WO2021208105A1 - 数据处理方法及通信设备、计算机存储介质 - Google Patents
数据处理方法及通信设备、计算机存储介质 Download PDFInfo
- Publication number
- WO2021208105A1 WO2021208105A1 PCT/CN2020/085444 CN2020085444W WO2021208105A1 WO 2021208105 A1 WO2021208105 A1 WO 2021208105A1 CN 2020085444 W CN2020085444 W CN 2020085444W WO 2021208105 A1 WO2021208105 A1 WO 2021208105A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- communication device
- configuration information
- data
- division
- information
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5066—Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5017—Task decomposition
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W8/00—Network data management
- H04W8/22—Processing or transfer of terminal data, e.g. status or physical capabilities
Definitions
- This application relates to the field of communication technology, and in particular to a data processing method, communication equipment, and computer storage medium.
- AI Artificial Intelligence
- ML Machine Learning
- the data processing of the AI/ML model can be completed through cooperation between the terminal and the network device or other terminals; specifically, the AI/ML model can be divided into multiple sub-models, and the terminal performs data through some of the multiple sub-models For processing, network equipment or other terminals perform data processing through another part of the multiple sub-models.
- the segmentation method of AI/ML model in the prior art is fixed, that is to say, all data in the data to be processed are processed through the fixed segmentation method, which results in the available processing of the terminal in some cases Resources cannot meet the size of the resources to be occupied by the fixed partitioning method; therefore, data processing performed by the fixed partitioning method will cause the problem of unbalanced resource occupancy.
- the embodiments of the present application provide a data processing method, a communication device, and a computer-readable storage medium, which improve the balance of resource occupation in the data processing process and the resource utilization rate in the data processing process.
- the embodiment of the present application provides a data processing method, which is applied to a first communication device, and includes:
- the first configuration information is used to configure N division modes and the j-th part corresponding to the i-th division mode among the N division modes; where N is an integer greater than or equal to 1 , I is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1;
- the N types of segmentation methods are segmentation methods in which the data processing model is segmented into at least two sub-processing models by preset segmentation positions;
- the embodiment of the present application provides a data processing method, which is applied to a second communication device, and includes:
- At least one piece of configuration information is determined; wherein, each piece of configuration information in the at least one piece of configuration information is used to configure N types of partitioning methods, and the j-th part corresponding to the i-th type of partitioning methods among the N types of partitioning methods; wherein, N is an integer greater than or equal to 1, i is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1;
- the N types of segmentation methods are segmentation methods in which the data processing model is segmented into at least two sub-processing models by preset segmentation positions;
- An embodiment of the present application provides a first communication device, and the communication device includes:
- the first obtaining unit is configured to obtain first configuration information; the first configuration information is used to configure N division modes and the j-th part corresponding to the i-th division mode among the N division modes; where N Is an integer greater than or equal to 1, i is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1;
- the N division modes are division modes in which the data processing model is divided into at least two sub-processing models by preset division positions.
- An embodiment of the present application provides a second communication device, and the communication device includes:
- the second processing unit is configured to determine at least one piece of configuration information; wherein, each piece of configuration information in the at least one piece of configuration information is used to configure N division modes, N parts into which data is divided, and an i-th division mode
- N is an integer greater than or equal to 1
- i is greater than or equal to 1 and less than or equal to N
- j is greater than or equal to 1 and less than or equal to M
- M is an integer greater than 1
- the second sending unit is configured to send the at least one configuration information to the first communication device; the at least one configuration information includes the first configuration information.
- An embodiment of the present application provides a first communication device, and the communication device includes:
- the communication device includes: a first transceiver, a first processor, and a first memory storing a computer program;
- the first processor is configured to execute the above-mentioned data processing method applied to the first communication device when running the computer program stored in the first memory in conjunction with the first transceiver.
- An embodiment of the present application provides a second communication device, and the communication device includes:
- the second processor is configured to execute the above-mentioned data processing method applied to the second communication device when running the computer program stored in the second memory in conjunction with the second transceiver.
- the embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and the computer program is executed by a first processor to realize the steps of the data processing method; or; the computer program is processed by a second processor.
- the steps of the data processing method described by the device are also provided.
- the first communication device can configure the N division modes of the data processing model and the jth division mode corresponding to the i-th division mode among the N division modes according to the configuration information. Parts. That is to say, the first communication device can process the j-th part of the data by using the i-th segmentation method, that is, use different segmentation methods for processing in different parts of the data, so that the N-type segmentation method is used to process the data.
- the processing resources occupied at the time match the available processing resources of the first communication device; the balance of resource occupation in the data processing process and the resource utilization rate in the data processing process are improved.
- FIG. 1 is a schematic diagram of a data transmission scenario of a communication network provided by an embodiment of this application;
- FIG. 2 is a schematic diagram of a network architecture for offloading of AI/ML operations according to an embodiment of the application
- FIG. 3 is a schematic diagram of a network architecture in which a terminal and a network device cooperate to perform AI/ML model operations according to an embodiment of the application;
- FIG. 4 is a schematic diagram of a network architecture of a communication system provided by an embodiment of this application.
- FIG. 5 is a schematic flow chart 1 of a data processing method provided by an embodiment of this application.
- FIG. 6 is a schematic structural diagram of a CNN model provided by an embodiment of the application.
- FIG. 7 is a first schematic diagram of switching a division mode provided by an embodiment of this application.
- FIG. 8 is a second schematic flow diagram of a data processing method provided by an embodiment of this application.
- FIG. 9 is a second schematic diagram of switching a division mode provided by an embodiment of this application.
- FIG. 10 is a third schematic diagram of switching of a division mode provided by an embodiment of this application.
- FIG. 11 is a fourth schematic diagram of switching of a division mode provided by an embodiment of this application.
- FIG. 12 is a third schematic flowchart of a data processing method provided by an embodiment of this application.
- FIG. 13 is a first structural diagram of a communication device provided by an embodiment of this application.
- FIG. 14 is a second schematic structural diagram of a communication device provided by an embodiment of this application.
- 15 is a schematic diagram 1 of the hardware structure of a communication device provided by an embodiment of this application.
- FIG. 16 is a second schematic diagram of the hardware structure of a communication device provided by an embodiment of this application.
- eMBB enhanced mobile broadband communication
- mMTC large-scale machine communication
- URLLC ultra-reliable low-latency communication
- AI/ML applications AI/ML applications.
- eMBB can support the transmission of large data traffic between the terminal and the network device (such as 5G cloud or 5G edge computing device), such as the transmission of multimedia data such as video and audio between the terminal and the network device.
- the network device such as 5G cloud or 5G edge computing device
- mMTC/URLLC refers to the support of low-latency and highly reliable data transmission between a large number of terminals and network equipment, such as remote sensors, smart cars and other terminals and network equipment for the transmission of sensing data and control commands; AI/ML
- the application can support the transmission of AI/ML models between terminals and network devices.
- AI/ML applications face three major challenges, as follows:
- the terminal lacks the computing resources, storage resources and battery capacity required to run the AI/ML model completely locally;
- 3GPP Third Generation Partnership Project
- 3GPP Third Generation Partnership Project
- the network device needs to perform the operation of the AI/ML model, and also needs to act as a control node to process the data sent by the terminal and send it to the terminal Control instruction.
- the terminal cooperates with the network device to complete the operation of the AI/ML model.
- the AI/ML model can be divided into multiple sub-models, and the terminal performs data processing through some of the multiple sub-models, and The obtained processing result is transmitted to the network device, and the network device performs data processing through another part of the multiple sub-models; in this way, the terminal and the network device can cooperate to process the data.
- the segmentation method of AI/ML model in the prior art is fixed, that is to say, all data in the data to be processed are processed through this fixed segmentation method, which results in the available processing of the terminal in some cases Resources cannot meet the size of the resources to be occupied by the fixed partitioning method; therefore, data processing performed by the fixed partitioning method will cause the problem of unbalanced resource occupancy.
- FIG. 4 is a schematic diagram of a network architecture of a communication system to which an embodiment of the present application is applicable. Below, the wireless communication system involved in the embodiment of the present application is briefly introduced.
- the network architecture of the communication system includes terminal-to-network communication and device-to-device (D2D) communication;
- the communication system includes a network device 410, a terminal 420, and a terminal 430.
- the terminal-to-network communication includes communication between the terminal 420 and the network device 410, or the communication between the terminal 430 and the network device 410, and D2D communication includes the communication between the terminal 420 and the terminal 430.
- the network device 410 mentioned in the embodiment of this application may be a base station (Base Transceiver) of a global system of mobile communication (GSM) system or a code division multiple access (Code Division Multiple Access, CDMA) system.
- BTS can also be a base station (NodeB, NB) in a Wideband Code Division Multiple Access (WCDMA) system, or an evolved base station in a Long Term Evolution (LTE) system (evolved NodeB, eNB), access point (AP) or relay station, it can also be a base station in a 5G system (such as gNB or Transmission Point (TRP)), etc., or it can be a cloud wireless access network (Cloud Radio Access Network, CRAN) scenarios such as wireless controllers, wearable devices or vehicle-mounted devices.
- CRAN Cloud Radio Access Network
- the terminal 420 and the terminal 430 mentioned in the embodiments of this application may be User Equipment (UE), access terminal, UE unit, UE station, mobile station, mobile station, remote station, remote terminal, mobile equipment, UE Terminal, terminal, wireless terminal equipment, UE agent or UE device, etc. It can also be a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a wireless local loop (Wireless Local Loop, WLL) station, a personal digital processing (Personal Digital Assistant, PDA), and a wireless communication function.
- UE User Equipment
- UE unit User Equipment
- UE station mobile station
- mobile station mobile station
- remote station remote terminal
- mobile equipment UE Terminal
- terminal wireless terminal equipment
- UE agent or UE device etc. It can also be a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a wireless local loop (Wireless Local Loop, WLL) station, a personal digital processing (Pers
- PLMN Public Land Mobile Network
- the devices with communication functions in the network/system in the embodiments of the present application may be collectively referred to as communication devices.
- the communication device may include a network device 410 having a communication function, a terminal 420 and a terminal 430.
- an embodiment of the present application provides a data processing method, and the execution subject of the data processing method is the first communication device.
- the first communication device may be any one of the network device 410, the terminal 420, and the terminal 430 shown in FIG. 4.
- the second communication device when the first communication device is the network device 410, the second communication device is the terminal 420 or the terminal 430.
- the first communication device when the first communication device is the terminal 420, the second communication device may be the network device 410 or the terminal 430.
- the second communication device may be the network device 410 or the terminal 420.
- the data processing method includes step 510. in,
- Step 510 Obtain first configuration information.
- the first configuration information is used to configure the N division modes and the j-th part corresponding to the i-th division mode among the N division modes; where N is an integer greater than or equal to 1, and i is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1.
- the N division modes are division modes in which the data processing model is divided into at least two sub-processing models by preset division positions.
- the data processing model mentioned in the embodiment of the present application may be any type of AI/ML model, such as an AI/ML model for image recognition, an AI/ML model for natural language recognition, and so on.
- AI/ML model for image recognition
- AI/ML model for natural language recognition and so on.
- the embodiments of the application are not limited here.
- the N division modes in the first configuration information may be one division mode or multiple division modes among a plurality of preset division modes, which is not limited in this embodiment of the present application.
- the first communication device may also determine that the data is divided into M parts through the first configuration information.
- data refers to data that needs to be processed through a data processing model (data can also be referred to as data to be processed).
- the data can be any type of data, such as video data, text data, and audio. Data, etc., are not limited here in the embodiment of the present application.
- part mentioned in the embodiment of this application refers to a part of data in the data
- combination of the M parts in the embodiment of this application constitutes the whole of the data.
- each part of the data may be composed of discrete data units, that is, the data unit in each part of the data is not a continuous data unit.
- the video data can be divided into two parts, the odd-numbered frames in the video data constitute the first part, and the even-numbered frames in the video data constitute the second part.
- the above-mentioned data unit refers to the smallest unit for processing data.
- the data unit specifically refers to an image frame, that is, the data can be divided into image frames one by one for processing.
- the data when the data is text data, the data can be divided into a piece of data of a fixed size (10K) for processing, where a piece of data of a fixed size is a data unit.
- the first configuration information may configure the j-th part corresponding to the i-th splitting method among the N splitting methods. That is, the first configuration information may also configure the corresponding splitting method for each of the N splitting methods. part.
- the correspondence relationship here means that when the first communication device uses the i-th division method for processing, what is processed is the j-th part of the entire data. In other words, the first communication device can process the j-th part of the data through the sub-processing model corresponding to the i-th segmentation method.
- each partitioning method corresponds to a part of the data, and the partitioning method corresponds to the part of the data one-to-one.
- the first configuration information configures a split mode (that is, N is 1)
- M takes a value greater than N (that is, M is greater than 1); that is, even if the first configuration information configures a split mode, it still
- the data is divided into multiple (M) parts, and the j-th part corresponding to this division mode configured in the first configuration information; that is, the first communication device is in one of the M parts of the data Part of the use of this type of division.
- the processing resources of the first communication device that need to be occupied when each segmentation method is used for processing are different.
- the available processing resources of the first communication device cannot completely match the processing resources that need to be occupied by each division method. In most cases, the amount of available processing resources may be between the amount of processing resources that need to be occupied by different division methods.
- the available processing resource size of the first communication device at time 1 is the first available processing resource amount 1, where the first available processing resource amount 1 is greater than the division position 2 (that is, the second division method).
- the first processing resource amount 2 occupied is smaller than the first processing resource amount 3 occupied by the division position 3 (that is, the third division method). Therefore, using the split location 2 for data processing will cause a waste of available resources of the first communication device, and using the split location 3 for data processing, the amount of processing resources required exceeds the first available processing resource amount 1 again.
- the first communication device may configure N division modes for the data processing model through the first configuration information. Different parts of the data are processed using different partitioning methods, so that the average amount of processing resources occupied by the first communication device matches the processing resources currently available to the first communication device.
- the N division modes include only one division mode, and the first configuration information may configure the j-th part corresponding to the one division mode.
- the first communication device uses this division method to process the j-th part of the data.
- the first communication device does not use any segmentation method for processing other data parts in the data (that is, the other parts of the data are sent to the second communication device for processing), or the other parts of the data use specific Split the way to deal with.
- the specific division mode is a division mode agreed upon by the first communication device and the second communication device, and the specific division mode is any one of the division modes preset by the data processing model.
- the specific division manner may be a factory-configured division manner, or may be a division manner that is semi-statically configured by the second communication device.
- the configuration information only configures the first division mode and the part corresponding to the first division mode is the odd-numbered frame part of the data; in this way, the first communication device can use the above-mentioned first division in the odd-numbered frame of the data. In the even-numbered frame of the data, no splitting method is used for processing.
- the configuration information only configures the first split mode and the odd-numbered frame part of the data corresponding to the first split mode.
- the specific division method agreed upon by the first communication device and the second communication device is the second division method; in this way, the first communication device can use the above-mentioned first division method for processing in odd-numbered frames of data, and use the second division method in even-numbered frames of data. Split the way to deal with.
- the N division modes include two or more division modes
- the first communication device may according to the part of the data corresponding to each division of the N division modes, in the part where the data is different, Use different segmentation methods for processing.
- the N division methods include two division methods: the first division mode and the second division mode; wherein the corresponding part of the first division mode is the odd-numbered frame part of the data, and the second division mode The corresponding part is the even frame part of the data.
- the first communication device uses the first division method for processing in the odd-numbered frames of the data, and uses the second division method for processing in the even-numbered frames of the data, that is, alternately applies the first division method and the second division method. Process the data.
- the principle of the at least one division mode and the data information of the at least one division mode configured in the configuration information is that the first communication device uses the first communication device occupied by the at least one division mode The processing resources of matches the available processing resources of the first communication device.
- step 510 After obtaining the first configuration information in step 510, the following steps may be further included:
- Step 520 Use the sub-processing model corresponding to the i-th segmentation method to process the j-th part of the data.
- the first communication device can use the sub-processing model of this division method to process the jth part of the data; and, the first communication device does not use any division for other parts of the data except the jth part.
- the method is used for processing (that is, the other parts are sent to the second communication device for processing); that is, the data is processed by alternately using a split mode and a non-split mode configured in the first configuration information.
- the first communication device may also adopt a specific division method to process other parts of data except the j-th part, that is, alternately use the foregoing division method and the specific division method to process the data.
- the first communication device may Different parts of the data are processed using sub-processing models corresponding to different segmentation methods.
- the sub-processing model corresponding to the i-th segmentation method is used for processing in the j-th part
- the sub-processing model corresponding to the i+1-th segmentation method is used for processing in the j+1-th segment. That is, the sub-processing model corresponding to each of the N types of segmentation methods is alternately used to process each part of the data.
- the data is processed by traversing each of the multiple partitioning methods.
- the first configuration information configures two division modes for the data processing model, the data part corresponding to the first division mode is an odd frame part, and the data part corresponding to the second division mode is an even frame part;
- the first communication The device uses the first splitting method to process the first frame of data, and the second splitting method to process the second frame of data.
- the first communication device continues to traverse the two splitting methods, that is, the third frame of data is processed through the first splitting method.
- the fourth frame of data is processed through the second segmentation method, and the above two segmentation methods are alternately used in this way until the last frame of the data to be processed.
- the processing resources of the first communication device that need to be occupied when the first communication device uses N partitioning methods for data processing can be compared with the current processing resources of the first communication device. Match the available processing resources.
- the first communication device may use the i-th segmentation method to process the j-th part of the data, that is, use different segmentation methods for processing in different parts of the data.
- the processing resources occupied when processing data in the N split manners match the available processing resources of the first communication device; the balance of resource occupation in the data processing process and the resource utilization rate in the data processing process are improved.
- the data processing model is usually a convolutional neural network (Convolutional Neural Networks, CNN) model for image processing.
- CNN convolutional Neural Networks
- FIG. 6 which is composed of different data processing layers.
- the CNN model includes input layer, convolutional layer 1 to convolutional layer 5, excitation layer 1 to excitation layer 5, pooling layer 1 and pooling layer 2, normalization layer 1 and normalization layer 2, full Connection layer 1 and fully connected layer 2, and output layer.
- different data processing layers have different processing delays and output data volumes. For example, the processing delay corresponding to the fully connected layer 2 of the CNN model shown in Fig. 6 is 10ms, and the output data volume is 0.1MB.
- the CNN model shown in FIG. 6 has four optional segmentation positions, including segmentation position 0 to segmentation position 3.
- the first communication device executes the output processing layer before the pooling layer 2 corresponding to segmentation location 2, and transmits the output data obtained by the pooling layer 2 to the second communication device.
- the second communication device can continue to execute the remaining data processing layer from the normalization layer 2 after the segmentation position 2 according to the output data sent by the first communication device, until the calculation of the entire CNN model is completed, and the image recognition result is obtained.
- the segmentation position 0 is a special segmentation position.
- the first communication device does not perform any calculations, and directly sends the collected raw data (such as captured pictures) to the second communication device, and the second communication device Perform calculations for all layers of CNN. Therefore, the division position 0 does not occupy the processing resources of the first communication device, but only occupies the processing resources of the second communication device.
- segmentation position 1 the first communication device only performs part of the calculation of the data processing layer, segmentation position 1 needs to occupy the processing resource size of the first communication device as the first processing resource amount 1, and the processing resource size occupied by the second communication device is the first Two processing resources 1.
- the processing resource size of the first communication device that needs to be occupied is the first processing resource amount 2
- the processing resource size of the second communication device is the second processing resource amount 2
- the processing resource size of the first communication device that needs to be occupied is the first processing resource amount 3
- the processing resource size of the second communication device that needs to be occupied is the second processing resource amount 3.
- the amount of available processing resources of the first communication device may be between the amount of processing resources of the first communication device occupied by the aforementioned four divided positions.
- the available processing resource size of the first communication device is the first available processing resource amount 1
- the available processing resource size of the second communication device is the second available processing resource amount 1.
- the available processing resource size of the communication device is the first available processing resource amount 2
- the available processing resource size of the second communication device is the second available processing resource amount 2. It can be seen from Figure 7 that the first available processing resource amount 1 at time 1 and the first available processing resource amount 2 at time 2 are both higher than the first processing resource amount 2 that needs to be occupied by the division position 2, and at the same time lower than the division position.
- At least one division method is applied to some parts of the data, that is, during the data processing process, the first communication device switches between different division positions, for example, the first communication
- the device applies segmentation position 2 for processing in some frames (such as even-numbered frames), and applies segmentation position 3 for processing in other partial frames (such as odd-numbered frames), so that the average amount of processing resources occupied by the first communication device is the same as that of the first communication device.
- the processing resources are matched. In this way, the processing resources of the first communication device can be fully utilized, and the data processing efficiency can be improved.
- the j-th part is determined by at least one of the following information:
- the time domain period of the division method the time domain offset of the division method, and the type of data information
- the time domain period is used to indicate the time domain period of the jth partial data
- the time domain offset is used to indicate the time domain position of the jth partial data within the time domain period
- the type information of the data information is used for Indicates that the j-th part is a critical data unit or a non-critical data unit.
- the time domain period can be understood as the time or space interval between repeated occurrences of the same thing.
- the time domain period specifically refers to the time domain period of the j-th partial data. To put it simply, it means that each of the N division methods is repeatedly applied once, and the time domain period of the processed data portion processed by the first communication device.
- the configuration information includes two division positions (corresponding to two division methods): division position 2 and division 3; among them, the time domain period for applying the division method is 2 frames, that is, the first communication device applies division position 2 and division position once. 3
- the processed data is 2 frames.
- the time-domain offset of the division mode refers to the time-domain position of the j-th part of the data in the time-domain period when the i-th division mode is applied to process the j-th part.
- the division position includes division position 2 and division 3, the time domain offset of division position 2 is 1, and the time domain offset of division position 3 is 0. That is, the first communication device may use the division position 2 for data processing in the first image frame of a time domain period, and use the division position 3 for data processing in the 0th image frame of a time domain period.
- the part of the data corresponding to each division method can also be determined by the type of data information.
- the type of data information is used to indicate whether the j-th part is a key data unit or a non-critical data unit, that is, the type of data information is used to indicate whether each part in the data is a key data unit or a non-critical data unit.
- the key data unit and the non-key data unit correspond to different partitioning methods.
- the key data unit is very important to the user and has privacy sensitivity, so it is not suitable for transmission directly through the wireless network, and it is not suitable for processing through other communication devices. Therefore, all key data units can be placed in the first communication device for processing, and non-critical data units can be sent to the second communication device for processing.
- the key data unit can use the preset key segmentation method
- the non-key data unit can use the preset non-key segmentation method for processing.
- the key data unit in the data can be processed using the sub-processing model corresponding to the key segmentation method
- the non-key data unit in the data can be processed using the sub-processing model corresponding to the non-critical segmentation method.
- obtaining the first configuration information in step 510 can be implemented through the following steps:
- Step 5101 The first communication device receives at least one piece of configuration information
- Step 5102 The first communication device determines first configuration information from at least one piece of configuration information.
- the first configuration information may be configured by the second communication device to the first communication device.
- the second communication device may send at least one piece of configuration information to the first communication device, and determine the first configuration information from the at least one piece of configuration information.
- the first communication device may also determine the first configuration information from the at least one piece of configuration information according to an instruction of the second communication device.
- the data processing method provided in the embodiment of the present application further includes:
- Step 5101a Receive scheduling signaling
- the scheduling signaling is used to indicate the second configuration information among the multiple configuration information.
- Step 5101b Determine second configuration information from at least one piece of configuration information based on the scheduling information.
- the second configuration information is similar to the first configuration information, and both are used to configure the N division modes and the j-th part corresponding to the i-th division mode among the N division modes.
- the second configuration information may be the same as the first configuration information, or may be different from the first configuration information.
- the embodiments of the application are not limited here.
- the second communication device may determine one for the first communication device according to the available processing resources of the first communication device, or the state of the wireless channel between the first communication device and the second communication device The second configuration information is more suitable for the current wireless channel state and the available processing resource size.
- the second communication device may carry the second configuration information in the scheduling information and send it to the first communication device.
- the first communication device selects the second configuration information from a plurality of configuration information, and based on the first configuration information. Second, configure the information for data processing.
- the first communication device can determine the amount of data transmitted on the wireless channel according to the quality of the wireless channel status. When the wireless channel status is good, it can use the multiple configuration information to transmit the data volume on the wireless channel. When the wireless channel status is poor, the configuration information with a small amount of data transmitted on the wireless channel can be used.
- the output data volume of the segmentation position 1 is the sum of the output data volume of the convolutional layer 1, the excitation layer 1 and the pooling layer 1, and the output data volume of the segmentation position 3 is The sum of the output data of all data processing layers on the left side of the excitation layer 4. Therefore, the output data amount of the division position 1 is smaller than the output data amount of the division position 3. If the first communication device uses segmentation position 3 for data processing, when the second communication device detects that the current wireless channel status is poor, the current segmentation method can be reconfigured, and segmentation position 3 in the current configuration information can be modified to segmentation position 2. In this way, the amount of data transmitted in the wireless channel is reduced.
- the specific division manner may be a factory-configured division manner, or may be a division manner that is semi-statically configured by the second communication device.
- the data transmission volume of the configuration information can be reduced, and the transmission efficiency of the configuration information can be improved.
- the first communication device may also determine the first configuration information locally based on the available processing resource information.
- the specific determination method is the same as the method in which the second communication device determines the first configuration information based on the available processing resource information. For details, please refer to the following description.
- the data processing method provided in this application further includes:
- Step 501 Send available processing resource information to the second communication device.
- the available processing resource information includes at least one of the following: the amount of available computing resources of the first communication device, the amount of available storage resources, and the amount of available power.
- the first communication device may send the available processing resource information of the first communication device to the second communication device; in this way, after the second communication device receives the available processing resource information, the second communication device is The first communication device matches a division mode suitable for the currently available processing resources of the first communication device and data information corresponding to each division mode.
- step 501 further includes the following steps after sending the available processing resource information to the second communication device:
- Step 502 If the available processing resource information changes, obtain updated available processing resource information
- Step 503 Send the updated available processing resource information to the second communication device.
- the first communication device when the available resource processing resources of the first communication device change, the first communication device needs to switch from one type of configuration information to another type of configuration information. Based on this, the first communication device needs to send the updated available processing resource information to the second communication device. In this way, after receiving the updated available processing resource information, the second communication device can re-match the available processing resource information for the first communication device. It is applicable to the division mode of the currently available processing resources of the first communication device and the data information corresponding to each division mode.
- the first communication device is a terminal device
- the second communication device is a network device.
- the first communication device may obtain the first configuration information from the following information:
- MAC CE Media Access Control Element
- control channel information includes Downlink Control Information (DCI).
- DCI Downlink Control Information
- the embodiment of this application may send the at least one configuration information through a dedicated DCI format of DCI, or use a dedicated radio network temporary identifier (RNTI Radio For the generation of Network Tempory Identity (RNTI), the high-level configuration signaling can be Radio Resource Control (RRC) signaling.
- RRC Radio Resource Control
- the first communication device can use the i-th segmentation method to process the j-th part of the data, that is, use different segmentation methods for processing in different parts of the data, so that the The processing resources occupied when the data is processed in the N division modes are matched with the available processing resources of the first communication device; the balance of resource occupation in the data processing process and the resource utilization rate in the data processing process are improved.
- the data processing method includes step 810 to step 820.
- Step 810 Determine at least one piece of configuration information.
- each configuration information in at least one piece of configuration information is used to configure N partitioning methods when processing the data to be processed, the N parts into which the data is divided, and the j-th part corresponding to the i-th partitioning method;
- N is greater than Or an integer equal to 1, i is greater than or equal to 1 and less than or equal to N, and j is greater than or equal to 1 and less than or equal to N;
- the N division modes are division modes in which the data processing model is divided into at least two sub-processing models by preset division positions.
- Step 820 Send at least one piece of configuration information to the first communication device, where the at least one piece of configuration information includes the first configuration information.
- the second communication device is a network device
- at least one piece of configuration information is carried in at least one of the following information and sent to the first communication device:
- MAC CE Media Access Control Element
- control channel information includes Downlink Control Information (DCI).
- DCI Downlink Control Information
- the embodiment of this application may send the at least one configuration information through a dedicated DCI format of DCI, or use a dedicated radio network temporary identifier (RNTI Radio For the generation of Network Tempory Identity (RNTI), the high-level configuration signaling can be Radio Resource Control (RRC) signaling.
- RRC Radio Resource Control
- the j-th part is determined by at least one of the following information:
- the time domain period of the division method the time domain offset of the division method, and the type of data information
- the time domain period is used to indicate the time domain period of the jth part of the data; the time domain offset is used to indicate the time domain position of the jth part in the time domain period; the type of data information is used to indicate the jth part Part is a critical data unit or a non-critical data unit.
- step 810 the determination of at least one piece of configuration information in step 810 can be implemented through step 8101 and step 8102:
- Step 8101 Receive available processing resource information sent by the first communication device.
- Step 8102 Determine at least one piece of configuration information based on the available processing resource information.
- the second communication device can match the first communication device with N types of division methods that are suitable for the current available processing resources of the first communication device and The j-th part corresponding to each of the N division modes.
- the processing resources of the first communication device occupied by the N division modes matched by the second communication device match the available processing resources of the first communication device.
- the processing resources of the first communication device occupied by the N division methods can match the available processing resources.
- the processing resources of the first communication device occupied by the N division methods are equal to the available processing resources, or It is that the absolute value of the difference between the processing resources of the first communication device and the available processing resources occupied by the N division methods is smaller than the preset threshold.
- the embodiments of the application are not limited here.
- the second communication device can determine at least one piece of configuration information for the first communication device according to the available processing resources of the first communication device, and configure the N types of division methods in the data processing process through the configuration information and Data information of at least one division method.
- N types of division methods in the data processing process through the configuration information and Data information of at least one division method.
- the data can be processed by Some data parts apply N types of division methods, so that the processing resources occupied when the N types of division methods are used to process the data to be processed match the available processing resources of the first communication device; the balance of resource occupation in the data processing process is improved and Resource utilization during data processing.
- step 8102 determining at least one piece of configuration information based on the available processing resource information can be achieved through the following steps:
- Step 8102a The second communication device determines the initial parameters corresponding to the preset M division modes; M is an integer greater than or equal to 2; each initial parameter is a natural number;
- step 8102b the second communication device continuously adjusts the initial parameters based on the preset processing resource amounts respectively occupied by the preset M partitioning methods to obtain the average amount of processing resources occupied by the preset M partitioning methods;
- Step 8102c When the average amount of processing resources matches the available processing resources, the second communication device determines a target parameter with a non-zero value
- Step 8102d The second communication device determines N segmentation methods corresponding to the target parameter from the preset M segmentation methods
- Step 8102e Determine the N types of segmentation methods and the target parameters corresponding to the N types of segmentation methods as one piece of configuration information, and then obtain at least one piece of configuration information.
- M is used instead of the number of preset division modes, and M is an integer greater than or equal to 2.
- the second communication device obtains all the preset division modes, and determines initial parameters for each preset division mode, where each initial parameter is a natural number.
- the initial parameters here can represent the initial time domain offset of each segmentation method.
- the average amount of processing resources occupied by the M division methods is calculated by continuously adjusting the values of the initial parameters.
- the second communication device sums the product of the initial parameters corresponding to the M division modes and the amount of processing resources occupied by each division mode, and quotients the sum result with the sum of the current values of the M initial parameters;
- the result of the quotient is the average amount of processing resources occupied by the M division methods.
- the target parameter can indicate the time domain offset corresponding to each segmentation method.
- the segmentation method corresponding to the non-zero-valued parameter is the segmentation method to be applied by the data processing model; and the size of the non-zero-valued target parameter may indicate the time-domain offset of the applied segmentation method.
- the second communication device determines two partitioning methods corresponding to the target parameter from the preset M partitioning methods; if the two partitioning methods include non-segmentation The data processing model (ie, the division method corresponding to the division position 0 shown in Figure 6), or the two division methods include a specific division method, the first communication device may divide the data processing model in a non-divided manner or a specific division The method is removed, and the configuration information is generated only according to another segmentation method and the target parameter value corresponding to the segmentation method. In this way, the generated configuration information contains only one division method, that is, N is 1.
- the second communication device determines two or more division modes corresponding to the target parameter from the preset M division modes; and combines at least two division modes And the target parameter is determined as a configuration information.
- the generated configuration information contains multiple division methods, that is, N is greater than 1.
- the initial parameters of each division mode are set to a, b, c; the amount of processing resources occupied by the first division mode is X, and the second division mode is The amount of processing resources occupied is Y, and the amount of processing resources occupied by the third division method is Z; the second communication device can adjust the values of the initial parameters a, b, and c; calculate different initial parameters a by formula (1), Under the values of b and c, the average amount of processing resources occupied by these three partitioning methods:
- the average amount of processing resources occupied by the three division methods it is necessary to determine whether the calculated average amount of processing resources matches the amount of available processing resources of the first communication device, and if they do not match, continue to adjust The value of the initial parameter; if it matches, the non-zero value in the current initial parameter is determined, and the non-zero value parameter is determined as the target parameter.
- To determine whether the calculated average amount of processing resources matches the available processing resources of the first communication device can be determined by determining whether the difference between the average amount of processing resources and the available processing resources of the first communication device is less than a threshold.
- an image recognition task is taken as an example to describe the above-mentioned method of determining configuration information.
- the first available processing resource amount 1 which is just the average value between the first processing resource amount 2 and the first processing resource amount 3; at this time, the division position 2 and the division position 3 can be alternately used. To achieve matching with the first available processing resource amount 1.
- the division position 2 can be used in odd-numbered frames, and the division position 3 can be used in even-numbered frames.
- the amount of processing resources occupied by the first communication device is the first processing resource amount 2
- the amount of processing resources occupied by the first communication device is the first processing resource amount 3
- the average of the two is exactly equal to the first processing resource amount.
- the time domain period of the configurable division position is 2 frames, and the time domain offset of the division position 2 is 1.
- the time domain offset of division position 3 is 3. That is, a frame with a time domain offset of 1 adopts the division position 2 and a frame with a time domain offset of 3 adopts the division position 3. In this way, the average amount of processing resources occupied by the first communication device is exactly equal to the first available processing resource amount 1.
- the available processing resource amount of the first communication device is the first available processing resource amount 2 in FIG. 7, which is less than the first available processing resource amount 1, the average alternating mode in the previous example is still used. The amount of available processing resources of the first communication device is exceeded, and the first communication device cannot complete the image recognition task. At this time, the uneven alternation between the division position 2 and the division position 3 may be used to achieve the matching of the processing resource amount with respect to the first available processing resource amount 2.
- the division position 2 can be used in two of every three frames, and the division position 3 can be used in one of every three frames; in this way, the frames with the time domain offset of 1 and 2 occupy the first communication
- the processing resource amount of the device is the first processing resource amount 2
- the processing resource amount occupied by the first communication device is the first processing resource amount 3
- the final average value is exactly equal to the first available Processing resources 2.
- the first communication device alternately uses the division position 2 and the division position 3 for resource processing, so that the processing resources occupied by the first communication device in the process of data processing based on the above-mentioned method match the available processing resources of the first communication device, so, The computing power of the first communication device can be fully utilized, and the data processing efficiency can be improved.
- the configuration information example of the above-mentioned alternately using the division position 2 and the division position 3 is shown in Table 2.
- the time domain period of the configurable division position (that is, the division method) is 3 frames, and the time domain offset of the division position 2 is 1 or 2.
- the time domain offset of division position 3 is 3. That is, the frame with the time domain offset of 1 or 2 adopts the division position 2, and the frame with the time domain offset of 3 adopts the division position 3. In this way, the average amount of processing resources occupied by the first communication device is exactly equal to the first amount of available processing resources 2.
- Time domain offset Split position 2 3 frames 1,2 Split position 3 3 frames 3
- the configuration information may also be determined by the type of data information.
- step 810 determines at least one piece of configuration information, which can be implemented through step 8101' and step 8102':
- Step 8101' Obtain the type of data information
- Step 8102' based on the type of data information, determine whether the j-th part of the M parts of the data is a key data unit or a non-key data unit;
- Step 8103' if the j-th part is a key data unit, determine the key segmentation mode corresponding to the j-th part;
- Step 8103' if the j-th part is a non-critical data unit, determine that the j-th part corresponds to a non-critical segmentation mode;
- Step 8103' Determine a piece of configuration information based on the division mode corresponding to the j-th part of the M parts of the data.
- the second communication device may determine whether each part of the data is a critical data unit or a non-critical data unit according to the type of data information.
- the key data unit is very important to the user and has privacy sensitivity, so it is not suitable for direct transmission through the wireless network. All the key data units can be placed in the first communication device for processing, and the non-critical data units can be sent to the second communication device for processing.
- the key data unit can use the preset key segmentation method
- the non-key data unit can use the preset non-key segmentation method for processing.
- the second communication device may use the key segmentation method for the key data unit of the data, and use the non-critical segmentation method for the non-critical data unit of the data.
- the second communication device can judge each part of the data to determine whether each part belongs to non-critical data or a key data unit. If the jth part is a key data unit, then determine the jth part. Each part is processed using the preset key segmentation method. If the j-th part is a non-critical data unit, it is determined that the j-th part is processed using a preset non-critical division method.
- the data part in this way of determining at least one piece of configuration information by the type of data information, can be divided into two parts, that is, the value of M is 2; the first part includes key data units, and the second part Including non-critical data units.
- the first division method is the above-mentioned key division method
- the second division method is the non-key division method.
- an image recognition task is taken as an example to describe the above-mentioned method of determining configuration information.
- images may be divided into key frames (that is, pictures containing the original information of the complete image) and non-key frames (that is, only the incremental information or prediction information of the image).
- key frame contains the original image information, it is sensitive to privacy and is not suitable for transmission on the network.
- the non-key frame does not contain the original information of the image, and even if it is transmitted in the wireless channel, it will not leak private information. Therefore, a part of the segmentation method calculated on the terminal side can be used for key frames (as shown in the segmentation position 1 in Figure 6), and for non-key frames, the segmentation mode calculated on the second communication network can be used (as shown in Figure 6).
- Split position 0 shows
- the key frame can be processed by the key segmentation method to avoid transmitting the key frame carrying the original image data in the wireless channel.
- the remaining non-key frames can be divided into non-key frames, that is, they are all processed by the second communication device, so as to save the computing resources of the first communication device. In this way, while protecting the data privacy of the first communication device, the processing capability of the second communication device can also be fully utilized, saving valuable processing resources of the first communication device.
- the above configuration information example of alternately using key segmentation and non-critical segmentation is shown in Table 3.
- the time domain period of the configurable segmentation method is N frames
- the time domain offset of the key segmentation method is 0,
- the non-critical segmentation method is used.
- the time domain offset of the mode is non-zero frame.
- the first configuration information may include: initial configuration information; the first initial configuration information includes a first division manner and a length of a division data unit corresponding to the first division manner.
- the second communication device may send initial configuration information to the first communication device, where the initial configuration information may only include information related to one split mode, that is, only include the first split mode and the first split mode.
- step 830 after step 830 sends at least one piece of configuration information to the first communication device, it may further include the following steps:
- Step 840 Send scheduling information to the first communication device; the scheduling information is used to indicate second configuration information among the multiple configuration information.
- the second communication device when the available resource processing resources of the first communication device change, the second communication device needs to instruct the first communication device to switch from one type of configuration information to another type of configuration information.
- the second communication device may send scheduling information to the first communication device, specifically instructing the first communication device to select the second configuration information from a plurality of configuration information for data processing.
- the second communication device may determine the second configuration information based on the wireless channel state with the first communication device and/or the updated available processing resource information sent by the first communication device;
- the second configuration information generates scheduling information.
- the second communication device may determine a more suitable current wireless channel for the first communication device based on the available processing resources of the first communication device or the state of the wireless channel between the first communication device and the second communication device Status, and target configuration information for the size of available processing resources.
- the first communication device can determine the amount of data transmitted on the wireless channel according to the quality of the wireless channel status. When the wireless channel status is good, it can use the multiple configuration information to transmit the data volume on the wireless channel. When the wireless channel status is poor, the configuration information with a small amount of data transmitted on the wireless channel can be used.
- the second communication device after receiving the scheduling signaling, reconfigures the at least two division modes and the length of the divided data unit based on the second configuration information indicated by the scheduling signaling. And perform data processing based on the second configuration information.
- the second communication device can determine at least one piece of configuration information for the first communication device according to the available processing resources of the first communication device, and configure the N types of division methods in the data processing process through the configuration information and Data information of N kinds of division methods.
- N kinds of division methods are only applied to some parts of the data, and the same division method is not required to be used on all the data. In this way, it can be applied to certain data parts of the data.
- N kinds of partitioning methods so that the processing resources occupied when the at least one partitioning method is used to process the data to be processed matches the available processing resources of the first communication device; the balance of resource occupation in the data processing process and the data processing process are improved Resource utilization.
- the embodiment of the present application provides a data processing method. As shown in FIG. 12, the data processing method provided by the embodiment of the present application specifically includes the following steps:
- Step 1201 The first communication device sends available processing resource information to the second communication device.
- the available processing resource information includes at least one of the following: the amount of available computing resources of the first communication device, the amount of available storage resources, and the amount of available power.
- Step 1202 the second communication device determines at least one piece of configuration information based on the available processing resource information.
- Step 1203 The second communication device sends at least one piece of configuration information to the first communication device.
- At least one piece of configuration information includes the first configuration information.
- the first configuration information is used to configure the N division modes and the j-th part corresponding to the i-th division mode among the N division modes; where N is an integer greater than or equal to 1, and i is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1, and the N division modes are division modes in which the data processing model is divided into at least two sub-processing models by preset division positions.
- Step 1204 The first communication device uses the sub-processing model corresponding to the i-th segmentation method to process the j-th part of the data.
- the division method may divide the data processing model into a first part of the sub-processing model and a second part of the sub-processing model.
- the first part of the sub-processing model is operated by the first communication device; the second part of the sub-processing model is operated by the second communication device.
- the first part of the sub-processing model and the second part of the sub-processing model obtained are different.
- the first communication device processes the data to be processed according to the first partial sub-processing model corresponding to each of the N division modes.
- the first communication device alternately uses the first partial sub-processing models corresponding to the N division modes to process the data.
- Step 1205 The first communication device obtains the first output data obtained when the first sub-processing model corresponding to each segmentation mode is applied for data processing.
- the first communication device processes the to-be-processed data according to the first partial sub-processing model corresponding to the at least one segmentation manner to obtain the intermediate processing result, that is, the first output data.
- Step 1206 The first communication device sends the first output data to the second communication device.
- the first communication device may send the first output data outputted by the data processing model to the second communication device via the wireless link, and the first communication device may also send multiple output data after obtaining multiple first output data.
- Sending to the second communication device is not limited in the embodiment of the present application.
- Step 1207 The second communication device processes the first output data based on the first output data and the sub-processing model corresponding to the i-th segmentation mode to obtain a data processing result.
- the second communication device after receiving the first output data, processes the data to be processed according to the second partial sub-processing model corresponding to at least one division method. Get the final data processing result.
- the second communication device alternately uses the second partial sub-processing models corresponding to the N division modes to process the data.
- the first communication device can only apply N splitting methods to certain data parts in the data to be processed, and does not need to use the same splitting on all the data.
- N splitting methods to certain parts of the data
- the processing resources occupied when processing data using the N splitting methods match the available processing resources of the first communication device; improving the data processing process Balance of resource occupancy and resource utilization during data processing.
- the embodiment of the present application provides a communication device 13, which corresponds to a data processing method applied to the first communication device;
- Composition structure diagram one, the communication device 13 may include:
- the first obtaining unit 1301 is configured to obtain first configuration information; the first configuration information is used to configure N division modes and the j-th part corresponding to the i-th division mode among the N division modes; wherein, N is an integer greater than or equal to 1, i is greater than or equal to 1 and less than or equal to N, j is greater than or equal to 1 and less than or equal to M; M is an integer greater than 1;
- the N types of segmentation methods are segmentation methods in which the data processing model is segmented into at least two sub-processing models by preset segmentation positions.
- the first communication device further includes a first processing unit
- the first processing unit is configured to use the sub-processing model corresponding to the i-th segmentation mode to process the data of the j-th part.
- the j-th part is determined by at least one of the following information:
- the time domain period of the division method the time domain offset of the division method, and the type of data information
- the time domain period is used to indicate the time domain period of the jth partial data
- the time domain offset is used to indicate the time domain position of the jth part within the time domain period
- the type of data information is used to indicate that the j-th part is a key data unit or a non-key data unit.
- the communication device includes a first receiving unit and a first determining unit
- the first receiving unit is configured to receive at least one piece of configuration information
- the first determining unit is configured to determine the first configuration information from the at least one piece of configuration information.
- the first receiving unit is configured to receive the sent scheduling signaling
- the first processing unit is configured to determine second configuration information from at least one piece of configuration information according to the scheduling information.
- the first sending unit 1303 is configured to send available processing resource information to the second communication device.
- the available processing resource information includes at least one of the following: the amount of available computing resources, the amount of available storage resources, and the amount of available power of the first communication device.
- the communication device further includes a sending unit
- the first obtaining unit 1301 is configured to obtain updated available processing resource information if the available processing resource information changes;
- the first sending unit is configured to send the updated available processing resource information to the second communication device.
- the communication device proposed in the application embodiment can only apply N types of segmentation methods to certain parts of the data, and does not need to use the same segmentation method for all the data; in this way, it can be used in certain parts of the data.
- Some parts apply N kinds of division methods, so that the processing resources occupied when using the N kinds of division methods to process the data to be processed match the available processing resources of the first communication device; improve the balance of resource occupation and data processing in the data processing process Resource utilization in the process.
- FIG. 14 is a schematic diagram of the composition structure of a communication device provided by an embodiment of the application Second, the communication device 14 may include:
- the second determining unit 1401 is configured to determine at least one piece of configuration information; wherein, each piece of configuration information in the at least one piece of configuration information is used to configure N kinds of division modes, N parts into which data is divided, and an i-th division mode
- N is an integer greater than or equal to 1
- i is greater than or equal to 1 and less than or equal to N
- j is greater than or equal to 1 and less than or equal to M
- M is an integer greater than 1.
- the second sending unit 1402 is configured to send the at least one piece of configuration information to the first communication device; the at least one piece of configuration information includes the first configuration information.
- the j-th part is determined by at least one of the following information:
- the time domain period of the division mode the time domain offset of the division mode, and the type of the data information
- the time domain period is used to indicate the time domain period of the jth partial data; the time domain offset is used to indicate the time domain position of the jth part within the time domain period;
- the type of the data information is used to indicate that the j-th part is a key data unit or a non-key data unit.
- the communication device further includes a second receiving unit
- the second receiving unit is configured to receive available processing resource information sent by the first communication device
- the second determining unit 1401 is configured to determine the at least one piece of configuration information based on the available processing resource information.
- the second determining unit is configured to determine the initial parameters corresponding to the preset M division modes; M is an integer greater than or equal to 2; each of the initial parameters is a natural number; Based on the preset processing resources occupied by the preset M partitioning methods, the initial parameters are continuously adjusted to obtain the average processing resource occupied by the preset M partitioning methods; when the average processing resource When the available processing resources match, determine the target parameter with a non-zero value; from the preset M partitioning methods, determine the N partitioning methods corresponding to the target parameter; combine the N partitioning methods with the The target parameters corresponding to the N division modes are determined as one piece of configuration information, and then the at least one piece of configuration information is obtained.
- the communication device further includes a second acquiring unit
- the second acquiring unit is configured to acquire the type of data information
- the second determining unit is configured to determine, based on the type of the data information, that the jth part of the M parts of data is a key data unit or a non-key data unit; if the jth part is a key data unit, Determine that the j-th part corresponds to a key segmentation method; if the j-th part is a non-critical data unit, determine that the j-th part corresponds to a non-critical segmentation method; based on the data corresponding to the j-th part of the M parts Split mode, determine a piece of configuration information.
- the second sending unit 1402 is configured to send scheduling information to the first communication device; the scheduling information is used to indicate the second configuration information in the plurality of configuration information.
- the communication device further includes a generating unit
- the second determining unit is configured to determine second configuration information based on a wireless channel state with the first communication device and/or updated available processing resource information sent by the first communication device;
- the generating unit is configured to generate the scheduling information based on the second configuration information.
- the at least one piece of configuration information is carried in at least one of the following information and sent to the first communication device:
- Control channel information media access control control unit MAC CE information, high-level configuration signaling, and application layer control information.
- the communication device proposed in the embodiment of the present application can determine at least one piece of configuration information for the first communication device according to the available processing resources of the first communication device, and configure the corresponding N split modes and N split modes in the data processing process through the configuration information. In this way, only some parts of the data are applied with N kinds of division methods, and the same division method does not need to be used on all the data. In this way, N kinds of division methods can be applied to some data parts of the data to be processed.
- the division method makes the processing resources occupied when using the N division methods to process the data to be processed match the available processing resources of the first communication device; it improves the balance of resource occupation in the data processing process and the resource utilization in the data processing process Rate.
- FIG. 15 is a schematic diagram of the hardware composition structure of a terminal according to an embodiment of the application.
- the communication device includes a first transceiver 1501, a first processor 1502, and a first memory 1503 storing a computer program.
- the first processor 1502 is configured to execute the method steps in the embodiment corresponding to FIG. 5 and FIG. 12 when running the computer program.
- bus system 1504. is used to implement connection and communication between these components.
- the bus system 1504 also includes a power bus, a control bus, and a status signal bus.
- various buses are marked as the bus system 1504 in Fig. 15.
- FIG. 16 is a schematic diagram of the hardware composition structure of a terminal according to an embodiment of the application.
- the communication device includes a second transceiver 1601, a second processor 1602, and a second memory 1603 storing a computer program.
- the second processor 1602 when configured to run the computer program, it executes the method steps in the aforementioned embodiment corresponding to FIG. 8 and FIG. 12.
- bus system 1604 is used to implement connection and communication between these components.
- bus system 1604 also includes a power bus, a control bus, and a status signal bus.
- various buses are marked as the bus system 1604 in FIG. 16.
- non-volatile memory can be Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (Erasable Programmable Read-Only Memory) , EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Magnetic Random Access Memory (FRAM), Flash Memory, Magnetic Surface Memory , CD-ROM, or CD-ROM (Compact Disc Read-Only Memory, CD-ROM); magnetic surface memory can be disk storage or tape storage.
- ROM Read Only Memory
- PROM Programmable Read-Only Memory
- EPROM Erasable Programmable Read-Only Memory
- EEPROM Electrically Erasable Programmable Read-Only Memory
- FRAM Magnetic Random Access Memory
- Flash Memory Magnetic Surface Memory
- CD-ROM Compact Disc Read-Only Memory
- magnetic surface memory can be disk storage or tape storage.
- the volatile memory may be a random access memory (Random Access Memory, RAM), which is used as an external cache.
- RAM Random Access Memory
- SRAM static random access memory
- SSRAM synchronous static random access memory
- dynamic random access memory dynamic random access memory
- DRAM Dynamic Random Access Memory
- SDRAM Synchronous Dynamic Random Access Memory
- DDRSDRAM Double Data Rate Synchronous Dynamic Random Access Memory
- ESDRAM Double Data Rate Synchronous Dynamic Random Access Memory
- ESDRAM Enhanced Synchronous Dynamic Random Access Memory
- SLDRAM synchronous connection dynamic random access memory
- SLDRAM synchronous connection dynamic random access memory
- DDRRAM direct memory bus random access memory
- DRRAM Direct Rambus Random Access Memory
- the methods disclosed in the foregoing embodiments of the present application may be applied to a processor or implemented by a processor.
- the processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in the form of software.
- the aforementioned processor may be a general-purpose processor, DSP, or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the like.
- the processor can implement or execute the methods, steps, and logical block diagrams disclosed in the embodiments of the present application.
- the general-purpose processor may be a microprocessor or any conventional processor or the like.
- the software module may be located in a storage medium, and the storage medium is located in a memory.
- the processor reads the information in the memory and completes the steps of the foregoing method in combination with its hardware.
- the embodiment of the present application also provides a computer storage medium, which is specifically a computer-readable storage medium.
- Computer instructions are stored thereon.
- the computer storage medium is located in the terminal, when the computer instructions are executed by the processor, any step in the above-mentioned data processing method in the embodiment of the present application is implemented.
- the disclosed device and method may be implemented in other ways.
- the device embodiments described above are only illustrative.
- the division of the units is only a logical function division, and there may be other divisions in actual implementation, such as: multiple units or components can be combined, or It can be integrated into another system, or some features can be ignored or not implemented.
- the coupling, or direct coupling, or communication connection between the components shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms. of.
- the units described above as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, 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 objectives of the solutions of the embodiments.
- the functional units in the embodiments of the present application can be all integrated into one processing unit, or each unit can be individually used as a unit, or at least two units can be integrated into one unit; the above-mentioned integrated units are both It can be implemented in the form of hardware, or in the form of hardware plus software functional units.
- the foregoing program can be stored in a computer readable storage medium. When the program is executed, it is executed. Including the steps of the foregoing method embodiment; and the foregoing storage medium includes: various media that can store program codes, such as a mobile storage device, ROM, RAM, magnetic disk, or optical disk.
- the aforementioned 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 may also be stored in a computer readable storage medium.
- the computer software product is stored in a storage medium and includes several instructions for A computer device (which may be a personal computer, a server, or a network device, etc.) is allowed to execute all or part of the methods described in the various embodiments of the present application.
- the aforementioned storage media include: removable storage devices, ROM, RAM, magnetic disks, or optical disks and other media that can store program codes.
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本申请实施例提供一种数据处理方法,应用于第一通信设备,所述方法包括:获取第一配置信息;所述第一配置信息用于配置N分割方式,以及所述N种分割方式中第i第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。本申请实施例同时提供一种通信设备和计算机存储介质。
Description
本申请涉及通信技术领域,尤其涉及一种数据处理方法及通信设备、计算机存储介质。
随着计算机科学技术的发展,基于人工智能(Artificial Intelligence,AI)/机器学习(Machine Learning,ML)模型的数据处理技术在终端中承担越来越重要的任务,例如基于AI/ML模型的图像识别任务、视频通话处理任务、增强现实或虚拟现实任务等。可以预见,在未来移动通信网络(例如,5G网络)中传输AI/ML模型的数据,将成为一种重要的业务。
目前,可以通过终端与网络设备或者其他终端合作,来完成AI/ML模型的数据处理;具体地,可以将AI/ML模型分割为多个子模型,终端通过多个子模型中的部分子模型进行数据处理,网络设备或者其他终端通过多个子模型中的另一部分子模型进行数据处理。然而,现有技术中AI/ML模型的分割方式是固定的,也就是说,待处理数据中所有的数据都通过该固定分割方式进行处理,这就造成在某些情况下,终端的可用处理资源并不能够满足该固定分割方式所要占用的资源大小;因此,通过固定分割方式进行的数据处理会产生资源占用不均衡的问题。
发明内容
本申请实施例提供了一种数据处理方法及通信设备、计算机可读存储介质,提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
本申请实施例提供一种数据处理方法,应用于第一通信设备中,包括:
获取第一配置信息;所述第一配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;
所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式;
本申请实施例提供一种数据处理方法,应用于第二通信设备中,包括:
确定至少一个配置信息;其中,所述至少一个配置信息中的每个配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;
所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式;
向所述第一通信设备发送至少一个配置信息,所述至少一个配置信息中包含第一配置信息。
本申请实施例提供了一种第一通信设备,所述通信设备包括:
第一获取单元,配置为获取第一配置信息;所述第一配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;
所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
本申请实施例提供了一种第二通信设备,所述通信设备包括:
第二处理单元,,配置为确定至少一个配置信息;其中,所述至少一个配置信息中的每个配置信息用于配置N种分割方式,数据被分成的N个部分,及第i种分割方式对应的第j个部分;N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数
第二发送单元,配置为向所述第一通信设备发送所述至少一个配置信息;所述至少一个配置信 息中包含第一配置信息。
本申请实施例提供了一种第一通信设备,所述通信设备包括:
所述通信设备包括:第一收发器、第一处理器和存储有计算机程序的第一存储器;
所述第一收发器、所述第一处理器和所述第一存储器之间通过通信总线进行通信;
所述第一处理器,配置为结合所述第一收发器,运行所述第一存储器中存储的所述计算机程序时,执行上述所述的应用于第一通信设备中的数据处理方法。
本申请实施例提供了一种第二通信设备,所述通信设备包括:
第二收发器、第二处理器和存储有计算机程序的第二存储器;
所述第二收发器、所述第二处理器和所述第二存储器之间通过通信总线进行通信;
所述第二处理器,配置为结合所述第二收发器,运行所述第二存储器中存储的所述计算机程序时,执行上述所述的应用于第二通信设备中的数据处理方法。
本申请实施例还提供了一种计算机可读存储介质,上存储有计算机程序,所述计算机程序被第一处理器执行实现所述数据处理方法的步骤;或者;所述计算机程序被第二处理器所述数据处理方法的步骤。
本申请实施例提供的数据处理方法及通信设备、计算机存储介质,第一通信设备能够根据配置信息,配置数据处理模型的N种分割方式以及N种分割方式中第i种分割方式对应的第j个部分。也就是说,第一通信设备可以采用第i种分割方式对数据的第j个部分进行处理,即在数据的不同部分采用不同的分割方式进行处理,如此,使得采用该N种分割方式处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
图1为本申请实施例提供的一种通信网络数据传输场景示意图;
图2为本申请实施例提供的一种AI/ML操作卸载的网络架构示意图;
图3为本申请实施例提供的一种终端与网络设备合作进行AI/ML模型操作的网络架构示意图;
图4为本申请实施例提供的一种通信***的网络架构示意图;
图5为本申请实施例提供的一种数据处理方法流程示意图一;
图6为本申请实施例提供的一种CNN模型结构示意图;
图7为本申请实施例提供的一种分割方式切换示意图一;
图8为本申请实施例提供的一种数据处理方法流程示意图二;
图9为本申请实施例提供的一种分割方式切换示意图二;
图10为本申请实施例提供的一种分割方式切换示意图三;
图11为本申请实施例提供的一种分割方式切换示意图四;
图12为本申请实施例提供的一种数据处理方法流程示意图三;
图13为本申请实施例提供的一种通信设备的结构示意图一;
图14为本申请实施例提供的一种通信设备的结构示意图二;
图15为本申请实施例提供的一种通信设备的硬件结构示意图一;
图16为本申请实施例提供的一种通信设备的硬件结构示意图二。
为使本申请的目的、技术方案和优点更加清楚,下面将结合附图对本申请实施方式作进一步地详细描述,所附附图仅供参考说明之用,并非用来限定本发明实施例。
需要说明的是,本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别不同对象,而不是用于描述特定顺序。此外,术语“包括”和“具有”以及它们任何变形,意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、***、产品或设备没有限定于已列出的步骤或单元,而是可选地还包括没有列出的步骤或单元,或可选地还包括对于这些过程、方法、产品或设备固有的其他步骤或单元。
参考图1所示的通信网络数据传输场景示意图,在未来的通信网络中,能够涉及以下应用场景:增强的移动宽带通信(eMBB),大规模机器类通信(mMTC),超可靠低时延通信(URLLC),以及 AI/ML应用。其中,eMBB能够支持终端与网络设备(例如5G云端或者5G边缘计算设备)之间的大数据流量传输,例如视频、音频等多媒体数据在终端与网络设备之间的传输。另外,mMTC/URLLC是指支持大量终端与网络设备之间的低时延与高可靠的数据传输,例如远程传感器、智能汽车等终端与网络设备的传感数据以及控制指令的传输;AI/ML应用能够支持AI/ML模型在终端和网络设备之间的传输。
实际应用中,对于终端来说,AI/ML应用面临三大挑战,具体如下:
挑战一、终端缺乏完全在本地运行AI/ML模型所需的计算资源、存储资源和电池容量;
挑战二、终端在本地执行AI/ML应用时,如何在多变的AI/ML数据业务以及网络环境下实时获取所需的AI/ML模型;
挑战三、终端如何参与AI/ML模型的训练。
针对挑战一,第三代合作伙伴计划(3
rd Generation Partnership Project,3GPP)已经设计了解决方案,即将AI/ML模型全部卸载到网络侧(例如,5G云端或5G边缘)进行操作。参考图2所述的AI/ML操作卸载的网络架构示意图,在该网络结构中,网络设备需要执行AI/ML模型的操作,还需要作为控制节点对终端发送的数据进行处理,以及向终端发送控制指令。
在未来的通信网络中需要极低的终端到终端的环回时延;将AI/ML模型全部卸载到网络侧进行操作,并且要达到毫秒级别的环回时延,不仅需要终端和网络设备支持URLLC,还需要无所不在的MEC部署以及完全的网络覆盖,然而,这些需求对于使用毫米波频段的5G网络是很难实现的,因此在终端本地进行AI/ML模型的操作是必不可少的。另外,将AI/ML模型全部卸载到网络侧有可能带来隐私泄露的风险,这是因为将很多终端将本地数据上传到网络侧可能带来隐私泄露,并且违法隐私保护法规和用户的意愿。
目前,一种可行的方法是终端与网络设备合作来完成AI/ML模型的操作。参考图3所示的终端与网络设备合作进行AI/ML模型操作的网络架构示意图,AI/ML模型可以被分割为多个子模型,终端通过多个子模型中的部分子模型进行数据处理,并将得到的处理结果传输给网络设备,网络设备通过多个子模型中的另一部分子模型进行数据处理;这样,实现终端和网络设备协同对数据进行处理。
但是,现有技术中AI/ML模型的分割方式是固定的,也就是说,待处理数据中所有的数据都通过该固定分割方式进行处理,这就造成在某些情况下,终端的可用处理资源并不能够满足该固定分割方式所要占用的资源大小;因此,通过固定分割方式进行的数据处理会产生资源占用不均衡的问题。
为解决上述问题,本申请实施例提供一种数据处理方法,图4为本申请实施例适用的一种通信***的网络架构示意图,下面,对本申请实施例涉及的无线通信***进行简单介绍。
如图4所示,该通信***的网络架构中包括终端到网络的通信,以及终端到终端(Device-to-Device,D2D)通信;该通信***包括网络设备410,终端420,终端430。其中,终端到网络的通信包括终端420与网络设备410之间的通信,或者终端430与网络设备410之间的通信,D2D通信包括终端420和终端430之间的通信。
需要明确的是,本申请实施例中提到的网络设备410可以是全球移动通讯(Global System of Mobile communication,GSM)***或码分多址(Code Division Multiple Access,CDMA)***的基站(Base Transceiver Station,BTS),也可以是宽带码分多址(Wideband Code Division Multiple Access,WCDMA)***中的基站(NodeB,NB),还可以是长期演进(Long Term Evolution,LTE)***中的演进型基站(evolved NodeB,eNB)、接入点(access point,AP)或者中继站,也可以是5G***中的基站(如gNB或传输点(Transmission Point,TRP))等,还可以是云无线接入网络(Cloud Radio Access Network,CRAN)场景下的无线控制器以及可穿戴设备或车载设备等。在此不作限定。
本申请实施例中提到的终端420以及终端430可以是用户设备(User Equipment,UE)、接入终端、UE单元、UE站、移动站、移动台、远方站、远程终端、移动设备、UE终端、终端、无线终端设备、UE代理或UE装置等。还可以是蜂窝电话、无绳电话、会话启动协议(Session Initiation Protocol,SIP)电话、无线本地环路(Wireless Local Loop,WLL)站、个人数字处理(Personal Digital Assistant,PDA)、具有无线通信功能的手持设备、计算设备或连接到无线调制解调器的其它处理设备、车载设备、无人机、可穿戴设备、机器人,未来5G网络中的终端或者未来演进的公共陆地移动网络(Public Land Mobile Network,PLMN)中的终端等。需要说明的是,在本发明实施例中并不限定终端420和终端430的具体类型。
可以理解的是,本申请实施例中网络/***中具有通信功能的设备可统一称为通信设备。以图4 示出的通信***的网络架构为例,通信设备可包括具有通信功能的网络设备410,终端420和终端430。
基于图4所示的通信***的网络架构示意图,本申请实施例提供一种数据处理方法,该数据处理方法的执行主体为第一通信设备。这里,第一通信设备可以是图4所示的网络设备410,终端420以及终端430中的任意一个。在一可行的示例中,当第一通信设备为网络设备410时,第二通信设备为终端420或者终端430。在另一可行的示例中,当第一通信设备为终端420时,第二通信设备可以为网络设备410或者终端430。在又一可行的示例中,当第一通信设备为终端430时,第二通信设备可以为网络设备410或者终端420。
请参见图5所示的一种数据处理方法流程示意图一,该数据处理方法包括步骤510。其中,
步骤510、获取第一配置信息;
第一配置信息用于配置N种分割方式,以及N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数。
这里,N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
本申请实施例所提及的数据处理模型可以是任意类型的AI/ML模型,例如图像识别的AI/ML模型,自然语言识别的AI/ML模型等。本申请实施例这里不做限定。
需要说明的是,数据处理模型的可选的预设分割位置可以有多个,也就是说,数据处理模型的预设分割方式有多种。在本申请实施例中,第一配置信息中的N种分割方式可以是多个预设分割方式中一种分割方式或者多种分割方式,本申请实施例这里不做限定。
在本申请提供的实施例中,第一通信设备还可以通过第一配置信息确定数据被分为M个部分。这里,数据是指需要通过数据处理模型进行处理的数据(数据也可以称为待处理数据),在本申请提供的实施例中,数据可以是任意类型的数据,例如视频数据,文本数据,音频数据等等,本申请实施例这里不做限定。
可以理解的是,本申请实施例所提及的部分是指数据中的一部分数据,并且,本申请实施例中M个部分组合起来即为数据的整体。
需要说明的是,数据中的每个部分可以是由离散的数据单元构成,也就是说,数据中的每个部分中的数据单元并不是连续的数据单元。例如,针对视频数据来说,可以将视频数据分为两个部分,视频数据中的奇数帧构成第一部分,视频数据中的偶数帧构成第二部分。
其中,上文所述的数据单元是指对数据进行处理的最小单位。在一可行的示例中,数据为视频数据时,数据单元具体指图像帧,也就是说可以将数据分割为一个一个的图像帧进行处理。在另一可行的示例中,数据为文本数据时,可以将数据分割为固定大小(10K)的一段一段数据进行处理,这里固定大小的一段数据即为一个数据单元。
进一步地,第一配置信息可以配置N种分割方式中的第i种分割方式对应的第j个部分,也就是说,第一配置信息还可以为N种分割方式中的每种分割方式配置对应的部分。
可以理解的是,数据中不同的部分对应不同的分割方式。在本申请提供的实施例中,这里的对应关系是指第一通信设备采用第i种分割方式进行处理时,所处理的是整个数据中的第j部分的数据。也就是说,第一通信设备可以通过第i种分割方式对应的子处理模型,处理数据中第j部分的数据。
需要说明的是,当第一配置信息配置两个或者两个以上的分割方式时(即N大于1),M可以与N相同,也就是说,当第一配置信息配置两个或两个以上的分割方式时,每个分割方式对应数据一个部分,分割方式与数据的部分一一对应。当第一配置信息配置一种分割方式时(即N为1),此时M取大于N的值(即M大于1);也就是说,即使第一配置信息配置一种分割方式,但是仍然对数据分为多个(M个)部分,并且第一配置信息中配置的这一种分割方式对应的第j个部分;也就是说,第一通信设备在数据的M个部分中的某一个部分使用该一种分割方式。
实际应用中,采用每种分割方式进行处理时需占用的第一通信设备的处理资源不同。第一通信设备的可用处理资源并不能够与每种分割方式需占用的处理资源完全匹配,大多数情况下,可用处理资源量可能介于不同分割方式需占用的处理资源量之间。
例如,在图7中,时刻1第一通信设备的可用处理资源大小为第一可用处理资源量1,其中,第一可用处理资源量1大于为分割位置2(即第二种分割方式)所占用第一处理资源量2,小于分割位置3(即第三种分割方式)所占用的第一处理资源量3。因此,使用分割位置2进行数据处理,会造成第一通信设备的可用资源的浪费,使用分割位置3进行数据处理,需要的处理资源量又超过了 第一可用处理资源量1。
基于此,在本申请提供的实施例中,第一通信设备可以通过第一配置信息为数据处理模型配置N种分割方式。在数据中不同的部分使用不同的分割方式进行处理,使得第一通信设备所占用的处理资源的平均量与当前第一通信设备可用的处理资源匹配。
在一种可能的实施方式中,N种分割方式仅包括一种分割方式,第一配置信息可以配置该一种分割方式对应的第j个部分。也就是说,第一通信设备采用该种分割方式对第j个部分的数据进行处理。除此之外,第一通信设备在数据中其他的数据部分上不采用任何分割方式进行处理(即数据中其他的部分发送至第二通信设备进行处理),或者在数据中的其他部分采用特定分割方式进行处理。这里,特定分割方式是第一通信设备和第二通信设备约定好的分割方式,特定分割方式是数据处理模型预设的分割方式中的任何一个。在本申请提供的实施方式中,特定分割方式可以是出厂配置好的分割方式,也可以是通过第二通信设备半静态配置的分割方式。
在一示例中,针对视频处理任务,配置信息仅配置第一分割方式以及第一分割方式对应的部分为数据的奇数帧部分;这样,第一通信设备可以在数据的奇数帧使用上述第一分割方式进行处理,在数据的偶数帧不采用任何分割方式处理。
在另一示例中,针对视频处理任务,配置信息仅配置第一分割方式以及第一分割方式对应数据的奇数帧部分。而第一通信设备和第二通信设备约定的特定分割方式为第二分割方式;这样,第一通信设备可以在数据的奇数帧使用上述第一分割方式进行处理,在数据的偶数帧采用第二分割方式进行处理。
在另一种可能的实施方式中,N种分割方式包括两种或者两种以上分割方式,第一通信设备可以根据N种分割方式中每种分割方对应数据的部分,在数据不同的部分,使用不同的分割方式进行处理。
示例性的,针对视频处理任务,N种分割方式包括两种分割方式:第一分割方式和第二分割方式;其中,第一分割方式的对应的部分为数据的奇数帧部分,第二分割方式对应的部分为数据的偶数帧部分。这样,第一通信设备根据第一配置信息,在数据的奇数帧使用第一分割方式进行处理,在数据的偶数帧采用第二分割方式进行处理,即交替应用第一分割方式和第二分割方式对数据进行处理。
在本申请提供的实施例中,配置信息中配置的至少一种分割方式以及至少一种分割方式的数据信息的原则是,第一通信设备使用该至少一种分割方式所占用的第一通信设备的处理资源与第一通信设备的可用处理资源相匹配。
基于前述实施例,步骤510取第一配置信息之后,还可以包括以下步骤:
步骤520、采用第i种分割方式对应的子处理模型,对第j个部分的数据进行处理。
在一种可能的实施方式中,N=1时,第一配置信息中仅配置一种分割方式,以及该一种分割方式对应的第j个部分。这时,第一通信设备可以采用该分割方式的子处理模型对数据的第j个部分进行处理;并且,第一通信设备对数据中除了第j个部分之外的其他部分,不采用任何分割方式进行处理(即其他的部分发送至第二通信设备进行处理);即交替采用第一配置信息中配置的一种分割方式和不分割方式对数据进行处理。另外,针对上述情况,第一通信设备还可以采用特定分割方式,对除了第j个部分之外的其他部分数据进行处理,即交替采用上述分割方式和特定分割方式对数据进行处理。
在另一种可能的实施方式中,N大于等于2时,第一配置信息中配置两个或者两个以上的分割方式,以及每个分割方式对应的数据的部分,则第一通信设备可以在数据中不同的部分,使用不同的分割方式对应的子处理模型对数据进行处理。可以理解为,在第j个部分使用第i种分割方式对应的子处理模型进行处理,在第j+1个部分使用第i+1种分割方式对应的子处理模型进行处理。即,交替使用N种分割方式中的每一种分割方式对应的子处理模型,对数据中各个部分进行处理。
需要说明的是,在交替使用N种分割方式中的每一种分割方式对数据进行处理过程中,通过对多种分割方式中每种分割方式进行遍历来对数据进行处理。示例性的,第一配置信息为数据处理模型配置了两种分割方式,第一分割方式对应的数据的部分为奇数帧部分,第二分割方式对应的数据的部分为偶数帧部分;第一通信设备采用第一种分割方式处理第一帧数据,采用第二种分割方式处理第二帧数据,第一通信设备继续遍历这两种分割方式,即通过第一种分割方式处理第三帧数据,通过第二种分割方式处理第四帧数据,如此交替使用上述两种分割方式,直至待处理数据的最后一帧。
这样,通过在数据的不同部分采用不同的分割方式进行数据处理,可以使得第一通信设备采用 N种分割方式进行数据处理时需要占用的第一通信设备的处理资源,与第一通信设备当前的可用处理资源相匹配。
综上所述,本申请实施例提供的数据处理方法中,第一通信设备可以采用第i种分割方式对数据的第j个部分进行处理,即在数据的不同部分采用不同的分割方式进行处理,如此,使得采用该N种分割方式处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
下面,以图像识别任务为例进行说明。
实际应用中,在图像识别任务中通常采用数据处理模型为卷积神经网络(Convolutional Neural Networks,CNN)模型进行图像处理。参考图6所示的一种CNN模型结构示意图,该CNN模型由不同的数据处理层构成。具体地,CNN模型包括输入层、卷积层1至卷积层5、激励层1至激励层5、池化层1和池化层2、归一化层1和归一化层2、全连接层1和全连接层2、以及输出层。并且,不同的数据处理层具有不同的处理时延和输出数据量。例如,图6所示的CNN模型全连接层2对应的处理时延为10ms,输出数据量为0.1MB。
图6所示的CNN模型具有四个可选的分割位置,包括分割位置0至分割位置3。以选择分割位置2为例进行说明,第一通信设备执行分割位置2对应的池化层2之前的输出处理层,并将池化层2得到的输出数据传输给第二通信设备,这样,第二通信设备可以根据第一通信设备发送的输出数据,从分割位置2之后的归一化层2继续执行剩下的数据处理层,直至完成整个CNN模型的运算,得出图像识别结果。
另外,不同分割位置所占用的处理资源大小也是不同的。如图7所示,分割位置0是一个特殊的分割位置,第一通信设备不执行任何运算,直接将采集到的原始数据(如拍摄的图片)发送至第二通信设备,由第二通信设备执行CNN的所有层的计算。所以,分割位置0并不占用第一通信设备的处理资源,仅占用第二通信设备的处理资源。对于分割位置1,第一通信设备只执行一部分数据处理层的计算,分割位置1需占用第一通信设备的处理资源大小为第一处理资源量1,占用第二通信设备的处理资源大小为第二处理资源量1。同样的,通过分割位置2进行处理时,需占用第一通信设备的处理资源大小为第一处理资源量2,占用第二通信设备的处理资源大小为第二处理资源量2;另外,通过分割位置3进行处理时,需占用第一通信设备的处理资源大小为第一处理资源量3,占用第二通信设备的处理资源大小为第二处理资源量3。其中,基于分割位置2进行数据处理时,占用的第一处理资源量2大于分割位置1所需的第一处理资源量1;分割位置3占用的第一处理资源量3大于分割位置2占用的第一处理资源量2。
实际应用中,某个时刻,第一通信设备的可用处理资源量可能介于上述四个分割位置所占用的第一通信设备的处理资源量之间。如图7所示,在时刻1,第一通信设备可用处理资源大小为第一可用处理资源量1,第二通信设备可用处理资源大小为第二可用处理资源量1;在时刻2,第一通信设备可用处理资源大小为第一可用处理资源量2,第二通信设备可用处理资源大小为第二可用处理资源量2。从图7可以看出,时刻1的第一可用处理资源量1和时刻2的第一可用处理资源量2均高于分割位置2需占用的第一处理资源量2,同时又低于分割位置3需占用的第一处理资源量3。即第一通信设备的可用处理资源不足以支持分割位置3的要求,而采用分割位置2,第一通信设备的可用处理资源又得不到充分利用,造成第一通信设备可用处理资源的浪费。
本申请实施例提供的数据处理方法,在数据的某些部分应用至少一种分割方式,也就是说,在数据处理过程中,第一通信设备在不同分割位置之间切换,例如,第一通信设备在部分帧(例如偶数帧)应用分割位置2处理,在另一部分帧(例如奇数帧)应用分割位置3进行处理,使得第一通信设备所占用的处理资源的平均量与第一通信设备可用处理资源匹配,如此,能够充分利用第一通信设备的处理资源,提高数据处理效率。
在一种可能的实施方式中,第j个部分通过以下信息中的至少一个来确定:
分割方式的时域周期,,分割方式的时域偏移量,以及数据信息的类型;
其中,时域周期用于指示第j个部分数据的时域周期;时域偏移量用于指示第j个部分数据在所述时域周期内的时域位置;数据信息的类型信息用于指示第j个部分是关键数据单元或非关键数据单元。
实际应用中,时域周期可以理解为,重复出现同一事物的时间或空间间隔。在本申请中,时域周期具体指第j个部分数据的时域周期。简单来说,就是重复应用一次N种分割方式中的每一种分割方式,第一通信设备处理的所处理的数据部分的时域周期。
示例性的,参考表1所示的一种示例性配置信息。其中,配置信息包括两种分割位置(对应两 种分割方式):分割位置2和分割3;其中,应用分割方式的时域周期为2帧,即第一通信设备应用一次分割位置2和分割位置3所处理的数据为2帧。
另外,在本申请提供的实施例中,不同分割方式对应不同的时域偏移量。分割方式的时域偏移量是指应用第i种分割方式对第j个部分进行处理时,数据的第j部分在时域周期内的时域位置。
示例性的,参考表1所示的一种示例性配置信息。其中,分割位置包括分割位置2和分割3,分割位置2的时域偏移量为1,分割位置3的时域偏移量为0。也就是说,第一通信设备可以在一个时域周期的第1个图像帧采用分割位置2进行数据处理,在一个时域周期的第0个图像帧采用分割位置3进行数据处理。
表1
时域周期 | 时域偏移量 | |
分割位置2 | 2帧 | 1 |
分割位置3 | 2帧 | 0 |
在另一种实施方式中,还可以通过数据信息的类型确定每种分割方式对应的数据的部分。其中,数据信息的类型用于指示第j个部分是关键数据单元还是非关键数据单元,也就是说,数据信息的类型用于指示数据中每个部分是关键数据单元还是非关键数据单元。
关键数据单元和非关键数据单元对应不同的分割方式。一般情况下,关键数据单元对于用户十分重要,具有隐私敏感性,因此不适宜直接通过无线网络进行传输,也不适宜通过其他的通信设备进行处理。因此,可以将关键数据单元全部放在第一通信设备进行处理,将非关键数据单元发送至第二通信设备进行处理。
这样,关键数据单元可以使用预设的关键分割方式,非关键数据单元使用预设的非关键分割方式进行处理。如此,可以在数据中的关键数据单元使用关键分割方式对应的子处理模型对数据进行处理,在数据中的非关键数据单元使用非关键分割方式对应的子处理模型进行数据处理。
基于上述实施例,在一种可能的实现方式中,步骤510获取第一配置信息可以通过以下步骤实现:
步骤5101、第一通信设备接收至少一个配置信息;
步骤5102、第一通信设备从至少一个配置信息中,确定第一配置信息。
在本申请提供的实施方式中,第一配置信息可以是第二通信设备配置给第一通信设备的。具体地,第二通信设备可以给第一通信设备发送至少一个配置信息,并从至少一个配置信息中确定第一配置信息。
在另一种可能的实现方式中,第一通信设备还可以根据第二通信设备的指示,来从至少一个配置信息中,确定第一配置信息。
基于上述实施例,步骤5101接收至少一个配置信息之后,本申请实施例提供的数据处理方法还包括:
步骤5101a、接收调度信令;
这里,调度信令用于指示多个配置信息中的第二配置信息。
步骤5101b、基于调度信息,从至少一个配置信息中确定第二配置信息。
这里,第二配置信息与第一配置信息类似,都是用于配置N种分割方式,以及N种分割方式中第i种分割方式对应的第j个部分。
在本申请提供的实施例中,第二配置信息可以与第一配置信息相同,也可以与第一配置信息不同。本申请实施例这里不做限定。
在本申请提供的实施方式中,第二通信设备可以根据第一通信设备的可用处理资源的情况,或者第一通信设备与第二通信设备之间的无线信道状态,为第一通信设备确定一个更适用于当前无线信道状态,以及可用处理资源大小的第二配置信息。
这里,第二通信设备可以将第二配置信息携带在调度信息中发送给第一通信设备,第一通信设备在接收到调度信息后,从多个配置信息中选择第二配置信息,并基于第二配置信息进行数据处理。
需要说明的是,第一通信设备可以根据无线信道状态的优劣,来确定无线信道上传输的数据量,当无线信道状态较好时,可以采用在无线信道上传输数据量多配置信息,当无线信道状态较差时,可以采用在无线信道上传输数据量少的配置信息。
示例性的,参考图6所述的CNN模型结构示意图,分割位置1的输出数据量为卷积层1,激励层1以及池化层1输出数据量之和,分割位置3的输出数据量为激励层4左侧所有数据处理层的输出数据量之和。因此,分割位置1的输出数据量小于分割位置3的输出数据量。若第一通信设备使 用分割位置3进行数据处理,当第二通信设备检测到当前无线信道状态较差,可将当前分割方式进行重配置,将当前配置信息中分割位置3修改为分割位置2,如此,降低在无线信道中的传输数据量。
在本申请提供的实施方式中,特定分割方式可以是出厂配置好的分割方式,也可以是通过第二通信设备半静态配置的分割方式。
这样,通过仅配置一种分割方式的相关信息,可以降低配置信息的数据传输量,提高配置信息的传输效率。
在一种可能的实施方式中,第一通信设备还可以在本地基于可用处理资源信息,确定第一配置信息。其中,具体的确定方式与第二通信设备基于可用处理资源信息确定第一配置信息的方式相同,详见下文中的描述。
在一种可能的实施方式中,步骤510获取第一配置信息之前,本申请提供的数据处理方法还包括:
步骤501、向第二通信设备发送可用处理资源信息。
其中,可用处理资源信息包括以下至少之一:第一通信设备的可用计算资源量,可用存储资源量以及可用电量。
在本申请提供的实施例中,第一通信设备可以向第二通信设备发送第一通信设备的可用处理资源信息;这样,第二通信设备在接收到可用处理资源信息之后,第二通信设备为第一通信设备匹配出适用于第一通信设备当前可用处理资源的分割方式以及每种分割方式对应的数据信息。
在一个可能的实施方式中,步骤501向第二通信设备发送可用处理资源信息之后,还包括以下步骤:
步骤502、若可用处理资源信息发生变化,获取更新后的可用处理资源信息;
步骤503、将更新后的可用处理资源信息发送至第二通信设备。
在本申请提供的实施例中,当第一通信设备的可用资处理资源发生变化时,第一通信设备需要从一种配置信息切换到另一种配置信息。基于此,第一通信设备需要将更新好的可用处理资源信息发送给第二通信设备,这样,第二通信设备在接收到更新后的可用处理资源信息之后,可以为第一通信设备重新匹配出适用于第一通信设备当前可用处理资源的分割方式,以及每种分割方式对应的数据信息。
在一种可能的实施方式中,第一通信设备为终端设备,第二通信设备为网络设备。
其中,第二通信设备为网络设备时,第一通信设备可以在以下信息中获取第一配置信息:
控制信道信息、媒体接入控制的控制单元(Media Access Control Control Element,MAC CE)信息、高层配置信令、以及应用层控制信息。
其中,控制信道信息包括下行控制信息(Downlink Control Information,DCI),本申请实施例可以通过DCI一种专用的DCI格式发送所述至少一个配置信息,又或者采用专用无线网络临时标识符(RNTI Radio Network Tempory Identity,RNTI)的生成,高层配置信令可以是无线资源控制(Radio Resource Control,RRC)信令。
本申请实施例提出的数据处理方法,第一通信设备可以采用第i种分割方式对数据的第j个部分进行处理,即在数据的不同部分采用不同的分割方式进行处理,如此,使得采用该N种分割方式处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
请参见图8所示的一种数据处理方法流程示意图,该数据处理方法包括步骤810至步骤820。
步骤810、确定至少一个配置信息。
其中,至少一个配置信息中的每个配置信息用于配置处理待处理数据时的N种分割方式,数据被分成的N个部分,及第i种分割方式对应的第j个部分;N为大于或等于1的整数,i大于等于1且小于等于N,j为大于等于1且小于等于N;
N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
步骤820、向第一通信设备发送至少一个配置信息,至少一个配置信息中包含第一配置信息。
在一种可能的实施方式中,第二通信设备为网络设备的情况下,将至少一个配置信息承载在以下至少一种信息中发送至所述第一通信设备:
控制信道信息、媒体接入控制的控制单元(Media Access Control Control Element,MAC CE)信息、高层配置信令、以及应用层控制信息。
其中,控制信道信息包括下行控制信息(Downlink Control Information,DCI),本申请实施例可 以通过DCI一种专用的DCI格式发送所述至少一个配置信息,又或者采用专用无线网络临时标识符(RNTI Radio Network Tempory Identity,RNTI)的生成,高层配置信令可以是无线资源控制(Radio Resource Control,RRC)信令。
在一种可能的实施方式中,第j个部分通过以下信息中的至少一个来确定:
分割方式的时域周期,分割方式的时域偏移量,以及数据信息的类型;
其中,时域周期用于指示第j个部分数据的时域周期;时域偏移量用于指示第j个部分在时域周期内的时域位置;数据信息的类型用于指示第j个部分是关键数据单元或非关键数据单元。
这里,对于分割方式的时域周期,分割方式的时域偏移量以及数据信息的类型的解释见上述实施例中的描述,本申请实施例这里不再赘述。
在一种可行的实施方式中,步骤810确定至少一个配置信息,可以通过步骤8101和步骤8102实现:
步骤8101、接收第一通信设备发送的可用处理资源信息;
步骤8102、基于可用处理资源信息,确定至少一个配置信息。
在本申请提供的实施例中,第二通信设备在接收到可用处理资源信息之后,第二通信设备可以为第一通信设备匹配出适用于第一通信设备当前可用处理资源的N种分割方式以及N种分割方式中每种分割方式对应的第j个部分。
这里,第二通信设备匹配出的N种分割方式所占用的第一通信设备的处理资源与所述第一通信设备的可用处理资源相匹配。
需要说明的是,N种分割方式所占用的第一通信设备的处理资源与可用处理资源相匹配可以是,N种分割方式所占用的第一通信设备的处理资源与可用处理资源相等,也可以是N种分割方式所占用的第一通信设备的处理资源与可用处理资源两者差值的绝对值小于预设阈值。本申请实施例这里不做限定。
本申请实施例提出的数据处理方法,第二通信设备可以根据第一通信设备的可用处理资源,为第一通信设备确定至少一个配置信息,通过配置信息配置在数据处理过程中N种分割方式以及至少一种分割方式的数据信息,这样,仅在数据中的某些数据部分分别应用N种分割方式,不需要固定的在全部的数据上使用同一分割方式,如此,可以通过在待处理数据的某些数据部分应用N种分割方式,使得使用该N种分割方式处理待处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
基于此,在一种可能的实施方式中,步骤8102中,基于可用处理资源信息,确定至少一个配置信息,可以通过以下步骤实现:
步骤8102a、第二通信设备确定预设M种分割方式分别对应的初始参数;M为大于或等于2的整数;每个初始参数均为自然数;
步骤8102b、第二通信设备基于预设M种分割方式分别占用的预设处理资源量,对初始参数进行持续调整,得到预设M种分割方式占用的处理资源平均量;
步骤8102c、第二通信设备当处理资源平均量与可用处理资源匹配时,确定非零值的目标参数;
步骤8102d、第二通信设备从预设M种分割方式中,确定出与目标参数对应的N种分割方式;
步骤8102e、将N种分割方式和N种分割方式对应的目标参数,确定为一个配置信息,进而得到至少一个配置信息。
在本申请提供的实施例中,数据处理模型的预设分割方式可以有多个,这里使用M代替预设分割方式的个数,M为大于或等于2的整数。
在本申请的一些实施例中,第二通信设备获取所有的预设分割方式,为每个预设分割方式确定初始的参数,其中,每个初始参数均为自然数。这里的初始参数可以表征每种分割方式的初始时域偏移量。
本申请提供的实施例中,通过不断调整初始参数的取值,来计算M种分割方式所占用的处理资源平均量。
具体地,第二通信设备对M个分割方式对应的初始参数与每个分割方式所占用的处理资源量乘积求和,并将求和结果与当前M个初始参数取值的总和求商;求商的结果即为M种分割方式所占用的处理资源平均量。
进一步地,每次计算出M种分割方式所占用的处理资源平均量后,需要判断计算出来的处理资源平均量与第一通信设备的可用处理资源是否匹配,若不匹配,则继续调整初始参数的取值;若匹配,则确定当前初始参数中非零值,将非零值的参数确定为目标参数。这里,目标参数可以指示每 种分割方式对应的时域偏移量。
这里,非零值的参数对应的分割方式,即为数据处理模型要应用的分割方式;且非零值的目标参数的大小可以指示应用的分割方式的时域偏移量。
在一种可能的实现方式中,第二通信设备在得到目标参数后,从预设M种分割方式中,确定出与目标参数对应的两种分割方式;若这两种分割方式中包含不分割数据处理模型的方式(即图6所示的分割位置0对应的分割方式),或者这两种分割方式中包含特定分割方式,则第一通信设备可以将不分割数据处理模型的方式或者特定分割方式去除,仅根据另外一个分割方式,以及该分割方式对应的目标参数值生成配置信息。这样,生成的配置信息中仅包含一种分割方式,即N为1。
在另一种可能的实现方式中,第二通信设备在得到目标参数后,从预设M种分割方式中,确定出与目标参数对应的两种以上的分割方式;并将至少两种分割方式和目标参数,确定为一个配置信息。这样,生成的配置信息中包含多种分割方式,即N大于1。
第二通信设备需要给第一通信设备配置多个配置信息时,则重复上述步骤,得到多个配置信息。这里不再赘述。
示例性的,当预设分割方式为三种时,设置每种分割方式的初始参数为a,b,c;其中第一种分割方式所占用的处理资源量为X,第二种分割方式所占用的处理资源量为Y,第三种分割方式所占用的处理资源量为Z;第二通信设备可以调整初始参数a,b,c的取值;通过公式(1)计算不同初始参数a,b,c的取值下,这3种分割方式所占用的处理资源平均量:
进一步,将每次计算出的三种分割方式所占用的处理资源平均量后,需要判断计算出来的处理资源平均量与第一通信设备的可用处理资源量是否匹配,若不匹配,则继续调整初始参数的取值;若匹配,则确定当前初始参数中非零值,将非零值的参数确定为目标参数。
这里判断计算出来的处理资源平均量与第一通信设备的可用处理资源是否匹配,可以通过判断处理资源平均量与第一通信设备的可用处理资源量之间的差值是否小于阈值来判断。
下面,以图像识别任务为例对上述确定配置信息的方式进行说明。
参考图7,第一可用处理资源量1,正好介于第一处理资源量2和第一处理资源量3之间的平均值;这时,可交替采用分割位置2和分割位置3之间交替来实现与第一可用处理资源量1的匹配。
参考图9,可在奇数帧采用分割位置2,在偶数帧采用分割位置3。这样对奇数帧,占用第一通信设备的处理资源量为第一处理资源量2,对偶数帧,占用第一通信设备的处理资源量为第一处理资源量3,两者的平均正好等于第一可用处理资源量1。通过在分割位置2和分割位置3之间交替切换,使得第一通信设备基于上述方式进行数据处理的过程中所占用的处理资源与第一通信设备可用处理资源匹配,如此,能够充分利用第一通信设备的可用处理资源,提高数据处理效率。
进一步,上述交替采用分割点2和分割点3的配置信息示例如表1所示,可配置分割位置(即分割方式)的时域周期为2帧,分割位置2的时域偏移量为1,分割位置3的时域偏移量为3。即,时域偏移量为1的帧采用分割位置2,时域偏移量为3的帧采用分割位置3。这样第一通信设备占用处理资源的平均量正好等于第一可用处理资源量1。
在另一可行的示例中,如果第一通信设备的可用处理资源量为图7中的第一可用处理资源量2,小于第一可用处理资源量1,采用上一示例中的平均交替模式仍超出了第一通信设备的可用处理资源量,第一通信设备无法完成图像识别任务。此时可以采用分割位置2和分割位置3之间的不平均交替来实现针对第一可用处理资源量2的处理资源量的匹配。
如图10所示,可以在每3帧中的两帧采用分割位置2,在每3帧中的一帧采用分割位置3;这样时域偏移量为1和2的帧,占用第一通信设备的处理资源量为第一处理资源量2,而在域偏移量为3的帧,占用第一通信设备的处理资源量为第一处理资源量3,最终的平均值正好等于第一可用处理资源量2。这样,第一通信设备交替采用分割位置2和分割位置3进行资源处理,使得第一通信设备基于上述方式进行数据处理的过程中所占用的处理资源与第一通信设备可用处理资源匹配,如此,能够充分利用第一通信设备的计算能力,提高数据处理效率。
上述交替采用分割位置2和分割位置3的配置信息示例如表2所示,可配置分割位置(即分割方式)的时域周期为3帧,分割位置2的时域偏移量为1或2,分割位置3的时域偏移量为3。即,时域偏移量为1或2的帧采用分割位置2,时域偏移量为3的帧采用分割位置3。这样第一通信设备占用处理资源的平均量正好等于第一可用处理资源量2。
表2
时域周期 | 时域偏移量 | |
分割位置2 | 3帧 | 1,2 |
分割位置3 | 3帧 | 3 |
另一种可行的实施方式中,还可以通过数据信息的类型确定配置信息。
具体地,步骤810确定至少一个配置信息,可以通过步骤8101’和步骤8102’实现:
步骤8101’、获取数据信息的类型;
步骤8102’、基于数据信息的类型,确定数据的M个部分中第j个部分是关键数据单元或非关键数据单元;
步骤8103’、若第j个部分是关键数据单元,确定第j个部分对应关键分割方式;
步骤8103’、若第j个部分是非关键数据单元,确定第j个部分对应非关键分割方式;
步骤8103’、基于数据的M个部分中第j个部分对应的分割方式,确定一个配置信息。
在本申请的一些实施例中,第二通信设备可以根据数据信息的类型,确定数据中每个部分是关键数据单元还是非关键数据单元。
一般情况下,关键数据单元对于用户十分重要,具有隐私敏感性,因此不适宜直接通过无线网络进行传输。可以将关键数据单元全部放在第一通信设备进行处理,将非关键数据单元发送至第二通信设备进行处理。
这样,关键数据单元可以使用预设的关键分割方式,非关键数据单元使用预设的非关键分割方式进行处理。第二通信设备可在数据的关键数据单元使用关键分割方式,在数据的非关键数据单元使用非关键分割方式。
在本申请提供的实施例中,第二通信设备可以对数据的每个部分进行判断,确定每个部分属于非关键数据还是关键数据单元,若第j个部分为关键数据单元,则确定第j个部分使用预设的关键分割方式进行处理。若第j个部分为非关键数据单元,则确定第j个部分使用预设的非关键分割方式进行处理。
在本申请提供的实施例中,通过数据信息的类型确定至少一个配置信息这种方式中,数据的部分可以分为两部分,即M取值为2;第一部分包括关键数据单元,第二部分包括非关键数据单元。另外,在这种情况下,分割方式也包括两种,即N的取值也为2,其中,第一分割方式为上述关键分割方式,第二分割方式为非关键分割方式。
下面,以图像识别任务为例对上述确定配置信息的方式进行说明。
实际应用中,图像可能分为关键帧(即包含完整图像原始信息的图片)和非关键帧(即只包含图像的增量信息或预测信息)。关键帧因包含原始图像信息,具有隐私敏感性,不适宜在网络中传输。非关键帧不包含图像的原始信息,即使在无线信道中传输,也不会泄露隐私信息。因此可以针对关键帧采用一部分在终端侧计算的分割方式(如图6所示的分割位置1),而对非关键帧采用全部在第二通信网络的计算的分割模式(如图6所示的分割位置0)。
具体第,如图11所示,假设每N帧出现一个关键帧,其余为非关键帧,则可以采用关键分割方式处理关键帧,以避免将携带原始图像数据的关键帧在无线信道中传输,而其余非关键帧可采用非关键分割方式,即全部通过第二通信设备处理,从而节省第一通信设备计算资源。这样,可以在保护第一通信设备数据隐私性的同时,也能充分利用第二通信设备的处理能力,节省第一通信设备宝贵的处理资源。
上述交替采用关键分割方式和非关键分割方式的配置信息示例如表3所示,可配置分割方式的时域周期为N帧,采用关键分割方式的时域偏移量为0,采用非关键分割方式的时域偏移量为非0的帧。
表3
时域周期 | 时域偏移量 | |
关键分割方式 | N帧 | ≠1 |
非关键分割方式 | N帧 | =1 |
在一种可能的实施方式中,第一配置信息可以包括:初始配置信息;所述第一初始配置信息包括第一分割方式及所述第一分割方式对应的分割数据单元长度。
在本申请提供的实施例中,第二通信设备可以向第一通信设备发送初始配置信息,这里的初始配置信息可以仅包括一种分割方式的相关信息,即仅包括第一分割方式和第一分割方式对应的分割 数据单元长度。
在另一可能的实施方式中,步骤830向第一通信设备发送至少一个配置信息之后,还可以包括以下步骤:
步骤840、向第一通信设备发送调度信息;调度信息用于指示多个配置信息中的第二配置信息。
在本申请提供的实施例中,当第一通信设备的可用资处理资源发生变化时,第二通信设备需要指示第一通信设备需要从一种配置信息切换到另一种配置信息。
具体地,第二通信设备可以向第一通信设备发送调度信息,具体指示第一通信设备从多个配置信息中选择第二配置信息进行数据处理。
在本申请提供的实施例中,第二通信设备可以基于与第一通信设备之间的无线信道状态和/或第一通信设备发送的更新后的可用处理资源信息,确定第二配置信息;基于第二配置信息,生成调度信息。
具体地,第二通信设备可以根据第一通信设备的可用处理资源的情况,或者第一通信设备与第二通信设备之间的无线信道状态,为第一通信设备确定一个更适用于当前无线信道状态,以及可用处理资源大小的目标配置信息。
需要说明的是,第一通信设备可以根据无线信道状态的优劣,来确定无线信道上传输的数据量,当无线信道状态较好时,可以采用在无线信道上传输数据量多配置信息,当无线信道状态较差时,可以采用在无线信道上传输数据量少的配置信息。
进一步,第二通信设备在接收到调度信令后,基于调度信令指示的第二配置信息,对至少两种分割方式以及分割数据单元长度进行重配置。并基于第二配置信息进行数据处理。
本申请实施例提出的数据处理方法,第二通信设备可以根据第一通信设备的可用处理资源,为第一通信设备确定至少一个配置信息,通过配置信息配置在数据处理过程中N种分割方式以及N种分割方式的数据信息,这样,仅在数据中的某些部分应用N种分割方式,不需要固定的在全部的数据上使用同一分割方式,如此,可以通过在数据的某些数据部分应用N种分割方式,使得使用该至少一种分割方式处理待处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
基于上述实施例,本申请实施例提供一种数据处理方法,如图12所示,本申请实施例提供的数据处理方法具体包括以下步骤:
步骤1201、第一通信设备向第二通信设备发送可用处理资源信息。
在本申请提供的实施例中,可用处理资源信息包括以下至少之一:第一通信设备的可用计算资源量,可用存储资源量以及可用电量。
步骤1202、第二通信设备基于可用处理资源信息,确定至少一个配置信息。
步骤1203、第二通信设备向第一通信设备发送至少一个配置信息。
在本申请提供的实施例中,至少一个配置信息中包含第一配置信息。
其中,第一配置信息用于配置N种分割方式,以及N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
步骤1204、第一通信设备采用第i种分割方式对应的子处理模型,对第j个部分的数据进行处理。
这里,分割方式可以将数据处理模型分割为第一部分子处理模型和第二部分子处理模型。第一部分子处理模型由第一通信设备操作;第二部分子处理模型由第二通信设备操作。
不同的分割方式,得到的第一部分子处理模型和第二部分子处理模型不同。
在本申请提供的实施例中,第一通信设备按照N种分割方式中每一种分割方式对应的第一部分子处理模型对待处理数据进行处理。
可以理解为,第一通信设备交替使用N种分割方式对应的第一部分子处理模型来对数据进行处理。
步骤1205、第一通信设备获取应用每一种分割方式对应的第一子处理模型进行数据处理时,得到的第一输出数据。
这里第一通信设备按照至少一种分割方式对应的第一部分子处理模型对待处理数据进行处理,得到中间处理结果,即第一输出数据。
步骤1206、第一通信设备将第一输出数据发送至第二通信设备。
这里,第一通信设备可以通过无线链路向第二通信设备发送数据处理模型每次输出的第一输出数据,第一通信设备也可以在得到多个第一输出数据后,将多个输出数据向所述第二通信设备发送,本申请实施例这里不做限定。
步骤1207、第二通信设备基于第一输出数据,以及第i种分割方式对应的子处理模型对第一输出数据进行处理,得到数据处理结果。
在本申请提供的实施例中,第二通信设备接收到第一输出数据后,按照至少一种分割方式对应的第二部分子处理模型对待处理数据进行处理。得到最终的数据处理结果。
可以理解为,第二通信设备基于第一输出数据,交替使用N种分割方式对应的第二部分子处理模型来对数据进行处理。
综上所述,本申请实施例提供的数据处理方法中,第一通信设备可以仅在待处理数据中的某些数据部分应用N种分割方式,不需要固定的在全部的数据上使用同一分割方式;如此,可以通过在数据的某些部分应用N种分割方式,使得使用该N种分割方式处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
基于实施例一同一发明构思,本申请实施例提供了一种通信设备13,对应于一种应用于第一通信设备中的数据处理方法;图13为本申请实施例提供的一种通信设备的组成结构示意图一,该通信设备13可以包括:
第一获取单元1301,配置为获取第一配置信息;所述第一配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;
所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式.
在本申请的一些实施例中,所述第一通信设备还包括第一处理单元;
第一处理单元,配置为采用所述第i种分割方式对应的子处理模型,对所述第j个部分的数据进行处理。
在本申请的一些实施例中,所述第j个部分通过以下信息中的至少一个来确定:
分割方式的时域周期,分割方式的时域偏移量,以及数据信息的类型;
其中,所述时域周期用于指示第j个部分数据的时域周期;所述时域偏移量用于指示所述第j个部分在所述时域周期内的时域位置;所述数据信息的类型用于指示所述第j个部分是关键数据单元或非关键数据单元。
在本申请的一些实施例中,所述通信设备包括第一接收单元和第一确定单元;
所述第一接收单元,配置为接收至少一个配置信息;
所述第一确定单元,配置为从所述至少一个配置信息中,确定所述第一配置信息。
在本申请的一些实施例中,所述第一接收单元,配置为接收所述发送的调度信令;
所述第一处理单元,配置为根据所述调度信息,从至少一个配置信息中确定第二配置信息。
在本申请的一些实施例中,所述第一发送单元1303,配置为向所述第二通信设备发送可用处理资源信息。
在本申请的一些实施例中,所述可用处理资源信息包括以下至少之一:所述第一通信设备的可用计算资源量,可用存储资源量以及可用电量。
在本申请的一些实施例中,所述通信设备还包括发送单元,
所述第一获取单元1301,配置为若所述可用处理资源信息发生变化,获取更新后的可用处理资源信息;
所述第一发送单元,配置为将所述更新后的可用处理资源信息发送至所述第二通信设备。
可以理解的是,申请实施例提出的通信设备,可以仅在数据中的某些部分应用N种分割方式,不需要固定的在全部的数据上使用同一分割方式;如此,可以通过在数据的某些部分应用N种分割方式,使得使用该N种分割方式处理待处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
基于前述实施例,本申请实施例提供了一种通信设备14,对应于一种应用于第二通信设备中的数据处理方法;图14为本申请实施例提供的一种通信设备的组成结构示意图二,该通信设备14可以包括:
第二确定单元1401,配置为确定至少一个配置信息;其中,所述至少一个配置信息中的每个配 置信息用于配置N种分割方式,数据被分成的N个部分,及第i种分割方式对应的第j个部分;N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数。
第二发送单元1402,配置为向所述第一通信设备发送所述至少一个配置信息;所述至少一个配置信息中包含第一配置信息。
在本申请的一些实施例中,所述第j个部分通过以下信息中的至少一个来确定:
分割方式的时域周期,分割方式的时域偏移量,以及所述数据信息的类型;
其中,所述时域周期用于指示所述第j个部分数据的时域周期;所述时域偏移量用于指示所述第j个部分在所述时域周期内的时域位置;所述数据信息的类型用于指示所述第j个部分是关键数据单元或非关键数据单元。
在本申请的一些实施例中,所述通信设备还包括第二接收单元;
第二接收单元,配置为接收所述第一通信设备发送的可用处理资源信息;
所述第二确定单元1401,配置为基于所述可用处理资源信息,确定所述至少一个配置信息。
在本申请的一些实施例中,所述第二确定单元,配置为确定预设M种分割方式分别对应的初始参数;M为大于或等于2的整数;所述每个初始参数均为自然数;基于预设M种分割方式分别占用的预设处理资源量,对所述初始参数进行持续调整,得到所述预设M种分割方式占用的处理资源平均量;当所述处理资源平均量与所述可用处理资源匹配时,确定非零值的目标参数;从所述预设M种分割方式中,确定出与所述目标参数对应的N种分割方式;将所述N种分割方式和所述N种分割方式分别对应的目标参数,确定为一个配置信息,进而得到所述至少一个配置信息。
在本申请的一些实施例中,所述通信设备还包括第二获取单元;
所述第二获取单元,配置为获取数据信息的类型;
所述第二确定单元,配置为基于所述数据信息的类型,确定数据的M个部分中第j个部分是关键数据单元或非关键数据单元;若所述第j个部分是关键数据单元,确定所述第j个部分对应关键分割方式;若所述第j个部分是非关键数据单元,确定所述第j个部分对应非关键分割方式;基于数据的M个部分中第j个部分对应的分割方式,确定一个配置信息。
在本申请的一些实施例中,第二发送单元1402,配置为向所述第一通信设备发送调度信息;所述调度信息用于指示所述多个配置信息中的第二配置信息。
在本申请的一些实施例中,所述通信设备还包括生成单元;
所述第二确定单元,配置为基于与所述第第一通信设备之间的无线信道状态和/或所述第一通信设备发送的更新后的可用处理资源信息,确定第二配置信息;
所述生成单元,配置为基于所述第二配置信息,生成所述调度信息。
在本申请的一些实施例中,所述第二通信设备为网络设备的情况下,将所述至少一个配置信息承载在以下至少一种信息中发送至所述第一通信设备:
控制信道信息、媒体接入控制的控制单元MAC CE信息、高层配置信令、以及应用层控制信息。
本申请实施例提出的通信设备,可以根据第一通信设备的可用处理资源,为第一通信设备确定至少一个配置信息,通过配置信息配置在数据处理过程中N种分割方式以及N种分割方式对应数据的部分,这样,仅在数据中的某些部分应用N种分割方式,不需要固定的在全部的数据上使用同一分割方式,如此,可以通过在待处理数据的某些数据部分应用N种分割方式,使得使用该N种分割方式处理待处理数据时所占用的处理资源,与第一通信设备可用处理资源匹配;提高了数据处理过程中资源占用的均衡性以及数据处理过程中的资源利用率。
需要说明的是:上述各单元的划分仅为示例性的,实际应用中,可以将终端的内部结构划分成不同的单元,以完成以上描述的全部或者部分功能。另外,上述实施例提供的数据处理方法实施例属于同一构思,其具体实现过程详见方法实施例,这里不再赘述。
基于上述通信设备的硬件实现,本申请实施例还提供了一种通信设备,对应于一种应用于第一通信设备中的数据处理方法;图15为本申请实施例的终端的硬件组成结构示意图,如图15所示,通信设备包括第一收发器1501,第一处理器1502,以及存储有计算机程序的第一存储器1503。
其中,所述第一处理器1502配置为运行所述计算机程序时,执行前述图5和图12对应的实施例中的方法步骤。
当然,实际应用时,如图15所示,该通信设备中的各个组件通过总线***1504耦合在一起。可理解,总线***1504用于实现这些组件之间的连接通信。总线***1504除包括数据总线之外,还包括电源总线、控制总线和状态信号总线。但是为了清楚说明起见,在图15将各种总线都标为总 线***1504。
基于上述通信设备的硬件实现,本申请实施例还提供了一种通信设备,对应于一种应用于第二通信设备中的数据处理方法图16为本申请实施例的终端的硬件组成结构示意图,如图16所示,通信设备包括第二收发器1601,第二处理器1602,以及存储有计算机程序的第二存储器1603。
其中,所述第二处理器1602配置为运行所述计算机程序时,执行前述图8和图12对应的实施例中的方法步骤。
当然,实际应用时,如图16所示,该通信设备中的各个组件通过总线***1604耦合在一起。可理解,总线***1604用于实现这些组件之间的连接通信。总线***1604除包括数据总线之外,还包括电源总线、控制总线和状态信号总线。但是为了清楚说明起见,在图16将各种总线都标为总线***1604。
可以理解,本实施例中的存储器可以是易失性存储器或非易失性存储器,也可包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(Read Only Memory,ROM)、可编程只读存储器(Programmable Read-Only Memory,PROM)、可擦除可编程只读存储器(Erasable Programmable Read-Only Memory,EPROM)、电可擦除可编程只读存储器(Electrically Erasable Programmable Read-Only Memory,EEPROM)、磁性随机存取存储器(Ferromagnetic Random Access Memory,FRAM)、快闪存储器(Flash Memory)、磁表面存储器、光盘、或只读光盘(Compact Disc Read-Only Memory,CD-ROM);磁表面存储器可以是磁盘存储器或磁带存储器。易失性存储器可以是随机存取存储器(Random Access Memory,RAM),其用作外部高速缓存。通过示例性但不是限制性说明,许多形式的RAM可用,例如静态随机存取存储器(Static Random Access Memory,SRAM)、同步静态随机存取存储器(Synchronous Static Random Access Memory,SSRAM)、动态随机存取存储器(Dynamic Random Access Memory,DRAM)、同步动态随机存取存储器(Synchronous Dynamic Random Access Memory,SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate Synchronous Dynamic Random Access Memory,DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced Synchronous Dynamic Random Access Memory,ESDRAM)、同步连接动态随机存取存储器(SyncLink Dynamic Random Access Memory,SLDRAM)、直接内存总线随机存取存储器(Direct Rambus Random Access Memory,DRRAM)。本申请实施例描述的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
上述本申请实施例揭示的方法可以应用于处理器中,或者由处理器实现。处理器可能是一种集成电路芯片,具有信号的处理能力。在实现过程中,上述方法的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以是通用处理器、DSP,或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。处理器可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤,可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于存储介质中,该存储介质位于存储器,处理器读取存储器中的信息,结合其硬件完成前述方法的步骤。
本申请实施例还提供了一种计算机存储介质,具体为计算机可读存储介质。其上存储有计算机指令,作为第一种实施方式,在计算机存储介质位于终端时,该计算机指令被处理器执行时实现本申请实施例上述数据处理方法中的任意步骤。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备和方法,可以通过其它的方式实现。以上所描述的设备实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,如:多个单元或组件可以结合,或可以集成到另一个***,或一些特征可以忽略,或不执行。另外,所显示或讨论的各组成部分相互之间的耦合、或直接耦合、或通信连接可以是通过一些接口,设备或单元的间接耦合或通信连接,可以是电性的、机械的或其它形式的。
上述作为分离部件说明的单元可以是、或也可以不是物理上分开的,作为单元显示的部件可以是、或也可以不是物理单元,即可以位于一个地方,也可以分布到多个网络单元上;可以根据实际的需要选择其中的部分或全部单元来实现本实施例方案的目的。
另外,在本申请各实施例中的各功能单元可以全部集成在一个处理单元中,也可以是各单元分别单独作为一个单元,也可以至少两个单元集成在一个单元中;上述集成的单元既可以采用硬件的形式实现,也可以采用硬件加软件功能单元的形式实现。
本领域普通技术人员可以理解:实现上述方法实施例的全部或部分步骤可以通过程序指令相关 的硬件来完成,前述的程序可以存储于一计算机可读取存储介质中,该程序在执行时,执行包括上述方法实施例的步骤;而前述的存储介质包括:移动存储设备、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
或者,本申请上述集成的单元如果以软件功能模块的形式实现并作为独立的产品销售或使用时,也可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请实施例的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机、服务器、或者网络设备等)执行本申请各个实施例所述方法的全部或部分。而前述的存储介质包括:移动存储设备、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
需要说明的是:本申请实施例所记载的技术方案之间,在不冲突的情况下,可以任意组合。
以上所述,仅为本申请的具体实施方式,但本申请的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本申请的保护范围之内。因此,本申请的保护范围应以所述权利要求的保护范围为准。
Claims (21)
- 一种数据处理方法,应用于第一通信设备,所述方法包括:获取第一配置信息;所述第一配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
- 根据权利要求1所述的方法,其中,所述获取第一配置信息之后,还包括:采用所述第i种分割方式对应的子处理模型,对所述第j个部分的数据进行处理。
- 根据权利要求1或2所述的方法,其中,所述第j个部分通过以下信息中的至少一个来确定:分割方式的时域周期,分割方式的时域偏移量,以及数据信息的类型;其中,所述时域周期用于指示第j个部分数据的时域周期;所述时域偏移量用于指示所述第j个部分在所述时域周期内的时域位置;所述数据信息的类型用于指示所述第j个部分是关键数据单元或非关键数据单元。
- 根据权利要求1-3任一项所述的方法,其中,所述获取第一配置信息之前,还包括:接收至少一个配置信息;从所述至少一个配置信息中,确定所述第一配置信息。
- 根据权利要求4所述的方法,其中,所述接收至少一个配置信息之后,还包括:接收调度信息;根据所述调度信息,从所述至少一个配置信息中确定第二配置信息。
- 根据权利要求1-5任一项所述的方法,其中,所述获取第一配置信息之前,还包括:向第二通信设备发送可用处理资源信息。
- 根据权利要求6所述的方法,其中,所述可用处理资源信息包括以下至少之一:所述第一通信设备的可用计算资源量,可用存储资源量以及可用电量。
- 根据权利要求6或7所述的方法,其中,所述向第二通信设备发送可用处理资源信息之后,还包括:若所述可用空闲处理资源信息发生变化,则获取更新后的可用处理资源信息;将所述更新后的可用处理资源信息发送至所述第二通信设备。
- 一种数据处理方法,应用于第二通信设备,所述方法包括:确定至少一个配置信息;其中,所述至少一个配置信息中的每个配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式;向第一通信设备发送至少一个配置信息;所述至少一个配置信息中包含第一配置信息。
- 根据权利要求9所述的方法,其中,所述第j个部分通过以下信息中的至少一个来确定:分割方式的时域周期,分割方式的时域偏移量,以及所述数据信息的类型;其中,所述时域周期用于指示所述第j个部分数据的时域周期;所述时域偏移量用于指示所述第j个部分在所述时域周期内的时域位置;所述数据信息的类型用于指示所述第j个部分是关键数据单元或非关键数据单元。
- 根据权利要求9或10所述的方法,其中,所述确定至少一个配置信息,包括:接收所述第一通信设备发送的可用处理资源信息;基于所述可用处理资源信息,确定所述至少一个配置信息。
- 根据权利要求11所述的方法,其中,所述基于所述可用处理资源信息,确定至少一个配置信息,包括:确定预设M种分割方式分别对应的初始参数;M为大于或等于2的整数;所述每个初始参数均为自然数;基于预设M种分割方式分别占用的预设处理资源量,对所述初始参数进行持续调整,得到所述 预设M种分割方式占用的处理资源平均量;当所述处理资源平均量与所述可用处理资源匹配时,确定非零值的目标参数;从所述预设M种分割方式中,确定出与所述目标参数对应的N种分割方式;将所述N种分割方式和所述N种分割方式分别对应的目标参数,确定为一个配置信息,进而得到所述至少一个配置信息。
- 根据权利要求9或10所述的方法,其中,所述确定至少一个配置信息,包括:获取数据信息的类型;基于所述数据信息的类型,确定数据的M个部分中第j个部分是关键数据单元或非关键数据单元;若所述第j个部分是关键数据单元,确定所述第j个部分对应关键分割方式;若所述第j个部分是非关键数据单元,确定所述第j个部分对应非关键分割方式;基于数据的M个部分中第j个部分对应的分割方式,确定一个配置信息。
- 根据权利要求9-13任一项所述的方法,其中,所述向第一通信设备发送所述至少一个配置信息之后,所述方法还包括:向所述第一通信设备发送调度信息;所述调度信息用于指示所述至少一个配置信息中的第二配置信息。
- 根据权利要求14所述的方法,其中,所述向所述第二通信设备发送调度信息之前,所述方法还包括:基于与所述第一通信设备之间的无线信道状态和/或所述第一通信设备发送的更新后的可用处理资源信息,确定第二配置信息;基于所述第二配置信息,生成所述调度信息。
- 根据权利要求9-15任一项所述的方法,其中,所述第二通信设备为网络设备的情况下,将所述至少一个配置信息承载在以下至少一种信息中发送至所述第一通信设备:控制信道信息、媒体接入控制的控制单元MAC CE信息、高层配置信令、以及应用层控制信息。
- 一种通信设备,所述通信设备包括:第一获取单元,配置为获取第一配置信息;所述第一配置信息用于配置N种分割方式,以及所述N种分割方式中第i种分割方式对应的第j个部分;其中,N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数;所述N种分割方式为通过预设分割位置将数据处理模型分割为至少两个子处理模型的分割方式。
- 一种通信设备,所述通信设备包括:第二确定单元,配置为确定至少一个配置信息;其中,所述至少一个配置信息中的每个配置信息用于配置N种分割方式,数据被分成的N个部分,及第i种分割方式对应的第j个部分;N为大于或等于1的整数,i大于等于1且小于等于N,j大于等于1且小于等于M;M为大于1的整数第二发送单元,配置为向所述第一通信设备发送所述至少一个配置信息;所述至少一个配置信息中包含第一配置信息。
- 一种通信设备,所述通信设备包括:第一收发器、第一处理器和存储有计算机程序的第一存储器;所述第一收发器、所述第一处理器和所述第一存储器之间通过通信总线进行通信;所述第一处理器,配置为结合所述第一收发器,运行所述第一存储器中存储的所述计算机程序时,执行权利要求1至8任一项所述方法的步骤。
- 一种通信设备,所述通信设备包括:第二收发器、第二处理器和存储有计算机程序的第二存储器;所述第二收发器、所述第二处理器和所述第二存储器之间通过通信总线进行通信;所述第二处理器,配置为结合所述第二收发器,运行所述第二存储器中存储的所述计算机程序时,执行权利要求9至16任一项所述方法的步骤。
- 一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被第一处理器执行实现权利要求1至8任一项所述方法的步骤;或者;所述计算机程序被第二处理器执行9至16任一项所述方法的步骤。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP20930985.5A EP4138428A4 (en) | 2020-04-17 | 2020-04-17 | DATA PROCESSING METHOD, COMMUNICATION DEVICE AND COMPUTER STORAGE MEDIA |
CN202080099787.5A CN115398942A (zh) | 2020-04-17 | 2020-04-17 | 数据处理方法及通信设备、计算机存储介质 |
PCT/CN2020/085444 WO2021208105A1 (zh) | 2020-04-17 | 2020-04-17 | 数据处理方法及通信设备、计算机存储介质 |
US17/967,020 US20230047386A1 (en) | 2020-04-17 | 2022-10-17 | Method for data processing, and communication device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2020/085444 WO2021208105A1 (zh) | 2020-04-17 | 2020-04-17 | 数据处理方法及通信设备、计算机存储介质 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/967,020 Continuation US20230047386A1 (en) | 2020-04-17 | 2022-10-17 | Method for data processing, and communication device |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021208105A1 true WO2021208105A1 (zh) | 2021-10-21 |
Family
ID=78083952
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/085444 WO2021208105A1 (zh) | 2020-04-17 | 2020-04-17 | 数据处理方法及通信设备、计算机存储介质 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20230047386A1 (zh) |
EP (1) | EP4138428A4 (zh) |
CN (1) | CN115398942A (zh) |
WO (1) | WO2021208105A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023198169A1 (zh) * | 2022-04-15 | 2023-10-19 | 展讯半导体(南京)有限公司 | 一种信息反馈方法及相关装置 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110208841A1 (en) * | 2010-02-22 | 2011-08-25 | Microsoft Corporation | Incrementally managing distributed configuration data |
CN105338297A (zh) * | 2014-08-11 | 2016-02-17 | 杭州海康威视***技术有限公司 | 一种视频数据存储及回放***、装置和方法 |
CN106650493A (zh) * | 2016-12-14 | 2017-05-10 | 咪咕文化科技有限公司 | 一种行为数据的获取方法和装置 |
CN107295360A (zh) * | 2016-04-13 | 2017-10-24 | 成都鼎桥通信技术有限公司 | 视频传输方法及装置 |
-
2020
- 2020-04-17 EP EP20930985.5A patent/EP4138428A4/en active Pending
- 2020-04-17 WO PCT/CN2020/085444 patent/WO2021208105A1/zh unknown
- 2020-04-17 CN CN202080099787.5A patent/CN115398942A/zh active Pending
-
2022
- 2022-10-17 US US17/967,020 patent/US20230047386A1/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110208841A1 (en) * | 2010-02-22 | 2011-08-25 | Microsoft Corporation | Incrementally managing distributed configuration data |
CN105338297A (zh) * | 2014-08-11 | 2016-02-17 | 杭州海康威视***技术有限公司 | 一种视频数据存储及回放***、装置和方法 |
CN107295360A (zh) * | 2016-04-13 | 2017-10-24 | 成都鼎桥通信技术有限公司 | 视频传输方法及装置 |
CN106650493A (zh) * | 2016-12-14 | 2017-05-10 | 咪咕文化科技有限公司 | 一种行为数据的获取方法和装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP4138428A4 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023198169A1 (zh) * | 2022-04-15 | 2023-10-19 | 展讯半导体(南京)有限公司 | 一种信息反馈方法及相关装置 |
Also Published As
Publication number | Publication date |
---|---|
EP4138428A4 (en) | 2023-06-14 |
CN115398942A (zh) | 2022-11-25 |
EP4138428A1 (en) | 2023-02-22 |
US20230047386A1 (en) | 2023-02-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11564271B2 (en) | User equipment category signaling in an LTE-5G configuration | |
WO2021142637A1 (zh) | 人工智能操作处理方法、装置、***、终端及网络设备 | |
WO2021179196A1 (zh) | 一种基于联邦学习的模型训练方法、电子设备及存储介质 | |
CN113570062B (zh) | 机器学习模型参数传递方法及装置 | |
WO2021142627A1 (zh) | 资源调度方法、装置及可读存储介质 | |
JP7150042B2 (ja) | 信号伝送方法、ネットワーク装置及び端末装置 | |
TWI744494B (zh) | 傳輸數據的方法和設備 | |
CN110933712B (zh) | 一种数据传输方法、终端及存储介质 | |
US11490433B2 (en) | Method and system for conflict resolution using a terminal device | |
WO2019028699A1 (zh) | 信号传输方法及相关设备 | |
CN103563462A (zh) | 用于配置分段载波的探测参考信号的方法和装置 | |
WO2022083634A1 (zh) | 资源配置方法、装置、设备及可读存储介质 | |
WO2021135264A1 (zh) | 通信资源的配置方法、装置、终端及可读存储介质 | |
WO2021218302A1 (zh) | 机器学习模型参数传递方法及装置 | |
TW201818675A (zh) | 波束賦形訊息的交互方法和網路設備 | |
AU2017417277A1 (en) | Data transmission method, terminal device and network device | |
WO2021208105A1 (zh) | 数据处理方法及通信设备、计算机存储介质 | |
JP6980171B2 (ja) | 情報の伝送方法および装置 | |
KR20230122668A (ko) | 직류 위치 처리 방법 및 관련 기기 | |
CN113055933A (zh) | 小区接入方法、用户设备和基站 | |
EP4187975A1 (en) | Method and device for determining handover target | |
US20230209442A1 (en) | Communication Method and Device | |
TW201924465A (zh) | 無線通訊方法和設備 | |
EP4113300A1 (en) | Model coordination method and apparatus | |
WO2023123477A1 (zh) | 无线通信方法、终端设备和网络设备 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20930985 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2020930985 Country of ref document: EP Effective date: 20221116 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |