Data processing method and device

Abstract

The invention discloses a data processing method and device.The method comprises the steps that in response to an obtained target processing task, a processing model corresponding to the target processing task is determined; target task data and target model data are transmitted to target electronic equipment, so that the target electronic equipment processes the target task data through the target model data, and the target model data are at least part of model data of the processing model. The target task data is data obtained by compiling the target processing task by the first electronic equipment.

Description

Technical Field

[0001] The present application relates to but is not limited to the field of computer technology, and in particular to a data processing method and device. Background Art

[0002] With the development of artificial intelligence (AI) technology, various applications for realizing AI functions are increasingly appearing in users' daily lives and work. For example, text-to-text models, text-to-graph models, and graph-to-graph models in generative large models. However, large models have the characteristics of large parameter file data volume, so the hardware and software computing resources of electronic devices running large models are required to be high.

[0003] In this way, when the computing power of the local electronic device is insufficient, it is necessary to call on the computing power resources of other electronic devices to run the large model. Therefore, how to efficiently use the computing resources of local electronic devices and other electronic devices to efficiently use the large model has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, the present application at least provides a data processing method and device.

[0005] The technical solution of this application is implemented as follows:

[0006] On the one hand, the present application provides a data processing method, comprising:

[0007] In response to obtaining a target processing task, determining a processing model corresponding to the target processing task;

[0008] The target task data and the target model data are transmitted to the target electronic device so that the target electronic device processes the target task data using the target model data; the target model data is at least part of the model data of the processing model; the target task data is the data obtained by the first electronic device compiling the target processing task.

[0009] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, the data processing method further includes:

[0010] Obtaining configuration information of the target electronic device, and using model data in a model data set obtained by parsing the processing model and matching the configuration information as target model data;

[0011] The configuration information includes at least one of the following:

[0012] Model configuration information of the target electronic device;

[0013] Model usage information of the target electronic device;

[0014] Resource usage of the target electronic device;

[0015] Network information of the target electronic device.

[0016] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, the data processing method further includes:

[0017] The model data in the storage area of ​​the first electronic device is obtained, and the model data in the storage area is compared with the model data set obtained by parsing the processing model to obtain first model difference data, and the first model difference data is used as the target model data.

[0018] In some embodiments, when the target task data and the target model data are transmitted to the target electronic device, the data processing method further includes:

[0019] Storing the target model data in a storage area of ​​the first electronic device;

[0020] The target model data includes at least one of the following:

[0021] All model data obtained by processing model analysis;

[0022] The first model difference data is obtained by comparing the model data in the storage area with the model data set.

[0023] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, the data processing method further includes:

[0024] The model data in the cache area of ​​the target electronic device is obtained, and the model data in the cache area is compared with the model data set obtained by parsing the processing model to obtain second model difference data, and the second model difference data is used as the target model data.

[0025] In some embodiments, the target task data includes a first portion of task data and a second portion of task data;

[0026] The data processing method also includes:

[0027] Using the model data that does not match the configuration information in the model data set obtained by parsing the processing model as the first model data, so that the first electronic device processes the first part of the task data using the first model data;

[0028] Using second model data matching the configuration information in the model data set obtained by parsing the processing model as target model data;

[0029] The target task data and the target model data are transmitted to the target electronic device, including: transmitting the target model data and the second part of the task data to the target electronic device, so that the target electronic device processes the second part of the task data using the target model data.

[0030] In some embodiments, transmitting the target task data and the target model data to the target electronic device includes:

[0031] Based on the reference data corresponding to the target processing task, the target model data is packaged to generate first packet header data;

[0032] Package the target task data to generate package data;

[0033] Transmitting data consisting of the first packet header data and the packet body data to a target electronic device;

[0034] The reference data includes at least one of the following:

[0035] The number of electronic devices targeted;

[0036] The device identification of the target electronic device;

[0037] A hash of the target model data.

[0038] On the other hand, the present application also provides a data processing method, applied to a target electronic device, comprising:

[0039] Obtain target task data and target model data corresponding to the target processing task;

[0040] The target task data is processed by the target model data using the hardware resources of the target electronic device to obtain the task result data.

[0041] In some embodiments, the target task data corresponding to the target processing task is processed by the target model data using the hardware resources of the target electronic device to obtain the task result data, further comprising:

[0042] Based on the reference data corresponding to the target processing task, the target processing task is executed using the hardware resources of multiple target electronic devices; or,

[0043] Based on the reference data corresponding to the target processing task, the target processing task is executed using multiple computing units of the target electronic device.

[0044] On the other hand, the present application also provides a data processing device, including:

[0045] a determining unit, configured to determine a processing model corresponding to the target processing task in response to obtaining the target processing task;

[0046] A transmission unit is used to transmit target task data and target model data to a target electronic device so that the target electronic device processes the target task data using the target model data. The target model data is at least part of the model data of the processing model, and the target task data is data obtained by compiling the target processing task by the first electronic device.

[0047] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0049] Figure 1 A schematic diagram of the implementation flow of a data processing method provided in this application;

[0050] Figure 2 A schematic diagram of the implementation flow of another data processing method provided in this application;

[0051] Figure 3 A schematic diagram of an implementation flow of an embodiment of a data processing method provided by the present application on a first electronic device side;

[0052] Figure 4 A schematic diagram of an implementation flow of an embodiment of a data processing method provided by the present application on a target electronic device side;

[0053] Figure 5 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application;

[0054] Figure 6 A schematic diagram of the structure of another data processing device provided in an embodiment of the present application;

[0055] Figure 7 A hardware entity schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions of the present application are further elaborated in detail below in conjunction with the drawings and embodiments. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0057] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0058] The terms "first / second / third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first / second / third" can be interchanged with a specific order or sequence where permitted so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing this application and are not intended to limit this application.

[0060] In order to better understand the data processing method provided by the embodiment of the present application, the solution adopted in the related art in which the local electronic device calls the computing resources of the server electronic device to run the large model is first explained below.

[0061] In the related art, the local end intercepts the local graphics card function by rewriting the local dynamic link library (DLL) and dispatches it to the server graphics card for execution. However, this solution has the following disadvantages:

[0062] First, the local central processing unit (CPU) parsing function and the server graphics processing unit (GPU) executing function are serial execution processes, resulting in low parallelism between the local CPU and the server GPU, and the problem of idle CPU and GPU resources;

[0063] Secondly, each call to the server GPU requires the transmission of the heterogeneous graphics card call function code from the local end to the server end, resulting in significant transmission overhead. At the same time, when the amount of data transmitted is large, the data transmission between the local end and the server end may be limited by the network bandwidth.

[0064] Thirdly, it is not possible to flexibly configure the calls of heterogeneous graphics card functions based on local and server computing resources, network resources, and user requirements.

[0065] Based on this, the present application provides a data processing method applied to a first electronic device, which can be executed by a processor of the first electronic device. The first electronic device may refer to a server, a laptop, a tablet computer, a desktop computer, a smart TV, a set-top box, a mobile device (such as a mobile phone, a portable video player, a personal digital assistant, a dedicated messaging device, a portable gaming device) and other devices with data processing capabilities.

[0066] Figure 1 A schematic diagram of the implementation flow of a data processing method provided in this application, such as Figure 1 As shown, the method includes the following steps S101 to S102:

[0067] Step S101, in response to obtaining a target processing task, determining a processing model corresponding to the target processing task.

[0068] Here, the target processing task refers to a model reasoning task that needs to be processed by a model, including but not limited to text-to-image tasks, text-to-text tasks, or image-to-image tasks.

[0069] The processing model refers to a model that can be used to perform the target processing task. For example, when the target processing task is a text-to-text task, the processing model can be a Stable Disffusion (SD) series, a Large Language Model (LLM), etc.; when the target processing task is a text-to-text task, the processing model can be an LLM; and so on.

[0070] In some embodiments, the processing model refers to the latest updated model used in the first electronic device to process the target processing task, including all updated model parameter files. For example, when using LLM to perform a task of a Vincent diagram, the processing model refers to the latest updated LLM.

[0071] Step S102, transmitting the target task data and the target model data to the target electronic device, so that the target electronic device processes the target task data using the target model data; the target model data is at least part of the model data of the processing model; the target task data is the data obtained by compiling the target processing task by the first electronic device.

[0072] Here, after the first electronic device determines the target processing task and the processing model, it further determines the target model data based on the processing model, and transmits the target task data and the target model data to the target electronic device, so that the target electronic device processes the target task data using the target model data.

[0073] In some embodiments, the first electronic device sends all model data of the processing model to the target electronic device as target model data.

[0074] In some embodiments, at least part of the model data of the processing model is pre-stored in the target electronic device. In this way, the first electronic device sends part of the model data of the processing model that is not stored in the target electronic device as the target model data to the target electronic device based on the part of the model data of the processing model stored in the target electronic device.

[0075] Here, the target task data is data obtained by compiling the target processing task by the first electronic device, that is, before sending the specific inference task to the target electronic device, the first electronic device pre-compiles the target processing task.

[0076] In the data processing method applied to the first electronic device provided by the present application, first, in response to obtaining the target processing task, the processing model corresponding to the target processing task is determined; then, the target task data and the target model data are transmitted to the target electronic device, so that the target electronic device processes the target task data using the target model data, wherein the target model data is at least part of the model data of the processing model, and the target task data is the data obtained by the first electronic device compiling the target processing task. In this way, on the one hand, when the first electronic device obtains the target processing task that needs to be completed using the processing model, and needs to call the target electronic device to assist in executing the target processing task, the first electronic device can send at least part of the model data of the processing model to the target electronic device, thereby realizing the use of the target electronic device to assist in executing the model reasoning task; on the other hand, the first electronic device pre-compiles the target processing task, and sends the compiled target task data to the target electronic device, which can save the target electronic device from compiling the target processing task, thereby improving the processing efficiency of the target processing task.

[0077] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, that is, before the above step S102, the data processing method further includes the following step S103:

[0078] Step S103, obtaining configuration information of the target electronic device, and using model data in a model data set obtained by parsing the processing model and matching the configuration information as the target model data;

[0079] The configuration information includes at least one of the following:

[0080] Model configuration information of the target electronic device;

[0081] Model usage information of the target electronic device;

[0082] The resource usage of the target electronic device;

[0083] The network information of the target electronic device.

[0084] Here, parsing the processing model refers to parsing the model file of the processing model to obtain data such as the input parameter name, input parameter type, input parameter order, value range, vector dimension, etc. of the processing model.

[0085] In some embodiments, parsing the processing model further includes the first electronic device parsing the model file, model parameters and reasoning logic instructions of the processing model into function codes applicable to the computing module (e.g., GPU) of the target electronic device according to the configuration information of the target electronic device, that is, the function codes of the instruction set corresponding to the computing module of the target electronic device; wherein the configuration information of the target electronic device at least includes the instruction set information applicable to the computing module of the target electronic device. In this way, the model data set corresponding to the processing model is the model data that can be recognized by the computing module of the target electronic device.

[0086] The model data matching the configuration information of the target electronic device refers to the model data determined based on the configuration information of the target electronic device and called to be executed by the target electronic device.

[0087] In some embodiments, the first electronic device may call monitoring services provided by the local and target electronic devices to obtain configuration information. For example, the resource monitoring module in the first electronic device obtains detailed configuration information of the local and target electronic devices by registering and connecting to the network addresses of the local and target electronic devices.

[0088] In some embodiments, the detailed configuration information may include the CPU model of the first electronic device and the target electronic device, resource usage, GPU model, driver version, GPU cache occupancy, CUDA (Compute Unified Device Architecture, CUDA) function usage, network information, model configuration information of the target electronic device, model usage, etc.

[0089] Here, the model configuration information of the target electronic device refers to the model information stored in the target electronic device; it may also refer to the model information loaded into the cache of the target electronic device.

[0090] In this way, when the model configuration information represents that the target electronic device has stored or loaded part of the model data of the processing model, the part of the model data in the model data set that is not stored or loaded in the target electronic device is used as the target model data; if the target electronic device does not store or load any model data of the processing model, the entire model data set is used as the target model data.

[0091] The model usage information of the target electronic device refers to the situation where the model loaded into the cache of the target electronic device is used or occupied. For example, after the target electronic device loads the large language model into the cache, it uses the large language model to perform related model reasoning, and then the large language model is in an occupied state.

[0092] In this way, when part or all of the model data in the model data set of the model characterizing the processing model in the model usage has been loaded into the cache of the target electronic device and another model inference task is being processed, the entire model data set is used as the target model data; when part of the model data of the model data set characterizing the model usage has been loaded into the cache of the target electronic device and is in an idle state, part of the model data in the model data set that has not been loaded into the cache of the target electronic device is used as the target model data.

[0093] The resource usage of the target electronic device refers to the usage of computing resources in the target electronic device. For example, the first electronic device can obtain the total amount of computing resources of the target electronic device, the total amount of computing resources currently in use, and the total amount of tasks to be processed in the task queue of the target electronic device, thereby calculating the resource usage of the target electronic device.

[0094] In this way, the first electronic device can determine the amount of target model data to be sent to the target electronic device based on the resource occupancy of the target electronic device. For example, when the resource occupancy represents that the idle computing resources of the target electronic device are relatively large, at least part of the model data with a relatively large amount of data can be determined from the model data set as the target model data; correspondingly, when the resource occupancy represents that the idle computing resources of the target electronic device are relatively small, part of the model data with a relatively small amount of data can be determined from the model data set as the target model data.

[0095] The network information of the target electronic device refers to the network connection status between the target electronic device and the first electronic device.

[0096] In some embodiments, the target electronic device communicates with the first electronic device via a mobile hotspot (Wi-Fi) or Ethernet's Transmission Control Protocol / Internet Protocol (TCP / IP).

[0097] In some embodiments, the network information includes information such as network connection speed, network latency, and bandwidth utilization.

[0098] In this way, when the network information represents that the network connection speed between the target electronic device and the first electronic device is slow, the network delay is large, or the bandwidth utilization rate is low, part of the model data with a smaller data volume in the model data set is used as the target model data; conversely, when the network information represents that the network connection speed between the target electronic device and the first electronic device is fast, the network delay is small, and the bandwidth utilization rate is high, at least part of the model data with a larger data volume in the model data set is used as the target model data.

[0099] In some embodiments, when the network information indicates that the network connection between the target electronic device and the first electronic device is extremely poor, the invocation of the computing module of the target electronic device to assist in executing the target processing task is canceled, or a new target electronic device is re-determined.

[0100] In the above embodiment, by obtaining the configuration information of the target electronic device, the first electronic device can dynamically adjust the calling method of the hardware resources of the target electronic device according to the real-time performance of the target electronic device and the local task processing requirements, and flexibly configure various parameters in the process of calling the target electronic device. Including but not limited to, the first electronic device can configure whether to call the target electronic device, whether to enable local GPU resources, and whether to perform parsing processing on the processing model and the target processing task in advance.

[0101] In some embodiments, the first electronic device determines whether to start the preheating mechanism based on the configuration information of the local and target electronic devices. For example, based on the configuration information, the first electronic device determines to call the target electronic device to perform the target processing task, and the network condition between the first electronic device and the target electronic device is good, and the first electronic device has sufficient computing resources to perform the preheating process, then the first electronic device determines to start the preheating mechanism; otherwise, the preheating mechanism is not started.

[0102] In some embodiments, under the preheating mechanism, the first electronic device performs compilation processing, packaging processing, compression processing, etc. on the model data of the processing model.

[0103] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, that is, before the above step S102, the data processing method provided by the present application further includes the following step S104:

[0104] Step S104, obtaining model data in a storage area of ​​a first electronic device, comparing the model data in the storage area with a model data set obtained by parsing the processing model to obtain first model difference data, and using the first model difference data as target model data.

[0105] Here, the storage area refers to an area in a storage device such as a memory, a cache or a hard disk of the first electronic device.

[0106] In some embodiments, after the first electronic device sends the model data to the target electronic device, the first electronic device stores the sent model data in a storage area of ​​the first electronic device.

[0107] In some embodiments, after the first electronic device compiles the model, it stores the compiled static resources and some dynamic resources in a storage area; wherein the static resources refer to the neural network data and parameter data of the model, and the dynamic resources refer to the logical instructions in the model reasoning process, and the partial dynamic resources stored in the storage area here refer to the function body in the logical instructions.

[0108] In this way, when the first electronic device determines the target model data, it can read the model data in the storage area and compare the model data in the storage area with the model data set to obtain first model difference data, that is, at least part of the model data in the model data set of the processing model that has not been sent to the target electronic device, and then send the first model difference data as the target model data to the target electronic device.

[0109] In some embodiments, when the target task data and the target model data are transmitted to the target electronic device, that is, when the first electronic device executes the above step S102, the following step S105 is also executed:

[0110] Step S105, storing the target model data in a storage area of ​​the first electronic device;

[0111] The target model data includes at least one of the following:

[0112] All model data obtained by parsing the processing model;

[0113] The first model difference data is obtained by comparing the model data in the storage area with the model data set.

[0114] Here, when sending the target model data to the target electronic device, the first electronic device stores the target model data locally so that when subsequently sending the processing model or updated model data of the processing model to the target electronic device, the stored target model data can be referred to to determine new target model data.

[0115] In some embodiments, when the first electronic device determines based on the model configuration information of the target electronic device that the target electronic device has loaded all model data of the processing model in the cache and the processing model is in an occupied state, the first electronic device will use all model data parsed from the processing model as target model data.

[0116] In some embodiments, when the first electronic device has stored partial model data obtained by parsing the processing model in a local storage area, the first electronic device uses first model difference data obtained by comparing the model data in the storage area with the model data set as target model data; when the first electronic device does not store any model data obtained by parsing the processing model in the local storage area, the first electronic device uses all model data obtained by parsing the processing model as target model data.

[0117] In some embodiments, before transmitting the target task data and the target model data to the target electronic device, that is, before the above step S102, the data processing method provided by the present application further includes the following step S106:

[0118] Step S106, obtaining model data in a cache area of ​​a target electronic device, comparing the model data in the cache area with a model data set obtained by parsing the processing model to obtain second model difference data, and using the second model difference data as target model data.

[0119] Here, first, the first electronic device obtains the model data in the cache area of ​​the target electronic device to determine the model data of the model currently in operation in the target electronic device;

[0120] Then, the first electronic device compares the model data obtained from the target electronic device with the model data set obtained by parsing in the first electronic device to obtain second model difference data, that is, to determine the model data in the model data set that has not been loaded into the cache area of ​​the target electronic device;

[0121] Afterwards, the first electronic device uses the second model difference data as the target model data to send the second model difference data to the target electronic device, and the target electronic device loads the second model difference data into the cache area.

[0122] In the above embodiment, the first electronic device reads the model data in the cache area of ​​the target electronic device, and determines the target model data based on the model data in the cache area. In this way, the target electronic device can directly call the part of the model data loaded in the cache area of ​​the target electronic device; at the same time, the first electronic device sends the target model data that is not loaded in the cache area of ​​the target electronic device to the target electronic device, so that the target electronic device can directly load the target model data into the cache area and call the target model data. In this way, when the amount of target model data is large, the speed of directly obtaining the target model data from the first electronic device and loading it into the cache area will be faster than the speed of loading the target model data from the local of the target electronic device, thereby shortening the time for the target electronic device to load the target model data and improving the processing efficiency of the target processing task.

[0123] In some embodiments, the target task data includes a first portion of task data and a second portion of task data.

[0124] Here, according to the processing order of the processing model for the target processing task, the target task data of the target processing task is divided into at least a first part of task data and a second part of task data. For example, when using a large language model to perform a text-generated image task, the text-generated image task can be divided into two parts of tasks: the first part of the task is a text-generated image task for expanding the text input by the user; the second part of the task is a text-generated image task for generating an image based on the expanded text data.

[0125] In this way, when the first electronic device has available computing resources, the model data set of the processing model can be split according to the amount of computing resources of the first electronic device and the operating logic of the processing model, so that the first electronic device and the target electronic device can respectively execute part of the model data of the processing model, thereby further improving the processing efficiency of the target processing task.

[0126] Thus, before transmitting the target task data and the target model data to the target electronic device, that is, before the above step S102, the data processing method provided by the present application further includes the following steps S107 to S108:

[0127] Step S107, using the model data that does not match the configuration information in the model data set obtained by parsing the processing model as first model data, so that the first electronic device processes the first part of the task data using the first model data;

[0128] Step S108: taking the second model data matching the configuration information in the model data set obtained by parsing the processing model as the target model data.

[0129] Here, based on the acquired configuration information of the target electronic device, model data executed locally (ie, first model data) and model data sent to the target electronic device (ie, second model data or target model data) are determined from the model data set of the processing model.

[0130] Model data that matches the configuration information of the target electronic device refers to model data that needs to be sent to the target electronic device for execution based on the configuration information of the target electronic device; correspondingly, model data that does not match the configuration information of the target electronic device refers to model data that needs to be executed locally based on the configuration information of the target electronic device.

[0131] For example, based on the model configuration information in the configuration information of the target electronic device, it is determined that part of the model data in the model data set has been loaded in the cache area of ​​the target electronic device, and the second model data related to the part of the model data loaded by the target electronic device can be used as the model data matching the configuration information of the target electronic device, that is, the target model data; at the same time, the first model data that is unrelated to the part of the model data loaded by the target electronic device is used as the model data executed locally on the first electronic device.

[0132] For another example, based on the resource occupancy in the configuration information of the target electronic device, it is determined that the available computing resources in the target electronic device are small, and the model data set can be split into first model data and second model data, wherein the data volume of the first model data is smaller than the data volume of the second model data, and the first model data is used as model data that matches the configuration information of the target electronic device, and the second model data is used as model data that is executed locally on the first electronic device.

[0133] For another example, based on the network information of the target electronic device in the configuration information of the target electronic device, it is determined that the network connection speed between the target electronic device and the first electronic device is high and the network delay is low, then the model data set can be split into first model data and second model data, wherein the data volume of the first model data is greater than the data volume of the second model data, and the first model data is used as model data that matches the configuration information of the target electronic device, and the second model data is used as model data that is executed locally on the first electronic device.

[0134] Here, the first part of the task data is the task data corresponding to the first model data. In this way, the first electronic device is used to run the first model data to implement model reasoning for the first part of the task data.

[0135] At the same time, the target task data and the target model data are transmitted to the target electronic device, that is, the above step S102, can be implemented as the following step S1021:

[0136] Step S1021: transmitting the target model data and a second part of the target task data to the target electronic device, so that the target electronic device processes the second part of the task data using the target model data.

[0137] Here, the second part of the task data is the task data corresponding to the target model data (ie, the second model data). In this way, the target model data is run by the target electronic device to implement model reasoning for the second part of the task data.

[0138] In the above embodiment, based on the configuration information of the target electronic device, the first model data and the second model data are determined from the model data set of the processing model, and the first model data and the second model data are respectively executed by the first electronic device and the target electronic device to complete the model reasoning of the first part of the task data and the second part of the task data. In this way, based on the configuration information of the target electronic device, the model reasoning can be executed by using the computing resources of the first electronic device and the target electronic device at the same time, so as to achieve the effect of improving the model reasoning speed.

[0139] In some embodiments, packaging processing is performed on the target model data and the target task data to prevent sticking and unpacking problems during data transmission, thereby improving the security and integrity of the target model data and the target task data during transmission.

[0140] Thus, in some embodiments, the transmitting of the target task data and the target model data to the target electronic device, that is, the above step S102, can be implemented as the following steps S1022 to S1024:

[0141] Step S1022, based on the reference data corresponding to the target processing task, the target model data is packaged and processed to generate first packet header data; wherein the reference data includes at least one of the following: the number of the target electronic device; the device identification of the target electronic device; the hash value of the target model data.

[0142] Here, the number of target electronic devices refers to the number of target electronic devices called by the first electronic device in one task processing.

[0143] In some embodiments, the number of target electronic devices may be one or more, that is, the first electronic device may call one or more target electronic devices to assist in executing the target processing task.

[0144] In this way, when the target model data and the target task data are packaged, the number of data packets can be determined according to the number of the called target electronic devices.

[0145] The device identification of the target electronic device refers to information that uniquely identifies the corresponding electronic device.

[0146] In this way, when the first electronic device calls multiple target electronic devices, the device identification information of the designated target electronic device can be filled in the first header of each data packet, so that during the data packet transmission process, the transmission route of the data packet can be determined based on the device identification in the first header.

[0147] The hash value of the target model data refers to the hash value obtained by performing a hash calculation on the target model data using a preset hash algorithm.

[0148] Here, filling in the hash value of the target model data in the first header part of the data packet can enable the target electronic device receiving the data packet to verify the integrity and legality of the data packet based on the hash value, thereby improving the security of data transmission.

[0149] Here, the target model data is simultaneously stored in the first header portion of the data packet to be transmitted to the corresponding target electronic device.

[0150] Step S1023, packaging the target task data to generate package body data.

[0151] Here, the target task data refers to information such as task functions and function parameters obtained after parsing the target processing task.

[0152] In some embodiments, when a first electronic device calls multiple target electronic devices, the target task data can be split into multiple target sub-tasks; based on the target model data sent to each target electronic device, the target sub-task sent to the corresponding target electronic device is determined; the target model data and target sub-task data corresponding to each target electronic device are packaged and processed, and a first packet header data is generated based at least on the device identification of the target electronic device.

[0153] Here, the target task data is packaged and processed to generate package data.

[0154] Step S1024, transmitting the data consisting of the first packet header data and the packet body data to a target electronic device;

[0155] Here, the data consisting of the first header data and the body data is transmitted to the corresponding target electronic device by using the communication network between the target electronic device and the first electronic device.

[0156] In some embodiments, when performing packaging processing on target model data and target task data, the target model data and target task data can be divided into dynamic resources and static resources and packaged separately; wherein, static resources include neural network data and corresponding parameter data in the target model data, and dynamic resources include logical instructions involved in performing model reasoning on the target task data.

[0157] In some embodiments, when the data volume of the target task data is less than a data volume threshold, the target task data is packaged based on the reference data to generate second packet header data containing the target model data and the target task data, and the second packet header data is transmitted to the target electronic device.

[0158] Here, when the data volume of the target task data is less than the data volume threshold, the target task data and the target model data are both packaged into a packet header, thereby improving data transmission efficiency.

[0159] In some embodiments, where the reference data includes the number of target electronic devices, the number of packaged data packets is determined based on the number of target electronic devices.

[0160] In some embodiments, when the reference data includes a device identification of the target electronic device, a packet header of the data packet is generated based on the device identification of the target electronic device, so that during data transmission, a transmission route of the data packet is determined based on the device identification.

[0161] In some embodiments, when the reference data includes a hash value of the target model data and a hash value of the target task data, a header of the data packet is generated based on the hash value of the target model data and the hash value of the target task to use the hash value as integrity verification data.

[0162] In the above embodiment, by packaging the target model data and the target task data and then sending them to the target electronic device, the data volume of the target model data and the target task data can be compressed, and the problems of packet sticking and unpacking can be reduced, thereby ensuring the integrity of data transmission; at the same time, in the case of multiple target electronic devices, the packet header data of the data packet can also be used to identify the target electronic device corresponding to each data packet.

[0163] On the other hand, the present application also provides a data processing method applied to a target electronic device. Figure 2 As shown, the data processing method includes the following steps S201 to S202:

[0164] Step S201, obtaining target task data and target model data corresponding to the target processing task.

[0165] Here, the target electronic device receives the target task data and the target model data corresponding to the target processing task from the first electronic device through the communication network between the target electronic device and the first electronic device.

[0166] The target task data and target model data corresponding to the target processing task are data determined by the first electronic device in response to the target processing task. In some embodiments, the specific implementation of the first electronic device determining the target task data and target model data can refer to the detailed description above.

[0167] Step S202: Process the target task data through the target model data using the hardware resources of the target electronic device to obtain task result data.

[0168] Here, the hardware resources of the target electronic device refer to the hardware computing resources that can be used or called by the target electronic device.

[0169] In some embodiments, the hardware resources of the target electronic device include at least one computing unit in the target electronic device, for example, at least one GPU.

[0170] In some embodiments, the hardware resources of the target electronic device also include hardware resources in other target electronic devices that can be called by the target electronic device.

[0171] Thus, in some embodiments, the target electronic device may utilize hardware resources of multiple target electronic devices to assist in executing the target processing task based on reference data corresponding to the target processing task.

[0172] Here, the reference data corresponding to the target processing task may include the number of target electronic devices required to perform the target processing task, or the amount of hardware computing resources required. In this way, the current target electronic device can determine at least one target electronic device with a specified number or a specified amount of hardware computing resources from multiple target electronic devices based on the reference data, and send the target task data corresponding to the target processing task to the determined at least one target electronic device.

[0173] The reference data corresponding to the target processing task may include the device identification of at least one target electronic device specified in the target processing task. Thus, the current target electronic device sends the target task data and the corresponding target model data to at least one target electronic device according to the device identification of the at least one target electronic device.

[0174] In some embodiments, based on reference data corresponding to the target processing task, the target processing task is executed using multiple computing units of the target electronic device.

[0175] Here, the target electronic device determines to use local hardware resources to execute the target processing task based on reference data corresponding to the target processing task, for example, information such as the device identification of the target electronic device specified in the target processing task and required hardware resources.

[0176] In some embodiments, the multiple computing units in the target electronic device may be computing units that execute the same or different instruction sets, or the multiple computing units may be computing units with different computing capabilities.

[0177] In this way, the first electronic device can select at least one computing unit with the highest processing efficiency from multiple computing units of the target electronic device based on the configuration information of the hardware resources in the target electronic device to execute the target processing task.

[0178] In this way, the target electronic device executes the target processing task using at least one computing unit selected by the first electronic device based on the reference data corresponding to the target processing task.

[0179] In some embodiments, after the target electronic device receives the target model data, it caches the target model data in a cache of the target electronic device (for example, in a cache corresponding to at least one computing unit specified by the first electronic device) so that when the target model data is subsequently called, the data transmission time between the first electronic device and the target electronic device can be reduced.

[0180] In some embodiments, the target electronic device also caches the function body portion in the parsed logic instruction in a cache of the target electronic device.

[0181] In this way, when the first electronic device sends the next target processing task that can be executed using the target model data to the target electronic device, the first electronic device will encapsulate the data list of the data involved in the target processing task into a hash judgment function and send it to the target electronic device, where the data list includes the processing model information corresponding to the next target processing task, the function body information of the logic instruction corresponding to the target processing task, etc.; then, the target electronic device determines whether the data specified in the data list is stored locally according to the data list, and feeds back the data not stored locally to the first electronic device; thereafter, the first electronic device sends the data not stored in the target electronic device according to the information fed back by the target electronic device. In this way, the data transmission overhead between the first electronic device and the target electronic device is reduced, and the overall operation efficiency is improved.

[0182] In some embodiments, after the target electronic device feeds back task result data for the target processing task to the first electronic device, the target electronic device releases the computing resources occupied by the target model data.

[0183] Next, combine Figure 3 and Figure 4 , an embodiment of the data processing method provided by the present application is described.

[0184] First, if Figure 3 As shown, the part executed by the first electronic device in this embodiment includes the following steps S301 to S309:

[0185] Step S301, obtaining a text image task; then, executing step S302;

[0186] Here, the text image task input by the user is obtained.

[0187] Step S302, obtaining configuration information of the local end and the server end; then, executing step S303;

[0188] Here, the local end is the first electronic device; the server end is the target electronic device that can be called by the first electronic device.

[0189] The local end obtains the configuration information of the local end and the server end through the monitoring service. The configuration information includes the network information between the local end and the server end, the model configuration information of the server end, the resource usage of the server end, and the resource usage of the local end.

[0190] Step S303, the local end determines whether to call the server end to execute the text map task; if not, execute step S304; if yes, execute step S305;

[0191] Here, the local end determines whether to call the server end to execute the text map task based on the obtained configuration information of the local end and the server end.

[0192] Step S304, using local hardware resources to process the text and image task;

[0193] Here, based on the network information, it is determined that the network connection between the local end and the server end is poor, or the resource usage of the server end is large, and then it is determined to use local hardware resources to process the text image task.

[0194] Step S305, using the CPU to parse the large language model and the text graph task; then, executing step S306;

[0195] Here, based on the network information, it is determined that the network connection between the local end and the server end is good, the server end has no cached large language model for executing the Wensheng graph task, and the server end resource occupancy is low, so it is determined to call the server end hardware resources to execute the Wensheng graph task and send the large language model to the server end.

[0196] Based on the instruction set corresponding to the computing unit on the server side, the local side uses the CPU to pre-parse the large language model and text graph tasks to determine the parameter data, function data, function parameters, etc. involved in the large language model and text graph tasks, and generates function instructions that can be executed by the server-side GPU.

[0197] Step S306, compile the parsed large language model and text graph task; then, execute step S307;

[0198] Here, the local end compiles the parsed large language model and text graph task to obtain the corresponding binary data.

[0199] Step S307, packaging the compiled large language model and text graph tasks; then, executing step S308;

[0200] Here, the compiled binary data is packaged and the device identification information corresponding to the server, the large language model and the hash value information corresponding to the text graph task are generated in the packet header.

[0201] Step S308, compressing and processing the packaged large language model and text graph task; then, executing step S309;

[0202] Here, compression processing is performed on the packaged large language model and text graph tasks to reduce the amount of data and shorten the network transmission time.

[0203] Step S309, sending the compressed large language model and text graph task to the server.

[0204] Here, the compressed large language model and text graph task are sent to the server through the communication network between the local end and the server.

[0205] Secondly, if Figure 4 As shown, the part executed by the target electronic device (ie, the server) in this embodiment includes the following steps S401 to S406:

[0206] Step S401, receiving the compressed large language model and text graph task from the local end; then, executing step S402;

[0207] Step S402, decompressing the large language model and the text graph task; then, executing step S403;

[0208] Step S403, performing unpacking processing on the decompressed large language model and text graph task; then, executing step S404;

[0209] Here, when the server performs unpacking processing on the decompressed large language model and text graph tasks, it verifies the hash value and device identification information in the data to determine the integrity of the received data packet.

[0210] Step S404: Decompile the unpacked large language model and text graph task

[0211] Here, the server uses the GPU decompilation module to perform decompilation processing on the unpacked large language model and text graph task to obtain a GPU function that can be executed by the server GPU.

[0212] Step S405, using the GPU to run the decompiled large language model and execute the text-to-graph task to obtain the task processing result; then, executing step S406;

[0213] Here, the server uses GPU resources to run the large language model, and uses the large language model to perform the text-to-image task and generate the corresponding image processing results.

[0214] Step S406, sending the task processing result to the local end.

[0215] Here, the server sends the task processing results to the local end through the communication network.

[0216] In the above embodiment, after the server performs unpacking processing on the large language model and text graph task and obtains the corresponding binary data, it can perform decompilation processing on the binary data and store the decompiled large model data, so that when the server is subsequently called by other electronic devices, the server can load the large language model from the local.

[0217] As can be seen from the above, when the local end calls the server end to execute the Wensheng graph task, the configuration information of the local end and the server end is obtained in advance, so as to flexibly determine whether to call the server end to execute the Wensheng graph task according to the configuration information; at the same time, before sending the large language model and Wensheng graph task to the server end, the local end pre-parses and compiles the large language model and Wensheng graph task, so that the server end does not need to perform parsing and compilation after receiving the data transmitted by the local end. The server end can directly load the received data into the cache and use the GPU to perform model reasoning, realizing the parallel processing of the local end CPU and the server end GPU, and improving the task processing speed. In addition, the server end directly loads the model data sent by the local end into the local cache for execution, which can improve the model loading speed and thus improve the task processing speed.

[0218] Based on the foregoing embodiments, the present application provides a data processing device, which includes the units included and the modules included in the units, and can be implemented by a processor in a computer device; of course, it can also be implemented by a specific logic circuit; in the implementation process, the processor can be a central processing unit CPU, a microprocessor (Microprocessor Unit, MPU), a digital signal processor (Digital Signal Processor, DSP) or a field programmable gate array (Field Programmable Gate Array, FPGA), etc.

[0219] Figure 5 A schematic diagram of the structure of a data processing device provided in this application is shown in FIG. Figure 5 As shown, the data processing device 500 includes: a determination unit 510 and a transmission unit 520, wherein:

[0220] A determination unit 510, configured to determine a processing model corresponding to the target processing task in response to obtaining the target processing task;

[0221] The transmission unit 520 is used to transmit the target task data and the target model data to the target electronic device so that the target electronic device processes the target task data using the target model data. The target model data is at least part of the model data of the processing model, and the target task data is the data obtained by compiling the target processing task by the first electronic device.

[0222] In some embodiments, the determining unit 510 is further configured to:

[0223] Obtaining configuration information of the target electronic device, and using model data that matches the configuration information in a model data set obtained by parsing the processing model as the target model data;

[0224] The configuration information includes at least one of the following:

[0225] Model configuration information of the target electronic device;

[0226] Model usage information of the target electronic device;

[0227] The resource usage of the target electronic device;

[0228] The network information of the target electronic device.

[0229] In some embodiments, the determining unit 510 is further configured to:

[0230] The model data in the storage area of ​​the first electronic device is obtained, and the model data in the storage area is compared with the model data set obtained by parsing the processing model to obtain first model difference data, and the first model difference data is used as the target model data.

[0231] In some embodiments, the apparatus 500 further includes a storage unit 530;

[0232] The storage unit 530 is used to store the target model data in a storage area of ​​the first electronic device;

[0233] The target model data includes at least one of the following:

[0234] All model data obtained by parsing the processing model;

[0235] The first model difference data is obtained by comparing the model data in the storage area with the model data set.

[0236] In some embodiments, the determining unit 510 is further configured to:

[0237] The model data in the cache area of ​​the target electronic device is obtained, and the model data in the cache area is compared with the model data set obtained by parsing the processing model to obtain second model difference data, and the second model difference data is used as the target model data.

[0238] In some embodiments, the target task data includes a first portion of task data and a second portion of task data;

[0239] The determining unit 510 is configured to:

[0240] using model data that does not match the configuration information in the model data set obtained by parsing the processing model as first model data, so that the first electronic device processes the first part of the task data using the first model data;

[0241] taking second model data matching the configuration information in the model data set obtained by parsing the processing model as target model data;

[0242] The target model data and a second part of the target task data are transmitted to the target electronic device, so that the target electronic device processes the second part of the task data using the target model data.

[0243] In some embodiments, the determining unit 510 is used to package the target model data based on the reference data corresponding to the target processing task to generate first header data; package the target task data to generate body data; wherein the reference data includes at least one of the following: the number of the target electronic device; the device identification of the target electronic device; the hash value of the target model data;

[0244] The transmission unit 520 is used to transmit the data consisting of the first packet header data and the packet body data to a target electronic device.

[0245] On the other hand, the present application also provides a data processing device that has been set on a target electronic device. Figure 6 As shown, the data processing device 600 includes a communication unit 610 and a processing unit 620; wherein:

[0246] The communication unit 610 is used to obtain target task data and target model data corresponding to the target processing task;

[0247] The processing unit 620 is used to process the target task data through the target model data using the hardware resources of the target electronic device to obtain task result data.

[0248] In some embodiments, the processing unit 620 is configured to:

[0249] Based on the reference data corresponding to the target processing task, the target processing task is executed using hardware resources of a plurality of target electronic devices; or,

[0250] Based on the reference data corresponding to the target processing task, the target processing task is executed using multiple computing units of the target electronic device.

[0251] The description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiment. For technical details not disclosed in the device embodiment of the present application, please refer to the description of the method embodiment of the present application for understanding.

[0252] It should be noted that in the embodiment of the present application, if the above-mentioned data processing method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific hardware, software or firmware, or any combination of hardware, software, and firmware.

[0253] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, some or all of the steps in the above method are implemented.

[0254] The embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium can be transient or non-transient.

[0255] An embodiment of the present application provides a computer program, including a computer-readable code. When the computer-readable code is run in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.

[0256] The embodiment of the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be implemented specifically by hardware, software or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium, and in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK) and the like.

[0257] It should be noted here that the description of the various embodiments above tends to emphasize the differences between the various embodiments, and the same or similar aspects can be referenced to each other. The description of the above device, storage medium, computer program and computer program product embodiments is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the embodiments of the device, storage medium, computer program and computer program product of this application, please refer to the description of the method embodiment of this application for understanding.

[0258] It should be noted that Figure 7 is a hardware entity schematic diagram of an electronic device in an embodiment of the present application, such as Figure 7 As shown, the hardware entity of the electronic device 700 includes: a processor 701, a communication interface 702 and a memory 703, wherein:

[0259] The processor 701 generally controls the overall operation of the electronic device 700 .

[0260] The communication interface 702 enables the electronic device to communicate with other terminals or servers through a network.

[0261] The memory 703 is configured to store instructions and applications executable by the processor 701, and can also cache data to be processed or processed by the processor 701 and each module in the electronic device 700 (for example, image data, audio data, voice communication data, and video communication data), which can be implemented by flash memory (FLASH) or random access memory (Random Access Memory, RAM). Data transmission can be performed between the processor 701, the communication interface 702, and the memory 703 through the bus 704.

[0262] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present application, the size of the serial numbers of the above-mentioned steps / processes does not mean the order of execution, and the execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments.

[0263] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0264] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. 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, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0265] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0266] In addition, all functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0267] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, etc., various media that can store program codes.

[0268] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0269] The above is only an implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A data processing method, applied to a first electronic device, comprising: In response to obtaining a target processing task, determining a processing model corresponding to the target processing task; The target task data and the target model data are transmitted to the target electronic device so that the target electronic device processes the target task data using the target model data; the target model data is at least part of the model data of the processing model; the target task data is the data obtained by compiling the target processing task by the first electronic device.

2. The method according to claim 1, before transmitting the target task data and the target model data to the target electronic device, further comprising: Obtaining configuration information of the target electronic device, and using model data that matches the configuration information in a model data set obtained by parsing the processing model as the target model data; The configuration information includes at least one of the following: Model configuration information of the target electronic device; Model usage information of the target electronic device; The resource usage of the target electronic device; The network information of the target electronic device.

3. The method according to claim 1, before transmitting the target task data and the target model data to the target electronic device, further comprising: The model data in the storage area of ​​the first electronic device is obtained, and the model data in the storage area is compared with the model data set obtained by parsing the processing model to obtain first model difference data, and the first model difference data is used as the target model data.

4. The method according to claim 3, wherein the transmitting the target task data and the target model data to the target electronic device further comprises: Storing the target model data in a storage area of ​​the first electronic device; The target model data includes at least one of the following: All model data obtained by parsing the processing model; The first model difference data is obtained by comparing the model data in the storage area with the model data set.

5. The method according to claim 1, before transmitting the target task data and the target model data to the target electronic device, further comprising: The model data in the cache area of ​​the target electronic device is obtained, and the model data in the cache area is compared with the model data set obtained by parsing the processing model to obtain second model difference data, and the second model difference data is used as the target model data.

6. The method according to claim 2, wherein the target task data comprises a first portion of task data and a second portion of task data; The method further comprises: using model data that does not match the configuration information in the model data set obtained by parsing the processing model as first model data, so that the first electronic device processes the first part of the task data using the first model data; taking second model data matching the configuration information in the model data set obtained by parsing the processing model as target model data; The transmitting the target task data and the target model data to the target electronic device includes: transmitting the target model data and the second part of the task data to the target electronic device, so that the target electronic device processes the second part of the task data using the target model data.

7. The method according to claim 1, wherein the step of transmitting the target task data and the target model data to the target electronic device comprises: Based on the reference data corresponding to the target processing task, the target model data is packaged to generate first packet header data; Packaging the target task data to generate package data; Transmitting data consisting of the first packet header data and the packet body data to a target electronic device; The reference data includes at least one of the following: the number of the target electronic devices; The device identification of the target electronic device; The hash value of the target model data.

8. A data processing method, applied to a target electronic device, the method comprising: Obtain target task data and target model data corresponding to the target processing task; The target task data is processed by the target model data using the hardware resources of the target electronic device to obtain task result data.

9. The method according to claim 8, wherein the step of processing the target task data corresponding to the target processing task using the hardware resources of the target electronic device through the target model data to obtain task result data further comprises: Based on the reference data corresponding to the target processing task, using the hardware resources of a plurality of target electronic devices to execute the target processing task; or, Based on the reference data corresponding to the target processing task, the target processing task is executed using multiple computing units of the target electronic device.

10. A data processing device, comprising: a determining unit, configured to determine a processing model corresponding to the target processing task in response to obtaining the target processing task; A transmission unit is used to transmit target task data and target model data to a target electronic device so that the target electronic device processes the target task data using the target model data. The target model data is at least part of the model data of the processing model, and the target task data is data obtained by compiling the target processing task by the first electronic device.

Images