Data processing method, electronic device, and computer readable storage medium

Abstract

The present disclosure relates to the fields of large model technology and data processing. Disclosed are a data processing method, an electronic device, and a computer readable storage medium. The method comprises: responding to a plurality of received query requests, wherein the query requests carry query data; encapsulating the plurality of pieces of query data to obtain batch handled data; and calling a target data processing model corresponding to the plurality of query requests to process the batch handled data to obtain reply results corresponding to the plurality of query data. The present disclosure solves the technical problem in the related art of low data processing efficiency.

Description

Data processing method, electronic device, and computer-readable storage medium Technical Field

[0001] The present disclosure relates to large model technology and data processing fields, and in particular, to a data processing method, an electronic device, and a computer-readable storage medium. Background Art

[0002] Currently, at the overall solution level, the existing inference service framework mainly supports the service-oriented data processing model, but focuses more on aspects such as model service scheduling and hardware acceleration, and pays less attention to the underlying general model calls. When multiple data need to be processed, they are generally processed in sequence by calling pre-set models, which leads to low data processing efficiency.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far.

[0004] Summary of the Invention

[0005] The embodiments of the present disclosure provide a data processing method, an electronic device, and a computer-readable storage medium to at least solve the technical problem of low data processing efficiency in related technologies.

[0006] According to one aspect of an embodiment of the present disclosure, a data processing method is provided, comprising: responding to multiple received query requests, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; and calling target data processing models corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data.

[0007] According to one aspect of an embodiment of the present disclosure, a data processing method is also provided, including: responding to multiple query requests received in a generative dialogue interface, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling a target large model corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data; and displaying the reply results in the generative dialogue interface.

[0008] According to one aspect of an embodiment of the present disclosure, a data processing method is also provided, including: responding to multiple query requests received in a generative dialogue interface, wherein the query requests carry query data, and the query data at least contains character information; encapsulating the multiple query data to obtain batch data; calling target data processing models corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data, wherein the type of the reply result includes one of the following: text, image, video and voice; and displaying the reply result in the generative dialogue interface.

[0009] According to one aspect of an embodiment of the present disclosure, a data processing method is also provided, including: receiving multiple query requests sent by at least one client in a dialogue mode, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling target data processing models corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data; sending the reply results to the corresponding client, wherein the reply results are displayed by the corresponding client.

[0010] According to one aspect of an embodiment of the present disclosure, a data processing method is also provided, including: obtaining multiple query requests displayed in a generative dialogue interface in a dialogue mode by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is multiple query requests, and the query requests carry query data; encapsulating the multiple query data to obtain batch data; using a target data processing model corresponding to the multiple query requests to process the batch data to obtain reply results corresponding to the multiple query data; outputting the reply result by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the reply result.

[0011] According to one aspect of an embodiment of the present disclosure, an electronic device is further provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes any one of the methods in the above embodiments when running.

[0012] According to one aspect of an embodiment of the present disclosure, a computer-readable storage medium is further provided, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the methods in the above embodiments.

[0013] In an embodiment of the present disclosure, a plurality of query requests are received, wherein the query requests carry query data; the plurality of query data are encapsulated to obtain batch data; a target data processing model corresponding to the plurality of query requests is called to process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the inference efficiency of the plurality of query data; it is easy to notice that the plurality of query data can be optimized by encapsulating the plurality of query data into batch data, and the target data processing model can be called according to the services required by the plurality of query requests, so that the target data processing module can quickly process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the processing efficiency of the plurality of query data, and further solving the technical problem of low data processing efficiency in related technologies.

[0014] It is easy to note that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present disclosure, and do not constitute a limitation of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0016] FIG1 is a schematic diagram of a hardware environment of a virtual reality device according to a data processing method according to an embodiment of the present disclosure;

[0017] FIG2 is a flow chart of a data processing method according to Embodiment 1 of the present disclosure;

[0018] FIG3 is a diagram of a service framework architecture according to an embodiment of the present disclosure;

[0019] FIG4 is a flow chart of a data processing method according to Embodiment 2 of the present disclosure;

[0020] FIG5 is a flow chart of a data processing method according to Embodiment 3 of the present disclosure;

[0021] FIG6 is a flow chart of a data processing method according to Embodiment 4 of the present disclosure;

[0022] FIG7 is a flow chart of a data processing method according to Embodiment 5 of the present disclosure;

[0023] FIG8 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 6 of the present disclosure;

[0024] FIG9 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 7 of the present disclosure;

[0025] FIG10 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 8 of the present disclosure;

[0026] FIG11 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 9 of the present disclosure;

[0027] FIG12 is a schematic diagram of an accelerated deployment device for a deep learning model according to Embodiment 10 of the present disclosure;

[0028] FIG13 is a structural block diagram of a computer terminal according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present disclosure.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. 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 units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] The technical solution provided by the present disclosure is mainly implemented using large-scale model technology. The large model here refers to a deep learning model with large-scale model parameters, which can usually contain hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. The large model can also be called a cornerstone model / foundation model (Foundation Model). The large model is pre-trained by large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks and has good generalization capabilities, such as large-scale language models (LLMs) and multi-modal pre-training models.

[0032] It should be noted that when the large model is actually applied, the pre-trained model can be fine-tuned through a small number of samples, so that the large model can be applied to different tasks. For example, the large model can be widely used in fields such as natural language processing (NLP), computer vision, etc., and can be specifically applied to computer vision tasks such as visual question answering (VQA), image caption (IC), and image generation. It can also be widely used in natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. Therefore, the main application scenarios of the large model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. In the embodiment of the present disclosure, data processing through a data processing model in a generative dialogue scenario is used as an example for explanation.

[0033] First, some nouns or terms that appear in the description of the embodiments of the present disclosure are subject to the following explanations:

[0034] Generative AI models (Artificial Intelligence Generated Content, AIGC): These can generate images based on text, or generate images based on text and images.

[0035] Model Service: This can be used in the service framework to call the model service through requests to complete some reasoning tasks;

[0036] Model Pipeline: This can be achieved by integrating the model pipeline code into the service framework, starting and loading the model through scripts, and completing some inference tasks.

[0037] Upsampling: It can be the process of improving the sampling accuracy of an image. For example, it can be the process of adjusting the image from 256*256 precision to 512*512 precision.

[0038] Example 1

[0039] According to an embodiment of the present disclosure, a data processing method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0040] Taking into account the huge amount of model parameters of the large model and the limited computing resources of the mobile terminal, Figure 1 is a schematic diagram of the hardware environment of a virtual reality device according to a data processing method of an embodiment of the present disclosure; the above-mentioned data processing method provided by the embodiment of the present disclosure can be applied to the application scenario shown in Figure 1, but is not limited thereto. In the application scenario shown in Figure 1, the large model is deployed in a server 10, and the server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client devices 20 here may include but are not limited to: smart phones, tablet computers, laptops, PDAs, personal computers, smart home devices, car-mounted devices, etc. The client device 20 can interact with the user through a graphical user interface to implement the call to the large model, thereby implementing the method provided by the embodiment of the present disclosure.

[0041] In an embodiment of the present disclosure, a system consisting of a client device and a server can perform the following steps: the client device generates multiple query requests in a generative dialogue interface; the server executes the received multiple query requests, encapsulates the multiple query data, and obtains batch data; invokes target data processing models corresponding to the multiple query requests, processes the batch data, and obtains response results corresponding to the multiple query data; and displays the response results in the generative dialogue interface. It should be noted that the present embodiment can be performed in the client device if the operating resources of the client device can meet the deployment and operation requirements of the large model.

[0042] In the above operating environment, the present disclosure provides a data processing method as shown in Figure 2. Figure 2 is a flow chart of the data processing method according to Example 1 of the present disclosure. As shown in Figure 2, the method may include the following steps:

[0043] Step S202: responding to multiple query requests received in the generative dialogue interface.

[0044] The query request carries query data, and the query data includes at least one of the following: text and image.

[0045] The above-mentioned generative dialogue interface can respond in real time based on the user's input. The generative dialogue interface can include a user input box and a system reply box.

[0046] The above-mentioned generative dialogue interface can be applied to a variety of scenarios, such as customer service interaction scenarios in the e-commerce field, doctor question-and-answer scenarios in the medical field, student question-and-answer scenarios in the education field, and search scenarios in the search field. There is no limitation on the application scenarios of the generative dialogue interface here.

[0047] The above-mentioned inquiry data includes at least one of the following: text, image, video, audio, etc., and the modality of the inquiry data is not limited here.

[0048] In an optional embodiment, the user can enter inquiry data in the user input box, where the inquiry data can be questions, instructions or other content that requires interaction. Multiple inquiry requests carrying the inquiry data can be generated, and reply results can be obtained by processing the multiple inquiry requests, and the reply results can be displayed in the system reply box.

[0049] In another optional embodiment, multiple query requests can first reach the model inference service framework (Aquila), which can forward the multiple query requests to the service layer for subsequent processing. The model inference service framework is mainly used to package the model inference process into a service provided to the outside world.

[0050] Step S204: encapsulate the multiple query data to obtain batch data.

[0051] The above encapsulation is used to represent the combination of multiple query data in order to improve the overall computational efficiency of the model.

[0052] In an optional embodiment, a plurality of query data may be encapsulated by a batch handler to obtain batch data (batch); optionally, the plurality of query data may be converted into a plurality of vectors, the plurality of vectors may be combined into a matrix, and the matrix may be determined as the batch data.

[0053] In another optional embodiment, the query data is transformed into a vector and entered into the model. Encapsulating multiple query data essentially combines these vectors into a matrix. A matrix optimization algorithm can be used to improve the overall efficiency of matrix operations. Therefore, inputting the resulting batch data into the model achieves higher processing efficiency than inputting multiple query data into the model individually. The aforementioned matrix optimization algorithm can be any algorithm that can improve matrix computational efficiency, and no limitation is imposed herein.

[0054] Step S206 , calling the target data processing model corresponding to the multiple query requests, processing the batch data, and obtaining response results corresponding to the multiple query data.

[0055] The above-mentioned data processing model can be a large-scale pre-trained model (LLM) or a generative artificial intelligence model (AIGC), but is not limited to this. The data processing model can be set according to actual needs.

[0056] In an optional embodiment, the required processing methods for multiple query data can be determined based on the attribute information or request types of the multiple query requests, and the target data processing model can be determined based on the required processing methods. The batch data can be processed through the target data processing model, so that the response results corresponding to the multiple query data can be obtained.

[0057] In another optional embodiment, when multiple query requests are for creating pictures based on text, or for creating pictures based on text and pictures, it is indicated that the processing method of the multiple query data is to generate images. At this time, it can be determined that the target data processing model is AIGC. Through AIGC, pictures can be created based on text or text and pictures, thereby obtaining generated pictures, and the generated pictures are returned to the user as a reply result.

[0058] In another optional embodiment, when multiple query requests are for data search based on text, it is indicated that the processing method of the multiple query data is to search for content related to the query data. At this time, the target data processing model can be determined to be LLM. Through LLM, data search can be performed based on multiple query data to obtain search results, and the search results are returned to the user as a reply result.

[0059] Step S208: Display the reply result in the generative dialogue interface.

[0060] The above-mentioned reply results can be text or images, and there is no limitation on the form of the reply results.

[0061] In an optional embodiment, the reply results corresponding to the multiple query data can be displayed in the generated dialogue interface when the query request corresponding to the multiple query data is received. Optionally, the reply results can be displayed in the system reply box of the generative dialogue interface.

[0062] The LLM model and AIGC model described above share similarities in that they require batch inference and take a long time. To ensure code consistency, the inference service frameworks of the LLM and AIGC can be integrated to form a unified inference service framework. This unifies the multimodal and AIGC tools for image and video processing, storage, and conversion, significantly improving code reuse. Batch inference refers to the process of inputting multiple inference samples into the model for inference. Compared to inference on individual samples, batch inference fully utilizes the parallel computing capabilities of hardware resources, improving inference speed.

[0063] Through the above steps, multiple query requests received in the generative dialogue interface are responded to, wherein the query requests carry query data, and the query data includes at least one of the following: text, image; the multiple query data are encapsulated to obtain batch data; the target data processing model corresponding to the multiple query requests is called to process the batch data to obtain reply results corresponding to the multiple query data; the reply results are displayed in the generative dialogue interface, thereby improving the reasoning efficiency of the multiple query data; it is easy to notice that the multiple query data can be optimized by encapsulating the multiple query data into batch data, and the target data processing model can be called according to the services required by the multiple query requests, so that the target data processing module can quickly process the batch data and obtain reply results corresponding to the multiple query data, thereby improving the processing efficiency of the multiple query data, thereby solving the technical problem of low data processing efficiency in related technologies.

[0064] In the above-mentioned embodiment of the present disclosure, a target data processing model corresponding to a plurality of query requests is called to process batch data to obtain a response result corresponding to the plurality of query data, including: calling a target data processing service based on the attribute information of the plurality of query requests; constructing a target data processing model through the target data processing service; and processing the batch data using the target data processing model to obtain a response result.

[0065] The attribute information of the plurality of inquiry data may be the type and format of the inquiry data, etc., and no limitation is imposed on the attribute information of the inquiry data.

[0066] In an optional embodiment, the target data processing service can be called based on the attribute information of multiple query requests at the abstract layer, where the abstract layer is a programming concept that separates the underlying implementation details from the upper-level logic. In this context, the abstract layer is used to call the target data processing service through attribute information.

[0067] In another optional embodiment, a build function (Build) can be used to build a target data processing model based on the target data processing service, and batch data can be input into the target data processing model to obtain a response result.

[0068] In the above embodiment of the present disclosure, a target data processing service is called based on the attribute information of multiple query requests, including: determining a target proxy object from proxy objects corresponding to multiple data processing models based on the attribute information of multiple query requests; and calling the target data processing service through the target proxy object.

[0069] The above-mentioned multiple data processing models can be models of the same type, or they can be models of different types. The data processing model can be an LLM model, an AIGC model, an open source model, or a closed source model. There is no limitation on the data processing model here, and it can be any data processing model.

[0070] The above-mentioned open source model refers to a model in which the source code of software or technology can be publicly, freely accessed and modified; the above-mentioned closed source model refers to a model in which the source code of software or technology is not public or accessible.

[0071] The proxy object corresponding to the above data processing model can be a server or software that provides intermediate services. The target data processing service corresponding to the query request can be called through the target proxy object to access different services more conveniently and securely.

[0072] The proxy objects corresponding to the above-mentioned multiple data processing models may be proxy objects of the LLM model, proxy objects of the AIGC model, or proxy objects of other models.

[0073] In another optional embodiment, the required model service can be determined at the proxy service layer based on the task type, name and other attribute information of multiple query requests, and the target proxy objects corresponding to multiple query requests can be determined according to the service types provided by different proxy objects, so as to call the target data processing service through the target proxy object.

[0074] In the above embodiments of the present disclosure, the target data processing model is a pre-deployed model or a model in a model pipeline.

[0075] The above-mentioned model pipeline can be a pre-deployed model pipeline or a non-pre-deployed model pipeline, which is not limited here.

[0076] In an optional embodiment, the model service (Model Service) or the model in the model pipeline (Model Pipeline) can be pre-deployed through the Elastic Algorithm Service (EAS).

[0077] The pre-deployed model service mentioned above can use different request methods (Http, RPC) to call the model service framework to complete some inference tasks. Model serving refers to deploying trained machine learning models into a production environment for prediction and inference. It usually involves loading the model into memory and providing an API or service to receive input data and return the model's prediction results. The goal of model serving is to provide an efficient, scalable, and reliable system to process real-time requests and make predictions on new input data in real time.

[0078] The above-mentioned pre-deployed model pipeline can be to integrate the model pipeline code into the service framework, start and load the model through the script, and complete some inference tasks. Model pipelines refer to combining multiple models together to form a pipeline or chain structure to perform multiple stages of data processing and prediction. The output of each stage serves as the input of the next stage and ultimately generates the final prediction result. Model pipelines can contain multiple different types of models, such as feature extraction models, conversion models, integration models, etc. The goal of model pipelines is to achieve more complex tasks such as feature engineering, data processing, model fusion, etc. by combining multiple models.

[0079] The above-mentioned Model serving focuses on deploying a single model to a production environment for real-time prediction, while Model pipelines focuses on combining multiple models into a pipeline to achieve more complex tasks.

[0080] When deploying a model using the deployment model pipeline, the model can be updated directly within the framework, and the model is transparent to developers, allowing them to directly view the model's current status. It should be noted that the loading process during the model startup phase is somewhat lengthy, and the service's external heartbeat check requires additional judgment. That is, in addition to checking whether the service is running, other judgments are required to ensure service availability, performance, and data consistency, and the model weight file needs to be maintained by the developer.

[0081] When deploying a model using a deployed model service, you can call a model that has less impact on the startup of this service, thereby achieving the purpose of fast calling.

[0082] In the above embodiment of the present disclosure, the method further includes: calling target data processing models corresponding to multiple query requests, processing batch data, and obtaining partial results in the reply results; and outputting the partial results.

[0083] In an optional embodiment, target data processing models corresponding to multiple query requests can be called. When the amount of data is large, the throughput can be improved by grouping multiple query data into batch data. The batch data can be processed by stream processing, so that the user can receive partial results in the reply results during the data processing process, and the partial results can be displayed in the generative dialogue interface to output the partial results. The partial results can also be output in other ways, thereby improving the user experience.

[0084] The above-mentioned stream processing capability can be modified by coordinating the decoding process (decoding) of the model. Currently, the complete result is generally generated first and then returned, which requires the user to wait for a long time. Through stream processing capabilities, partial results can be generated and returned, so that the user can continue to receive the generated results until the complete result is finally received, thereby improving the user experience.

[0085] In another optional embodiment, when implementing the upsampling step, partial results in the reply results can be returned step by step for higher resolution images, so that the user can learn relevant information by viewing partial results while waiting for the generation of all reply results, thereby reducing the user's waiting time and improving the user experience.

[0086] In the above embodiment of the present disclosure, multiple query data are encapsulated to obtain batch data, including: parsing the query request to obtain query data; formatting the query data to obtain converted data; and encapsulating multiple converted data to obtain batch data.

[0087] The format conversion mentioned above refers to converting the query data into a unified format. This can improve the efficiency of data packaging, as the converted data is in a unified format. Format conversion can also be referred to as pre-processing, which is the processing performed before query data is input into the model.

[0088] In an optional embodiment, the format conversion can be a process of converting the query data into matrix data. For example, when the query data is a picture, the picture can be converted into matrix data; when the query data is text, the text can be converted into matrix data. This is only an example, and the specific conversion method can also be set according to the actual usage scenario.

[0089] In an optional embodiment, when the query request is a picture address, the picture address can be parsed first to determine whether the picture can be downloaded from the picture address. If it can be downloaded, it means that the parsing is successful, and the downloaded picture can be the above-mentioned query data. If it cannot be downloaded, it means that the parsing failed. At this time, the query request can be re-parsed or a prompt message indicating that the parsing failed can be fed back so that the user can re-initiate the query request.

[0090] In another optional embodiment, the query request can be parsed first to obtain query data, and the query data can be pre-processed to obtain multiple matrix data, that is, the multiple pre-processed data mentioned above. The multiple matrix data can be combined into one matrix, thereby realizing the process of encapsulating the multiple pre-processed data to obtain batch data.

[0091] In another optional embodiment, the query request can be parsed through the service schema to obtain query data, and then the query data can be pre-processed using the service pre-processor function to obtain pre-processed data. Multiple pre-processed data can be packaged to obtain batch data.

[0092] In the above-mentioned embodiment of the present disclosure, format conversion is performed on the query data to obtain converted data, including: verifying the query data to obtain a verification result of the query data, wherein the verification result is used to indicate whether the query data meets the data processing requirements; if the verification result is that the query data meets the data processing requirements, format conversion is performed on the query data to obtain format-converted data.

[0093] The above-mentioned data processing requirements may be pre-set requirements that the query data needs to meet, for example, when the query data is a picture, requirements for picture size setting, requirements for picture resolution setting, etc. No limitation is imposed on the data processing requirements here.

[0094] In one optional method, the query data can be verified first to determine whether the query data meets the data processing requirements. If the verification result shows that the query data meets the data processing requirements, the query data can be format converted, that is, the query data can be converted into matrix data to obtain the converted data.

[0095] In the above embodiment of the present disclosure, the method further includes: converting the format of the reply result to obtain a converted result; and displaying the converted result in the generative dialogue interface.

[0096] The above-mentioned format conversion can also be used to indicate the conversion of the format of the reply result. Since the format of the converted result is unified, the converted result can be displayed in the generative dialogue interface for easy viewing by the user. The above-mentioned format conversion can be post-processing. During the stream processing process, the post-processing can be to determine whether it is necessary to stop displaying the reply result in advance through placeholders. For example, the output result of the stream processing process is the generated text information and the placeholder containing some special characters. The placeholder can be used to determine whether it is necessary to stop displaying the reply result.

[0097] During batch processing, post-processing can be to determine whether additional information needs to be added to the response result, for example, highlighting or bolding key information, or finding pictures that match the response result and displaying them to the user.

[0098] In an optional embodiment, the reply result can be formatted to add additional information to the reply result or set the display effect of the reply result, so that the converted result can better meet the user's viewing needs, thereby improving the user experience.

[0099] The LLM model and the AIGC model have the same service requirements when deploying services, including: parsing service requests to ensure that data meets standard requirements before entering the model; the pre-processing stage before the data enters the model can form appropriate batch data for model inference; the model inference time is long, and partial results need to be returned in stages to meet user needs; for multimodal data processing, it is basically the same.

[0100] When the LLM model and the AIGC model have the same service requirements during service deployment, the present disclosure proposes the same reasoning service framework, which can ensure that the data formats of both LLM and AIGC are verified to be qualified by setting the pattern type; by setting the interface type to support the basic capabilities of batch processing, stream processing, and step return, it can ensure that the services of LLM and AIGC can be quickly integrated and service-oriented when more models are added; tools can be used to refine multimodal data so that it can be used by various models; for AIGC scenarios, multiple models are generally required to be connected in series to complete a generation task. Through the reasoning service framework provided by the present disclosure, services from multiple models can be connected in series and generated based on the model loading method in the model pipeline.

[0101] Figure 3 is a service framework architecture diagram according to an embodiment of the present disclosure. As shown in Figure 3, multiple query requests are received through the model reasoning service framework interface, and then the multiple query requests are forwarded to the model service layer. The model service layer (LLM / AIGC Service) can provide a service schema (Service Schema), a service pre-processor, a batch / stream processing program, and a service post-processor. Among them, the service schema can be used to parse the query request to obtain query data. The service schema can also verify the query data to obtain a verification result of the query data. When the verification result shows that the query data meets the data processing requirements, the service pre-processor is used to pre-process the query data to obtain pre-processed data, and the batch / stream processing program (Batch / Steaming Handler) is used to encapsulate the multiple pre-processed data to obtain batch data.

[0102] As shown in Figure 3, the service proxy layer (Service Proxy) can provide large-scale pre-trained model service proxy (LLM service proxy), generative artificial intelligence model service proxy (AIGC Service Proxy), embedded service proxy (Embedding Service Proxy), and other service proxies (Other Service Proxys); based on the attribute information of multiple query requests, the target data processing service can be called in the service proxy layer (Service Proxy) to build a target data processing model through the target data processing service.

[0103] As shown in Figure 3, Model Services include large-scale pre-trained model services, generative AI model services, and other model services; Model Pipelines include large-scale pre-trained model pipelines, generative AI model pipelines, and other model pipelines. The target data processing model can be used to process batch data and generate response results. After obtaining the response results, they can be pushed to users through the Elastic Algorithm Service (EAS) inference service platform.

[0104] As shown in Figure 3, tools can be provided to provide additional functionality for the model inference service. The tools include logging, image utils, storage, module builder, and errors.

[0105] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0106] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0107] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present disclosure.

[0108] Example 2

[0109] According to an embodiment of the present disclosure, a data processing method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0110] FIG4 is a flow chart of a data processing method according to Embodiment 2 of the present disclosure. As shown in FIG4 , the method includes the following steps:

[0111] Step S402: responding to multiple query requests received in the generative dialogue interface.

[0112] The query request carries query data, and the query data includes at least one of the following: text and image.

[0113] Step S404: encapsulate the multiple query data to obtain batch data.

[0114] Step S406: calling the target large model corresponding to the multiple query requests, processing the batch data, and obtaining response results corresponding to the multiple query data.

[0115] The aforementioned target large model may refer to a machine learning or deep learning model with a large number of parameters and powerful computing power.

[0116] Step S408: Display the reply result in the generative dialogue interface.

[0117] Through the above steps, multiple query requests received in the generative dialogue interface are responded to, wherein the query requests carry query data; the multiple query data are encapsulated to obtain batch data; the target large model corresponding to the multiple query requests is called to process the batch data to obtain reply results corresponding to the multiple query data; the reply results are displayed in the generative dialogue interface, thereby improving the reasoning efficiency of the multiple query data; it is easy to notice that the multiple query data can be optimized by encapsulating the multiple query data into batch data, and the target data processing model can be called according to the services required by the multiple query requests, so that the target data processing module can quickly process the batch data and obtain reply results corresponding to the multiple query data, thereby improving the processing efficiency of the multiple query data, and thus solving the technical problem of low data processing efficiency in related technologies.

[0118] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0119] Example 3

[0120] According to an embodiment of the present disclosure, a data processing method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0121] FIG5 is a flow chart of a data processing method according to Embodiment 3 of the present disclosure. As shown in FIG5 , the method includes the following steps:

[0122] Step S502: respond to multiple query requests received in the generative dialogue interface.

[0123] The inquiry request carries inquiry data, and the inquiry data at least includes character information.

[0124] The above-mentioned character information may be character information in a text.

[0125] In an optional embodiment, the user may generate the aforementioned multiple query requests by inputting text in the generative dialogue interface, wherein the text may be character information.

[0126] Step S504: encapsulate the multiple query data to obtain batch data.

[0127] Step S506 , calling the target data processing model corresponding to the multiple query requests, processing the batch data, and obtaining response results corresponding to the multiple query data.

[0128] The response type includes one of the following: text, image, video, and voice.

[0129] In an optional embodiment, the target data processing model can process batch data to obtain different types of response results. That is, the user can input text and obtain text, image, and voice response results, thereby improving the user experience.

[0130] Step S508: Display the reply result in the generative dialogue interface.

[0131] Through the above steps, multiple query requests received in the generative dialogue interface are responded to, wherein the query request carries query data, and the query data at least contains character information; the multiple query data are encapsulated to obtain batch data; the target data processing model corresponding to the multiple query requests is called to process the batch data to obtain reply results corresponding to the multiple query data, wherein the type of the reply result includes one of the following: text, image, video and voice; the reply result is displayed in the generative dialogue interface, thereby improving the reasoning efficiency of the multiple query data; it is easy to notice that the multiple query data can be optimized by encapsulating the multiple query data into batch data, and the target data processing model can be called according to the services required by the multiple query requests, so that the target data processing module can quickly process the batch data to obtain reply results corresponding to the multiple query data, thereby improving the processing efficiency of the multiple query data, thereby solving the technical problem of low data processing efficiency in related technologies.

[0132] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0133] Example 4

[0134] According to an embodiment of the present disclosure, a data processing method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0135] FIG6 is a flow chart of a data processing method according to Embodiment 4 of the present disclosure. As shown in FIG6 , the method includes the following steps:

[0136] Step S602: receiving multiple query requests sent by at least one client in a conversation mode.

[0137] The query request carries query data.

[0138] The at least one client mentioned above may be a client used by the user. The user may enter a conversation mode through the client. In the conversation mode, the client may receive multiple query requests sent by the user.

[0139] In an optional embodiment, the user may send a voice message through the client in a conversation mode, thereby generating multiple inquiry requests according to the voice message sent by the user.

[0140] Step S604: encapsulate the multiple query data to obtain batch data.

[0141] Step S606: calling the target data processing model corresponding to the multiple query requests, processing the batch data, and obtaining response results corresponding to the multiple query data.

[0142] Step S608: Send the reply result to the corresponding client, wherein the reply result is displayed by the corresponding client.

[0143] Through the above steps, multiple query requests sent by at least one client in a dialogue mode are received, wherein the query requests carry query data; the multiple query data are encapsulated to obtain batch data; the target data processing model corresponding to the multiple query requests is called to process the batch data to obtain reply results corresponding to the multiple query data; the reply results are sent to the corresponding client, wherein the reply results are displayed by the corresponding client, thereby improving the reasoning efficiency of the multiple query data; it is easy to notice that the multiple query data can be optimized by encapsulating the multiple query data into batch data, and the target data processing model can be called according to the services required by the multiple query requests, so that the target data processing module can quickly process the batch data to obtain reply results corresponding to the multiple query data, thereby improving the processing efficiency of the multiple query data, and thus solving the technical problem of low data processing efficiency in related technologies.

[0144] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0145] Example 5

[0146] According to an embodiment of the present disclosure, a data processing method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0147] FIG7 is a flow chart of a data processing method according to Embodiment 5 of the present disclosure. As shown in FIG7 , the method includes the following steps:

[0148] Step S702: Acquire multiple query requests displayed in the generative dialogue interface in the dialogue mode by calling the first interface.

[0149] The first interface includes a first parameter, a parameter value of the first parameter is a plurality of query requests, and the query requests carry query data.

[0150] The first interface in the above steps may be an interface for data interaction between the cloud server and the client. The client may pass multiple query requests into the interface function as the first parameter of the interface function to achieve the purpose of uploading multiple query requests to the cloud server.

[0151] Step S704: encapsulate the multiple query data to obtain batch data.

[0152] Step S706 : Process the batch data using the target data processing models corresponding to the multiple query requests to obtain response results corresponding to the multiple query data.

[0153] Step S708: Output the reply result by calling the second interface.

[0154] The second interface includes a second parameter, and the parameter value of the second parameter is the reply result.

[0155] The above-mentioned second interface can be an interface for data interaction between the cloud server and the client. The cloud server can pass the reply result into the interface function as the second parameter of the interface function to achieve the purpose of sending the reply result to the client.

[0156] Through the above steps, a first interface is called to obtain multiple query requests displayed in a generative dialogue interface in a dialogue mode, wherein the first interface includes a first parameter, and the parameter value of the first parameter is multiple query requests. The query request carries query data, and the query data includes at least one of the following: text, image, and the multiple query data are encapsulated to obtain batch data. The batch data is processed using a target data processing model corresponding to the multiple query requests to obtain reply results corresponding to the multiple query data, and the reply results are output by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the reply result, thereby improving the reasoning efficiency of the multiple query data. It is easy to notice that the multiple query data can be optimized by encapsulating the multiple query data into batch data, and the target data processing model can be called according to the services required by the multiple query requests, so that the target data processing module can quickly process the batch data and obtain reply results corresponding to the multiple query data, thereby improving the processing efficiency of the multiple query data, thereby solving the technical problem of low data processing efficiency in the related art.

[0157] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0158] Example 6

[0159] According to an embodiment of the present disclosure, a data processing device for implementing the above-mentioned data processing method is also provided. Figure 8 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 6 of the present disclosure. As shown in Figure 8, the device 800 includes: a receiving module 802, an encapsulation module 804, and a calling module 806.

[0160] Among them, the receiving module is configured to respond to multiple query requests received, wherein the query requests carry query data; the encapsulation module is configured to encapsulate the multiple query data to obtain batch data; the calling module is configured to call the target data processing model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data.

[0161] It should be noted that the receiving module 802, encapsulation module 804, and calling module 806 correspond to steps S202 to S206 in Example 1. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above-mentioned modules can also be run as part of the device in the server 10 provided in Example 1.

[0162] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0163] In the above embodiment of the present disclosure, the calling module is also used to call the target data processing service based on the attribute information of multiple query requests, build a target data processing model through the target data processing service, and use the target data processing model to process the batch data to obtain a response result.

[0164] In the above embodiments of the present disclosure, based on the attribute information of multiple query requests, a target proxy object is determined from proxy objects corresponding to multiple data processing models, and a target data processing service is called through the target proxy object.

[0165] In the above embodiments of the present disclosure, the target data processing model is a pre-deployed model or a model in a model pipeline.

[0166] In the above embodiment of the present disclosure, the device further includes: an output module.

[0167] Among them, the calling module is also used to call the target data processing model corresponding to multiple query requests, process the batch data, and obtain partial results in the reply result; the output module is set to display the partial results in the generative dialogue interface.

[0168] In the above embodiment of the present disclosure, the encapsulation module is further used to parse the query request to obtain the query data, perform format conversion on the query data to obtain converted data, and encapsulate multiple converted data to obtain the batch data.

[0169] In the above-mentioned embodiment of the present disclosure, the encapsulation module is further used to verify the query data to obtain a verification result of the query data, wherein the verification result is used to characterize whether the query data meets the data processing requirements. When the verification result is that the query data meets the data processing requirements, the query data is format-converted to obtain the data after the format conversion.

[0170] In the above embodiment of the present disclosure, the display module is further used to convert the format of the reply result to obtain a converted result; and display the converted result in the generative dialogue interface.

[0171] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0172] Example 7

[0173] According to an embodiment of the present disclosure, a data processing device for implementing the above-mentioned data processing method is also provided. Figure 9 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 7 of the present disclosure. As shown in Figure 9, the device 900 includes: a receiving module 902, an encapsulation module 904, a calling module 906, and a display module 908.

[0174] Among them, the receiving module is configured to respond to multiple query requests received in the generative dialogue interface, wherein the query requests carry query data; the encapsulation module is configured to encapsulate multiple query data to obtain batch data; the calling module is configured to call the target large model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data; the display module is configured to display the reply results in the generative dialogue interface.

[0175] It should be noted that the receiving module 902, encapsulation module 904, calling module 906, and display module 908 correspond to steps S402 to S408 in Example 2. The examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above-mentioned modules can also be run as part of the device in the server 10 provided in Example 1.

[0176] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0177] Example 8

[0178] According to an embodiment of the present disclosure, a data processing device for implementing the above-mentioned data processing method is also provided. Figure 10 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 8 of the present disclosure. As shown in Figure 10, the device 1000 includes: a receiving module 1002, an encapsulation module 1004, a calling module 1006, and a display module 1008.

[0179] Among them, the receiving module is configured to respond to multiple query requests received in the generative dialogue interface, wherein the query request carries query data, and the query data at least contains character information; the encapsulation module is configured to encapsulate multiple query data to obtain batch data; the calling module is configured to call the target data processing model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data, wherein the type of the reply result includes one of the following: text, image, video and voice; the display module is configured to display the reply result in the generative dialogue interface.

[0180] It should be noted that the receiving module 1002, encapsulation module 1004, calling module 1006, and display module 1008 correspond to steps S502 to S508 in Example 3. The examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above-mentioned modules can also be run as part of the device in the server 10 provided in Example 1.

[0181] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0182] Example 9

[0183] According to an embodiment of the present disclosure, a data processing device for implementing the above-mentioned data processing method is also provided. Figure 11 is a schematic diagram of an accelerated deployment device for a deep learning model according to Example 9 of the present disclosure. As shown in Figure 11, the device 1100 includes: a receiving module 1102, an encapsulation module 1104, a calling module 1106, and a sending module 1108.

[0184] Among them, the receiving module is configured to receive multiple query requests sent by at least one client in a dialogue mode, wherein the query requests carry query data; the encapsulation module is configured to encapsulate multiple query data to obtain batch data; the calling module is configured to call the target data processing model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data; the sending module is configured to send the reply results to the corresponding client, wherein the reply results are displayed by the corresponding client.

[0185] It should be noted that the receiving module 1102, encapsulation module 1104, calling module 1106, and sending module 1108 correspond to steps S602 to S608 in Example 4. The examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above-mentioned modules can also be run as part of the device in the server 10 provided in Example 1.

[0186] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0187] Example 10

[0188] According to an embodiment of the present disclosure, a data processing device for implementing the above-mentioned data processing method is also provided. Figure 12 is a schematic diagram of an accelerated deployment device for a deep learning model according to embodiment 10 of the present disclosure. As shown in Figure 12, the device 1200 includes: a calling module 1202, an encapsulation module 1204, a processing module 1206, and an output module 1208.

[0189] Among them, the calling module is configured to obtain multiple query requests displayed in the generative dialogue interface in the dialogue mode by calling the first interface, wherein the first interface includes a first parameter, the parameter value of the first parameter is multiple query requests, and the query request carries query data; the encapsulation module is configured to encapsulate multiple query data to obtain batch data; the processing module is configured to use the target data processing model corresponding to the multiple query requests to process the batch data to obtain reply results corresponding to the multiple query data; the output module is configured to output the reply result by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the reply result.

[0190] It should be noted that the calling module 1202, encapsulation module 1204, processing module 1206, and output module 1208 correspond to steps S702 to S708 in Example 5. The examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above modules can also be run as part of the device in the server 10 provided in Example 1.

[0191] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0192] Example 11

[0193] The embodiment of the present disclosure may provide a computer terminal, which may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0194] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0195] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the data processing method: responding to multiple query requests received, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling the target data processing models corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data.

[0196] Optionally, Figure 13 is a block diagram of a computer terminal according to an embodiment of the present disclosure. As shown in Figure 13, the computer terminal A may include: one or more (only one is shown in the figure) processors 102, a memory 104, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0197] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the data processing method and device in the embodiments of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned data processing method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories may be connected to the terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0198] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: respond to multiple received query requests, where the query requests carry query data; encapsulate the multiple query data to obtain batch data; call the target data processing models corresponding to the multiple query requests, process the batch data, and obtain response results corresponding to the multiple query data.

[0199] Optionally, the processor may also execute the program code of the following steps: calling the target data processing service based on the attribute information of multiple query requests; building a target data processing model through the target data processing service; and processing the batch data using the target data processing model to obtain a response result.

[0200] Optionally, the processor may further execute program code of the following steps: determining a target proxy object from proxy objects corresponding to multiple data processing models based on attribute information of multiple query requests; and calling a target data processing service through the target proxy object.

[0201] Optionally, the processor may further execute program code of the following steps: the target data processing model is a pre-deployed model or a model in a model pipeline.

[0202] Optionally, the processor may also execute the program code of the following steps: calling target data processing models corresponding to multiple query requests, processing the batch data, and obtaining partial results in the reply result; and outputting the partial results.

[0203] Optionally, the processor may further execute program codes of the following steps: parsing the query request to obtain the query data; performing format conversion on the query data to obtain converted data; and encapsulating a plurality of the converted data to obtain the batch data.

[0204] Optionally, the processor may also execute the program code of the following steps: verifying the query data to obtain a verification result of the query data, wherein the verification result is used to indicate whether the query data meets the data processing requirements; and if the verification result is that the query data meets the data processing requirements, performing format conversion on the query data to obtain the data after the format conversion.

[0205] Optionally, the processor may also execute the program code of the following steps: performing format conversion on the reply result to obtain a converted result; and displaying the converted result in a generative dialogue interface.

[0206] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: respond to multiple query requests received in the generative dialogue interface, wherein the query request carries query data; encapsulate the multiple query data to obtain batch data; call the target large model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data; and display the reply results in the generative dialogue interface.

[0207] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: respond to multiple query requests received in the generative dialogue interface, wherein the query request carries query data, and the query data at least contains character information; encapsulate the multiple query data to obtain batch data; call the target data processing model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data, wherein the type of the reply result includes one of the following: text, image, video and voice; and display the reply result in the generative dialogue interface.

[0208] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: receive multiple query requests sent by at least one client in a dialogue mode, wherein the query requests carry query data; encapsulate the multiple query data to obtain batch data; call the target data processing model corresponding to the multiple query requests, process the batch data, and obtain reply results corresponding to the multiple query data; send the reply results to the corresponding client, wherein the reply results are displayed by the corresponding client.

[0209] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain multiple query requests displayed in the generative dialogue interface in the dialogue mode by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is multiple query requests, and the query request carries query data; encapsulate the multiple query data to obtain batch data; use the target data processing model corresponding to the multiple query requests to process the batch data to obtain reply results corresponding to the multiple query data; output the reply result by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the reply result.

[0210] According to an embodiment of the present disclosure, a plurality of query requests are received, wherein the query requests carry query data; the plurality of query data are encapsulated to obtain batch data; the target data processing models corresponding to the plurality of query requests are called to process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the inference efficiency of the plurality of query data; it is easy to notice that the plurality of query data can be optimized by encapsulating the plurality of query data into batch data, and the target data processing model can be called according to the services required by the plurality of query requests, so that the target data processing module can quickly process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the processing efficiency of the plurality of query data, and further solving the technical problem of low data processing efficiency in related technologies.

[0211] Those skilled in the art will appreciate that the structure shown in FIG13 is merely illustrative, and the computer terminal may also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile internet device (MID), a PAD, or other terminal device. FIG13 does not limit the structure of the above-mentioned electronic devices. For example, the computer terminal A may include more or fewer components (such as a network interface, a display device, etc.) than those shown in FIG13, or may have a configuration different from that shown in FIG13.

[0212] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0213] Example 12

[0214] The embodiment of the present disclosure further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the data processing method provided in the first embodiment.

[0215] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0216] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: responding to multiple query requests received, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling the target data processing models corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data.

[0217] Optionally, the above-mentioned storage medium is also configured to store program code for executing the following steps: calling the target data processing service based on the attribute information of multiple query requests; constructing a target data processing model through the target data processing service; and processing the batch data using the target data processing model to obtain a response result.

[0218] Optionally, the storage medium is further configured to store program code for executing the following steps: determining a target proxy object from proxy objects corresponding to multiple data processing models based on attribute information of multiple query requests; and calling a target data processing service through the target proxy object.

[0219] Optionally, the above-mentioned storage medium is further configured to store program codes for executing the following steps: the target data processing model is a pre-deployed model, or a model in a model pipeline.

[0220] Optionally, the storage medium is further configured to store program codes for executing the following steps: calling target data processing models corresponding to multiple query requests, processing the batch data, and obtaining partial results in the reply results; and outputting the partial results.

[0221] Optionally, the storage medium is further configured to store program codes for executing the following steps: parsing the query request to obtain the query data; formatting the query data to obtain converted data; and encapsulating multiple converted data to obtain the batch data.

[0222] Optionally, the storage medium is further configured to store program code for executing the following steps: verifying the query data to obtain a verification result of the query data, wherein the verification result is used to characterize whether the query data meets the data processing requirements; and if the verification result is that the query data meets the data processing requirements, format converting the query data to obtain the format-converted data.

[0223] Optionally, the storage medium is further configured to store program codes for executing the following steps: performing format conversion on the reply result to obtain a converted result; and displaying the converted result in a generative dialogue interface.

[0224] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: responding to multiple query requests received in the generative dialogue interface, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling the target large model corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data; and displaying the reply results in the generative dialogue interface.

[0225] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: responding to multiple query requests received in a generative dialogue interface, wherein the query request carries query data, and the query data contains at least character information; encapsulating the multiple query data to obtain batch data; calling a target data processing model corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data, wherein the type of the reply result includes one of the following: text, image, video, and voice; and displaying the reply result in the generative dialogue interface.

[0226] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: receiving multiple query requests sent by at least one client in a dialogue mode, wherein the query requests carry query data; encapsulating the multiple query data to obtain batch data; calling the target data processing model corresponding to the multiple query requests, processing the batch data, and obtaining reply results corresponding to the multiple query data; sending the reply results to the corresponding client, wherein the reply results are displayed by the corresponding client.

[0227] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: obtaining multiple query requests displayed in a generative dialogue interface in a dialogue mode by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is multiple query requests, and the query requests carry query data; encapsulating the multiple query data to obtain batch data; processing the batch data using a target data processing model corresponding to the multiple query requests to obtain reply results corresponding to the multiple query data; outputting the reply result by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the reply result.

[0228] According to an embodiment of the present disclosure, a plurality of query requests are received, wherein the query requests carry query data; the plurality of query data are encapsulated to obtain batch data; the target data processing models corresponding to the plurality of query requests are called to process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the inference efficiency of the plurality of query data; it is easy to notice that the plurality of query data can be optimized by encapsulating the plurality of query data into batch data, and the target data processing model can be called according to the services required by the plurality of query requests, so that the target data processing module can quickly process the batch data to obtain reply results corresponding to the plurality of query data, thereby improving the processing efficiency of the plurality of query data, and further solving the technical problem of low data processing efficiency in related technologies.

[0229] The serial numbers of the above-mentioned embodiments of the present disclosure are for description only and do not represent the advantages or disadvantages of the embodiments.

[0230] In the above embodiments of the present disclosure, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0231] In the several embodiments provided in the present disclosure, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0232] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0233] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0234] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0235] The above is only a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as within the scope of protection of the present disclosure.

Claims

1. A data processing method, comprising: Responding to a plurality of received query requests, wherein the query requests carry query data; Encapsulating a plurality of the query data to obtain batch processing data; The target data processing models corresponding to the plurality of query requests are called to process the batch data to obtain response results corresponding to the plurality of query data.

2. The method according to claim 1, wherein: Calling a target data processing model corresponding to the plurality of query requests to process the batch data to obtain a response result corresponding to the plurality of query data, including: Based on the attribute information of the plurality of query requests, calling a target data processing service; Building the target data processing model through the target data processing service; The target data processing model is used to process the batch data to obtain the response result.

3. The method according to claim 2, wherein: Based on the attribute information of the plurality of query requests, calling a target data processing service includes: Determining a target proxy object from proxy objects corresponding to a plurality of data processing models based on the attribute information of the plurality of query requests; The target data processing service is called through the target proxy object.

4. The method according to claim 2, wherein: The target data processing model is a pre-deployed model or a model in a model pipeline.

5. The method according to claim 1, wherein: The method further comprises: Calling target data processing models corresponding to the plurality of query requests to process the batch data to obtain partial results of the reply result; Output the partial results.

6. The method according to claim 1, wherein: The plurality of query data are packaged to obtain batch data, including: Parsing the query request to obtain the query data; Performing format conversion on the query data to obtain converted data; The plurality of converted data are packaged to obtain the batch processed data.

7. The method according to claim 6, wherein: The query data is formatted to obtain converted data, including: Verifying the query data to obtain a verification result of the query data, wherein the verification result is used to indicate whether the query data meets the data processing requirement; If the verification result is that the query data meets the data processing requirement, The data is format-converted to obtain the data after the format conversion.

8. The method according to claim 1, wherein: The method further comprises: Performing format conversion on the reply result to obtain a converted result; The converted result is displayed in a generative dialogue interface.

9. A data processing method, wherein: include: Responding to multiple query requests received in the generative dialogue interface, wherein the query requests carry query data; Encapsulating a plurality of the query data to obtain batch processing data; Calling a target large model corresponding to a plurality of the query requests, processing the batch data, and obtaining a response result corresponding to a plurality of the query data; The response result is displayed in the generative dialogue interface.

10. The method according to claim 9, wherein: Calling multiple target large models corresponding to the query requests, The batch data is processed to obtain a plurality of response results corresponding to the query data, including: Based on the attribute information of the plurality of query requests, calling a target data processing service; Constructing the target macro model through the target data processing service; The batch data is processed using the target large model to obtain the response result.

11. The method according to claim 10, wherein: Based on the attribute information of the plurality of query requests, calling a target data processing service includes: Determining a target proxy object from proxy objects corresponding to a plurality of data processing models based on the attribute information of the plurality of query requests; The target data processing service is called through the target proxy object.

12. A data processing method, wherein: include: Responding to multiple query requests received in the generative dialogue interface, wherein the query requests carry query data, and the query data at least includes character information; Encapsulating a plurality of the query data to obtain batch processing data; Calling a target data processing model corresponding to a plurality of the query requests to process the batch data to obtain a plurality of reply results corresponding to the query data, wherein the type of the reply result includes one of the following: text, image, video and voice; The response result is displayed in the generative dialogue interface.

13. The method according to claim 12, wherein: Calling a target data processing model corresponding to the plurality of query requests to process the batch data to obtain a response result corresponding to the plurality of query data, including: Based on the attribute information of the plurality of query requests, calling a target data processing service; Building the target data processing model through the target data processing service; The target data processing model is used to process the batch data to obtain the response result.

14. The method according to claim 13, wherein: Based on the attribute information of the plurality of query requests, calling a target data processing service includes: Determining a target proxy object from proxy objects corresponding to a plurality of data processing models based on the attribute information of the plurality of query requests; The target data processing service is called through the target proxy object.

15. A data processing method, wherein: include: Receiving a plurality of query requests sent by at least one client in a conversation mode, wherein the query requests carry query data; Encapsulating a plurality of the query data to obtain batch processing data; Calling a target data processing model corresponding to a plurality of the query requests to process the batch data to obtain a response result corresponding to a plurality of the query data; The reply result is sent to a corresponding client, wherein the reply result is displayed by the corresponding client.

16. The method according to claim 15, wherein: Calling a target data processing model corresponding to the plurality of query requests to process the batch data to obtain a response result corresponding to the plurality of query data, including: Based on the attribute information of the plurality of query requests, calling a target data processing service; Building the target data processing model through the target data processing service; The target data processing model is used to process the batch data to obtain the response result.

17. A data processing method, wherein: include: Acquiring multiple query requests displayed in a generative dialog interface in a dialog mode by calling a first interface, wherein the first interface includes a first parameter, a parameter value of the first parameter is multiple query requests, and the query request carries query data; Encapsulating a plurality of the query data to obtain batch processing data; Processing the batch data using target data processing models corresponding to the plurality of query requests to obtain response results corresponding to the plurality of query data; The reply result is outputted by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter is the reply result.

18. The method according to claim 17, wherein: The batch data is processed using a target data processing model corresponding to the plurality of query requests to obtain a response result corresponding to the plurality of query data, including: Based on the attribute information of the plurality of query requests, calling a target data processing service; Building the target data processing model through the target data processing service; The target data processing model is used to process the batch data to obtain the response result.

19. An electronic device, wherein: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 18 when running.

20. A computer-readable storage medium, wherein: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 18.

Images