the data set in the training process more intuitively and vividly, and the user may evaluate and further optimize the model.
In an embodiment, as shown in fig. 4, before the step 302 "test and quantize the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated, and obtain a quantization result", the method further includes:
s401, receiving data service operation triggered by a user based on a visual interface; the data service operation comprises a label verification operation and a data enhancement operation; the tag verification operation is used for verifying whether the tag corresponding to the data set to be operated is accurate, and the data enhancement operation is used for expanding the data set to be operated.
The label verification operation refers to the verification of the integrity and the accuracy of a label by a server according to a data set to be operated; the data enhancement operation refers to an operation that, in general, the number of data in a data set to be operated is small in actual situations, so that more data is expanded by using an existing data set, the data amount of deep learning model training is increased, and noise data is increased.
In this embodiment, after the server obtains the data set to be operated and the corresponding tag thereof, the server performs data enhancement operation on the data set to be operated, and performs tag verification operation on the tag corresponding to the data set to be operated. For example, the server may perform data expansion on the data set to be operated by way of up-down flipping, left-right flipping, diagonal flipping, noise adding, shifting, brightness adjustment, and the like, which is not limited in this embodiment; if the correct value of the label corresponding to the data set to be operated is "hair is black", the server may verify the label, and if the label appears "black" or "hair", both the labels are incomplete labels, which is not limited in this embodiment.
Optionally, the data enhancement operation includes: and expanding the data set to be operated by adopting at least one mode of up-down overturning, left-right overturning, diagonal overturning, noise adding and shifting.
In this embodiment, the server performs data expansion on the existing data in the data set to be operated in at least one of an up-down flip mode, a left-right flip mode, a diagonal flip mode, a noise adding mode, a shift mode, a brightness adjustment mode and the like, so that the data quantity reaches a target quantity. The image processing method comprises the steps of carrying out vertical mirror surface turning on an image, carrying out horizontal mirror surface turning on the image, carrying out diagonal turning on the image, carrying out mirror surface turning on the image by taking a diagonal line as a symmetry axis, adding noise to eliminate useless high-frequency features in a deep learning model by adding noise data at random, and moving the image along the x direction or the y direction or two directions.
In the embodiment, the server performs data enhancement operation on the data set to be operated in consideration of the possible condition that the number of data in the data set to be operated is small, so that the generalization capability and robustness of the deep learning model are improved to a certain extent.
S402, processing the data set to be operated and the label corresponding to the data set to be operated according to the data service operation to obtain a processed data set and a processed label.
In this embodiment, the server may perform data expansion operation on the data set to be operated by turning the data set upside down, or may perform data expansion operation by adjusting the brightness, to obtain an expanded data set; the server may determine the integrity and accuracy of the tag corresponding to the data set to be operated by using a comparative analysis method, and if the tag is incomplete or inaccurate, modify the tag to obtain a complete and accurate tag, which is not limited in this embodiment.
Further, step 302, "test and quantize the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated, and obtain a quantization result", includes:
and S403, testing and quantifying the target deep learning model according to the processed data set and the processed label to obtain a quantification result.
In this embodiment, the server obtains the data set to be operated after data enhancement and the label after passing the label verification, and performs training evaluation on the target deep learning model according to the data set to be operated and the label to obtain a training evaluation result of the model.
In this embodiment, the server performs data enhancement operation and tag verification operation on the acquired data sets to be operated and corresponding tags, respectively, and performs training on the target deep learning model under the condition that a sufficient number of data sets and tags are correct, so as to improve the accuracy of model training.
In one embodiment, in order to meet different requirements of different users, the user can also modify parameters of the objective function component based on the visual interface. As shown in fig. 5, before "obtaining parameters of the objective function component" in step 203, the method further includes:
s501, displaying parameter information of the objective function component on a visual interface.
In this embodiment, the server obtains an objective function component selected by a user on a visual interface, and displays all parameter information included in the component to the user on the visual interface, where the parameter information may be displayed in the visual interface in a graphical form or in a tabular form, and this embodiment is not limited to this.
And S502, receiving a modification instruction input by a user based on the visual interface.
The modification instruction refers to an instruction triggered by a user to set, modify or adjust parameters in each component according to own requirements.
In this embodiment, the server obtains a modification instruction triggered by a user setting, modifying or adjusting parameters in each component on the visual interface, and the user may modify the visual interface in a manual input manner, or modify the visual interface in a click selection manner or a text input manner, which is not limited in this embodiment. For example, the user may modify a parameter in the component constraint, and a modification instruction received by the server carries information such as the modified parameter and the original parameter in the component constraint.
S503, modifying the parameter information of the target function component according to the modification instruction.
In this embodiment, the server obtains a modification instruction, where the modification instruction may carry information such as a parameter value, a parameter name, and an original parameter value that are modified in a current modified component; the server modifies the corresponding parameter in the component according to the information carried in the modification instruction, which is not limited in this embodiment.
In this embodiment, the server obtains a modification instruction of the user to the parameter of the objective function component based on the visual interface, so that the user can adjust the parameter of the deep learning model according to the user's own requirement, and the obtained target deep learning model achieves the optimal training effect.
It should be understood that although the various steps in the flow charts of fig. 2-5 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-5 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternating with other steps or at least some of the sub-steps or stages of other steps.
In one embodiment, the above method is further explained from the deep learning model building system side. Fig. 6 is an architecture diagram of a deep learning model building system, which includes a visualization layer 11, a control layer 12 and an algorithm layer 13, wherein: the visualization layer 11 includes a model application module 111, a model development module 112, a model management module 113, and a data processing module 114 and a model evaluation module 115; the control layer 12 includes an algorithm calling module 121 and a database connection module 122; the algorithm layer 13 includes a data set processing algorithm module 131 and a component algorithm module 132.
Each layer and each module in the deep learning model building system are deployed on a server in a Docker mirror image mode, and efficient deep learning model building, evaluation, optimization and application services are provided for users in a graphical interface mode. The Docker mirror image deployment means that a deep learning model system based on imaging is made into a Docker mirror image file, and then the creation and the update of the system are realized through mirroring so as to achieve the purpose of rapid deployment of a system operating environment. The updating of the system can be completed in the Docker container template, and the system configuration and deployment are simplified by quickly copying the template after the updating is completed.
As shown in fig. 6, the deep learning model building system architecture diagram, the visualization layer 11 may perform operations of receiving all the user operations based on the visualization interface in the above method; the control layer 12 can execute algorithm calling and database connection operation in the deep learning model building training process in the method; the algorithm layer 13 can execute the algorithm execution operation of the components of the total deep learning model of the method; wherein, specifically:
the model application module 111 is used for acquiring an application type selected by a user based on a visual interface;
the model development module 112 is used for displaying a function component required when a target deep learning model corresponding to the application type is built on a visual interface according to the corresponding relation between the application type and the deep learning model;
the model management module 113 is configured to obtain a target function component selected by a user from function components required by the target deep learning model based on a visual interface, and obtain a parameter of the target function component; the method can also be used for displaying the parameter information of the objective function component on a visual interface; receiving a modification instruction input by a user based on a visual interface; modifying the parameter information of the target function component according to the modification instruction; the method is also used for returning to the step of executing the steps of acquiring the target function component selected by the user from the function components required by the target deep learning model based on the visual interface and acquiring the parameter of the target function component if the quantization result does not accord with the preset model index; the model index comprises at least one of a calculation power index, a data index, a model parameter index and a performance index;
the data processing model 114 is used for receiving a data set to be operated and a label corresponding to the data set to be operated, which are input by a user based on a visual interface; the data service operation triggered by a user based on the visual interface is received; the data service operation comprises a label verification operation and a data enhancement operation; the tag verification operation is used for verifying whether a tag corresponding to the data set to be operated is accurate, and the data enhancement operation is used for expanding the data set to be operated;
the model evaluation module 115 is used for displaying the test quantization process and the quantization result on a visual interface;
the algorithm calling module 121 is used for constructing a target deep learning model according to the target function component and the parameters of the target function component; the target deep learning model is tested and quantized according to the processed data set and the processed label to obtain a quantization result; the target deep learning model is tested and quantized according to the data set to be operated and the label corresponding to the data set to be operated to obtain a quantization result; the quantification result comprises at least one of a calculation force test result, a data test result, a model parameter test result and a performance test result;
the database connection module 122 is used for storing the target deep learning model; deep learning model set up by user and record of each training evaluation for connecting database
The data set processing algorithm module 131 is configured to process the data set to be operated and the tag corresponding to the data set to be operated according to the data service operation, so as to obtain a processed data set and a processed tag; the system comprises a data set to be operated, a data set processing module and a data processing module, wherein the data set to be operated is used for expanding a data set to be operated by adopting at least one mode of up-down overturning, left-right overturning, diagonal overturning, noise adding and shifting;
a component algorithm module 132, specifically configured to provide an algorithm basis for the model management module 03 of the visualization layer;
the implementation principle and technical effect of the deep learning model building system provided by the embodiment are similar to those of the method embodiment, and are not described herein again.
For specific limitations of the deep learning model building system, reference may be made to the above limitations of the deep learning model building system method, which are not described herein again. All layers and modules in the deep learning model building system can be completely or partially realized through software, hardware and a combination thereof. The above layers and modules may be embedded in a hardware form or independent from a processor in a computer device, or may be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the above layers and modules.
In one embodiment, as shown in fig. 7, there is provided a deep learning model building apparatus, including: an obtain application type module 701, a presentation module 702, an obtain component module 703, and a build model module 704, where:
an obtain application type module 701, configured to obtain an application type selected by a user based on a visual interface;
a display module 702, configured to display, on a visual interface, a function component required when a target deep learning model corresponding to an application type is built according to a correspondence between the application type and the deep learning model;
an obtaining component module 703, configured to obtain a target function component selected by a user from function components required by the target deep learning model based on a visual interface, and obtain a parameter of the target function component;
and a model building module 704, configured to build and store the target deep learning model according to the target function component and the parameters of the target function component.
In one embodiment, as shown in fig. 8, the apparatus further comprises a receiving module 705 and a test quantifying module 706, wherein:
the receiving module 705 is configured to receive a data set to be operated and a tag corresponding to the data set to be operated, which are input by a user based on a visual interface;
the test quantization module 706 is configured to perform test quantization on the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated, so as to obtain a quantization result; the quantification result comprises at least one of a calculation force test result, a data test result, a model parameter test result and a performance test result.
In one embodiment, as shown in fig. 9, the apparatus further comprises a display module 707;
and the display module 707 is configured to display the test quantization process and the quantization result on a visual interface.
In one embodiment, as shown in fig. 10, the apparatus further comprises a data processing module 708, wherein:
the receiving module 705 is further configured to receive a data service operation triggered by a user based on a visual interface; the data service operation comprises a label verification operation and a data enhancement operation; the tag verification operation is used for verifying whether a tag corresponding to the data set to be operated is accurate, and the data enhancement operation is used for expanding the data set to be operated;
the data processing module 708 is configured to process the data set to be operated and the tag corresponding to the data set to be operated according to the data service operation, so as to obtain a processed data set and a processed tag;
the test quantization module 706 is further configured to perform test quantization on the target deep learning model according to the processed data set and the processed tag, so as to obtain a quantization result.
In one embodiment, the data enhancement operation includes: and expanding the data set to be operated by adopting at least one mode of up-down overturning, left-right overturning, diagonal overturning, noise adding and shifting.
In one embodiment, as shown in fig. 11, the apparatus further includes a determining module 709;
a judging module 709, configured to return to execute the step of obtaining a target function component selected by the user from function components required by the target deep learning model based on the visual interface and obtaining a parameter of the target function component if the quantization result does not meet a preset model index; the model index includes at least one of a calculation power index, a data index, a model parameter index, and a performance index.
In one embodiment, as shown in fig. 12, the apparatus further comprises a modify parameter module 710, wherein:
the display module 702 is further configured to display parameter information of the objective function component on a visual interface;
the receiving module 705 is further configured to receive a modification instruction input by a user based on the visual interface;
and a parameter modification module 710 for modifying the parameter information of the target function component according to the modification instruction.
The implementation principle and technical effect of all the embodiments of the deep learning model building device are similar to those of the embodiments corresponding to the deep learning model building method, and are not described herein again.
For specific definition of the deep learning model building device, reference may be made to the above definition of the deep learning model building method, which is not described herein again. All or part of each module in the deep learning model building device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may also be a terminal, and its internal structure diagram may be as shown in fig. 1. The computer device comprises a processor, a memory, a database, a network interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a deep learning model building method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like. The database of the computer device is used for storing deep learning model building data.
Those skilled in the art will appreciate that the architecture shown in fig. 1 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring an application type selected by a user based on a visual interface;
displaying a function component required when a target deep learning model corresponding to the application type is built on a visual interface according to the corresponding relation between the application type and the deep learning model;
acquiring a target function component selected by a user from function components required by a target deep learning model based on a visual interface, and acquiring parameters of the target function component;
and constructing and storing a target deep learning model according to the target function component and the parameters of the target function component. The implementation principle and technical effect of the computer device provided by the above embodiment are similar to those of the above method embodiment, and are not described herein again.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring an application type selected by a user based on a visual interface;
displaying a function component required when a target deep learning model corresponding to the application type is built on a visual interface according to the corresponding relation between the application type and the deep learning model;
acquiring a target function component selected by a user from function components required by a target deep learning model based on a visual interface, and acquiring parameters of the target function component;
and constructing and storing a target deep learning model according to the target function component and the parameters of the target function component. The implementation principle and technical effect of the computer-readable storage medium provided by the above embodiments are similar to those of the above method embodiments, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A deep learning model building method is characterized by comprising the following steps:
acquiring an application type selected by a user based on a visual interface;
displaying a function component required when a target deep learning model corresponding to the application type is built on the visual interface according to the corresponding relation between the application type and the deep learning model;
acquiring a target function component selected by a user from function components required by the target deep learning model based on the visual interface, and acquiring parameters of the target function component;
and constructing and storing the target deep learning model according to the target function component and the parameters of the target function component.
2. The method of claim 1, further comprising:
receiving a data set to be operated and a label corresponding to the data set to be operated, which are input by a user based on the visual interface;
testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain a quantification result; the quantification result comprises at least one of a calculation force test result, a data test result, a model parameter test result and a performance test result.
3. The method of claim 2, further comprising:
and displaying the test quantification process and the quantification result on the visual interface.
4. The method according to claim 2 or 3, wherein before the testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain the quantified result, the method further comprises:
receiving data service operation triggered by a user based on the visual interface; the data service operation comprises a label verification operation and a data enhancement operation; the tag verification operation is used for verifying whether a tag corresponding to the data set to be operated is accurate or not, and the data enhancement operation is used for expanding the data set to be operated;
processing the data set to be operated and the label corresponding to the data set to be operated according to the data service operation to obtain a processed data set and a processed label;
the step of testing and quantifying the target deep learning model according to the data set to be operated and the label corresponding to the data set to be operated to obtain a quantification result includes:
and testing and quantifying the target deep learning model according to the processed data set and the processed label to obtain a quantification result.
5. The method of claim 4, wherein the data enhancement operation comprises:
and expanding the data set to be operated by adopting at least one mode of up-down overturning, left-right overturning, diagonal overturning, noise adding and shifting.
6. The method of claim 2, further comprising:
if the quantization result does not accord with the preset model index, returning to execute the step of selecting a target function component from function components required by the target deep learning model based on the visual interface and acquiring the parameter of the target function component by the acquiring user; the model index includes at least one of a calculation power index, a data index, a model parameter index, and a performance index.
7. The method of claim 1, wherein before obtaining the parameters of the objective function component, further comprising:
displaying parameter information of the objective function component on the visual interface;
receiving a modification instruction input by a user based on the visual interface;
and modifying the parameter information of the target function component according to the modification instruction.
8. An apparatus for building a deep learning model, the apparatus comprising:
the application type obtaining module is used for obtaining the application type selected by a user based on the visual interface;
the display module is used for displaying a function component required by the establishment of a target deep learning model corresponding to the application type on the visual interface according to the corresponding relation between the application type and the deep learning model;
the acquisition component module is used for acquiring a target function component selected by a user from function components required by the target deep learning model based on the visual interface and acquiring parameters of the target function component;
and the model building module is used for building and storing the target deep learning model according to the target function component and the parameters of the target function component.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030254.7A CN110991649A (en) | 2019-10-28 | 2019-10-28 | Deep learning model building method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030254.7A CN110991649A (en) | 2019-10-28 | 2019-10-28 | Deep learning model building method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110991649A true CN110991649A (en) | 2020-04-10 |
Family
ID=70082577
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911030254.7A Pending CN110991649A (en) | 2019-10-28 | 2019-10-28 | Deep learning model building method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110991649A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861020A (en) * | 2020-07-27 | 2020-10-30 | 深圳壹账通智能科技有限公司 | Model deployment method, device, equipment and storage medium |
| CN112114892A (en) * | 2020-08-11 | 2020-12-22 | 北京奇艺世纪科技有限公司 | Deep learning model obtaining method, loading method and selecting method |
| CN112256537A (en) * | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model running state display method and device, computer equipment and storage medium |
| CN112416301A (en) * | 2020-10-19 | 2021-02-26 | 山东产研鲲云人工智能研究院有限公司 | Deep learning model development method and device and computer readable storage medium |
| CN112528631A (en) * | 2020-12-03 | 2021-03-19 | 上海谷均教育科技有限公司 | Intelligent accompaniment system based on deep learning algorithm |
| CN112560344A (en) * | 2020-12-14 | 2021-03-26 | 北京云歌科技有限责任公司 | Method and device for constructing model servo system |
| CN112597239A (en) * | 2020-12-30 | 2021-04-02 | 厦门渊亭信息科技有限公司 | Visual modeling method and device and electronic equipment |
| CN112860779A (en) * | 2021-03-29 | 2021-05-28 | 中信银行股份有限公司 | Batch data importing method and device |
| CN113110833A (en) * | 2021-04-15 | 2021-07-13 | 成都新希望金融信息有限公司 | Machine learning model visual modeling method, device, equipment and storage medium |
| CN113157183A (en) * | 2021-04-15 | 2021-07-23 | 成都新希望金融信息有限公司 | Deep learning model construction method and device, electronic equipment and storage medium |
| CN113971032A (en) * | 2021-12-24 | 2022-01-25 | 百融云创科技股份有限公司 | Full-process automatic deployment method and system of machine learning model for code generation |
| WO2022205835A1 (en) * | 2021-04-02 | 2022-10-06 | 上海商汤科技开发有限公司 | Deep learning model production system, electronic device and storage medium |
| CN115456150A (en) * | 2022-10-18 | 2022-12-09 | 北京鼎成智造科技有限公司 | Reinforced learning model construction method and system |
| CN116974922A (en) * | 2023-07-25 | 2023-10-31 | 摩尔线程智能科技(北京)有限责任公司 | Performance analysis method, device, equipment and storage medium of deep learning model |
| CN117093259A (en) * | 2023-10-20 | 2023-11-21 | 腾讯科技（深圳）有限公司 | Model configuration method and related equipment |
| WO2023245522A1 (en) * | 2022-06-22 | 2023-12-28 | 极纳人工智能有限公司 | Method and apparatus for generating target deep learning model |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108319456A (en) * | 2018-01-29 | 2018-07-24 | 徐磊 | A kind of development approach for exempting to program deep learning application |
| US20180218287A1 (en) * | 2017-02-01 | 2018-08-02 | Facebook, Inc. | Determining performance of a machine-learning model based on aggregation of finer-grain normalized performance metrics |
| US20180285748A1 (en) * | 2017-03-31 | 2018-10-04 | Facebook, Inc. | Performance metric prediction for delivery of electronic media content items |
| US20180307978A1 (en) * | 2017-04-19 | 2018-10-25 | International Business Machines Corporation | Multi-Modal Construction of Deep Learning Networks |
| CN108734287A (en) * | 2017-04-21 | 2018-11-02 | 展讯通信（上海）有限公司 | Compression method and device, terminal, the storage medium of deep neural network model |
| CN108881446A (en) * | 2018-06-22 | 2018-11-23 | 深源恒际科技有限公司 | A kind of artificial intelligence plateform system based on deep learning |
| CN109325541A (en) * | 2018-09-30 | 2019-02-12 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern |
| CN109389219A (en) * | 2017-08-04 | 2019-02-26 | 三星电子株式会社 | The method and apparatus quantified for the parameter to neural network |
| CN109409533A (en) * | 2018-09-28 | 2019-03-01 | 深圳乐信软件技术有限公司 | A kind of generation method of machine learning model, device, equipment and storage medium |
| CN109816114A (en) * | 2018-12-29 | 2019-05-28 | 大唐软件技术股份有限公司 | A kind of generation method of machine learning model, device |
| CN110163269A (en) * | 2019-05-09 | 2019-08-23 | 北京迈格威科技有限公司 | Model generating method, device and computer equipment based on deep learning |
| CN110210624A (en) * | 2018-07-05 | 2019-09-06 | 第四范式（北京）技术有限公司 | Execute method, apparatus, equipment and the storage medium of machine-learning process |
| US20190274337A1 (en) * | 2018-03-08 | 2019-09-12 | Midea Group Co., Ltd. | Smart rice cookers capable of mixed grain cooking and abnormal conditions detection |
| CN110276456A (en) * | 2019-06-20 | 2019-09-24 | 山东大学 | A kind of machine learning model auxiliary construction method, system, equipment and medium |
| US20190313833A1 (en) * | 2018-04-16 | 2019-10-17 | Midea Group Co., Ltd. | Multi-purpose smart rice cookers |
-
2019
- 2019-10-28 CN CN201911030254.7A patent/CN110991649A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180218287A1 (en) * | 2017-02-01 | 2018-08-02 | Facebook, Inc. | Determining performance of a machine-learning model based on aggregation of finer-grain normalized performance metrics |
| US20180285748A1 (en) * | 2017-03-31 | 2018-10-04 | Facebook, Inc. | Performance metric prediction for delivery of electronic media content items |
| US20180307978A1 (en) * | 2017-04-19 | 2018-10-25 | International Business Machines Corporation | Multi-Modal Construction of Deep Learning Networks |
| CN108734287A (en) * | 2017-04-21 | 2018-11-02 | 展讯通信（上海）有限公司 | Compression method and device, terminal, the storage medium of deep neural network model |
| CN109389219A (en) * | 2017-08-04 | 2019-02-26 | 三星电子株式会社 | The method and apparatus quantified for the parameter to neural network |
| CN108319456A (en) * | 2018-01-29 | 2018-07-24 | 徐磊 | A kind of development approach for exempting to program deep learning application |
| US20190274337A1 (en) * | 2018-03-08 | 2019-09-12 | Midea Group Co., Ltd. | Smart rice cookers capable of mixed grain cooking and abnormal conditions detection |
| US20190313833A1 (en) * | 2018-04-16 | 2019-10-17 | Midea Group Co., Ltd. | Multi-purpose smart rice cookers |
| CN108881446A (en) * | 2018-06-22 | 2018-11-23 | 深源恒际科技有限公司 | A kind of artificial intelligence plateform system based on deep learning |
| CN110210624A (en) * | 2018-07-05 | 2019-09-06 | 第四范式（北京）技术有限公司 | Execute method, apparatus, equipment and the storage medium of machine-learning process |
| CN109409533A (en) * | 2018-09-28 | 2019-03-01 | 深圳乐信软件技术有限公司 | A kind of generation method of machine learning model, device, equipment and storage medium |
| CN109325541A (en) * | 2018-09-30 | 2019-02-12 | 北京字节跳动网络技术有限公司 | Method and apparatus for training pattern |
| CN109816114A (en) * | 2018-12-29 | 2019-05-28 | 大唐软件技术股份有限公司 | A kind of generation method of machine learning model, device |
| CN110163269A (en) * | 2019-05-09 | 2019-08-23 | 北京迈格威科技有限公司 | Model generating method, device and computer equipment based on deep learning |
| CN110276456A (en) * | 2019-06-20 | 2019-09-24 | 山东大学 | A kind of machine learning model auxiliary construction method, system, equipment and medium |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861020A (en) * | 2020-07-27 | 2020-10-30 | 深圳壹账通智能科技有限公司 | Model deployment method, device, equipment and storage medium |
| CN112114892A (en) * | 2020-08-11 | 2020-12-22 | 北京奇艺世纪科技有限公司 | Deep learning model obtaining method, loading method and selecting method |
| CN112416301A (en) * | 2020-10-19 | 2021-02-26 | 山东产研鲲云人工智能研究院有限公司 | Deep learning model development method and device and computer readable storage medium |
| CN112256537A (en) * | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model running state display method and device, computer equipment and storage medium |
| CN112256537B (en) * | 2020-11-12 | 2024-03-29 | 腾讯科技（深圳）有限公司 | Model running state display method and device, computer equipment and storage medium |
| CN112528631A (en) * | 2020-12-03 | 2021-03-19 | 上海谷均教育科技有限公司 | Intelligent accompaniment system based on deep learning algorithm |
| CN112528631B (en) * | 2020-12-03 | 2022-08-09 | 上海谷均教育科技有限公司 | Intelligent accompaniment system based on deep learning algorithm |
| CN112560344A (en) * | 2020-12-14 | 2021-03-26 | 北京云歌科技有限责任公司 | Method and device for constructing model servo system |
| CN112560344B (en) * | 2020-12-14 | 2023-12-08 | 北京云歌科技有限责任公司 | Method and device for constructing model servo system |
| CN112597239A (en) * | 2020-12-30 | 2021-04-02 | 厦门渊亭信息科技有限公司 | Visual modeling method and device and electronic equipment |
| CN112860779B (en) * | 2021-03-29 | 2024-05-24 | 中信银行股份有限公司 | Batch data importing method and device |
| CN112860779A (en) * | 2021-03-29 | 2021-05-28 | 中信银行股份有限公司 | Batch data importing method and device |
| WO2022205835A1 (en) * | 2021-04-02 | 2022-10-06 | 上海商汤科技开发有限公司 | Deep learning model production system, electronic device and storage medium |
| CN113157183A (en) * | 2021-04-15 | 2021-07-23 | 成都新希望金融信息有限公司 | Deep learning model construction method and device, electronic equipment and storage medium |
| CN113110833A (en) * | 2021-04-15 | 2021-07-13 | 成都新希望金融信息有限公司 | Machine learning model visual modeling method, device, equipment and storage medium |
| CN113971032A (en) * | 2021-12-24 | 2022-01-25 | 百融云创科技股份有限公司 | Full-process automatic deployment method and system of machine learning model for code generation |
| WO2023245522A1 (en) * | 2022-06-22 | 2023-12-28 | 极纳人工智能有限公司 | Method and apparatus for generating target deep learning model |
| CN115456150A (en) * | 2022-10-18 | 2022-12-09 | 北京鼎成智造科技有限公司 | Reinforced learning model construction method and system |
| CN116974922A (en) * | 2023-07-25 | 2023-10-31 | 摩尔线程智能科技(北京)有限责任公司 | Performance analysis method, device, equipment and storage medium of deep learning model |
| CN116974922B (en) * | 2023-07-25 | 2024-05-17 | 摩尔线程智能科技(北京)有限责任公司 | Performance analysis method, device, equipment and storage medium of deep learning model |
| CN117093259A (en) * | 2023-10-20 | 2023-11-21 | 腾讯科技（深圳）有限公司 | Model configuration method and related equipment |
| CN117093259B (en) * | 2023-10-20 | 2024-02-27 | 腾讯科技（深圳）有限公司 | Model configuration method and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110991649A (en) | Deep learning model building method, device, equipment and storage medium | |
| CN111126668B (en) | Spark operation time prediction method and device based on graph convolution network | |
| JP6182242B1 (en) | Machine learning method, computer and program related to data labeling model | |
| CN111797998B (en) | Method and system for generating combined features of machine learning samples | |
| CN110109750B (en) | Virtual resource acquisition method, device, computer equipment and storage medium | |
| CN109285105A (en) | Method of detecting watermarks, device, computer equipment and storage medium | |
| US20190339670A1 (en) | System and method for lattice structure design for additive manufacturing | |
| CN107886082B (en) | Method and device for detecting mathematical formulas in images, computer equipment and storage medium | |
| CN110414570B (en) | Image classification model generation method, device, equipment and storage medium | |
| JP2010002370A (en) | Pattern extraction program, technique, and apparatus | |
| CN111401472B (en) | Infrared target classification method and device based on deep convolutional neural network | |
| CN110990053A (en) | Method for creating and using machine learning scheme template and device | |
| US10984518B2 (en) | Methods and systems for assessing the quality of geospatial data | |
| CN111930932A (en) | Knowledge graph representation learning method and device in network space security field | |
| CN110517221B (en) | Gap positioning method and device based on real coordinates and storage medium | |
| CN111160394A (en) | Training method and device of classification network, computer equipment and storage medium | |
| CN111242167A (en) | Distributed image annotation method and device, computer equipment and storage medium | |
| CN114549849A (en) | Image recognition method and device, computer equipment and storage medium | |
| US20210182701A1 (en) | Virtual data scientist with prescriptive analytics | |
| CN116823758A (en) | Defect detection method, device, equipment and storage medium | |
| JP2006318232A (en) | Analytical mesh correction device | |
| CN115237097A (en) | Automatic driving simulation test method, device, computer equipment and storage medium | |
| CN110597874B (en) | Data analysis model creation method and device, computer equipment and storage medium | |
| CN111190574B (en) | Method, device, equipment and storage medium for selecting options of multi-stage linkage assembly | |
| CN110852400A (en) | Classification model evaluation method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 511300 No.78, west of Zhucun Avenue, Zhucun street, Zengcheng District, Guangzhou City, Guangdong Province Applicant after: CHINA ELECTRONIC PRODUCT RELIABILITY AND ENVIRONMENTAL TESTING RESEARCH INSTITUTE ((THE FIFTH ELECTRONIC RESEARCH INSTITUTE OF MIIT)(CEPREI LABORATORY)) Address before: 510610 No. 110 Zhuang Road, Tianhe District, Guangdong, Guangzhou, Dongguan Applicant before: CHINA ELECTRONIC PRODUCT RELIABILITY AND ENVIRONMENTAL TESTING RESEARCH INSTITUTE ((THE FIFTH ELECTRONIC RESEARCH INSTITUTE OF MIIT)(CEPREI LABORATORY)) |
|
| CB02 | Change of applicant information |