WO2023088268A1 - Ai信息的传输方法和设备 - Google Patents
Ai信息的传输方法和设备 Download PDFInfo
- Publication number
- WO2023088268A1 WO2023088268A1 PCT/CN2022/132085 CN2022132085W WO2023088268A1 WO 2023088268 A1 WO2023088268 A1 WO 2023088268A1 CN 2022132085 W CN2022132085 W CN 2022132085W WO 2023088268 A1 WO2023088268 A1 WO 2023088268A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- information
- model representation
- communication device
- model
- mode
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
Definitions
- the present application belongs to the field of communication technology, and specifically relates to a method and equipment for transmitting artificial intelligence (AI) information.
- the equipment may include communication equipment such as an AI information transmission device, a terminal or a network side device.
- AI model representations such as AI frameworks
- different AI model representations have different emphases and even support different development languages, and the descriptions and functions of AI models are also implemented in different ways, resulting in AI information cannot be transmitted between two or more communication devices using different AI model representations, which affects the performance of the communication system.
- Embodiments of the present application provide an AI information transmission method and device, which can solve the problem that communication system performance is affected because AI information cannot be transmitted between communication devices using different AI representation methods.
- a method for transmitting AI information including: a first communication device generates AI information in a first AI model representation mode; wherein, the AI information support in the first AI model representation mode is converted into AI information in the second AI model representation mode; the first communication device sends the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses the second AI Model representation.
- an AI information transmission device including: a generating module, configured to generate AI information in a first AI model representation mode; wherein, the AI information in the first AI model representation mode supports being converted AI information in the second AI model representation mode; a sending module, configured to send the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses the second AI Model representation.
- a communication device which includes a processor, a memory, and a program or instruction stored in the memory and operable on the processor, and the program or instruction is executed by the processor During execution, the method described in the first aspect is realized.
- a communication device including a processor and a communication interface, wherein the processor is configured to generate AI information in a first AI model representation mode; wherein the AI information in the first AI model representation mode The information support is converted into AI information in the second AI model representation mode, and the communication interface is used to send the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses The second AI model representation manner.
- a readable storage medium where a program or an instruction is stored on the readable storage medium, and when the program or instruction is executed by a processor, the method as described in the first aspect is implemented.
- a sixth aspect provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method as described in the first aspect .
- a computer program/program product is provided, the computer program/program product is stored in a non-transitory storage medium, and the computer program/program product is executed by at least one processor to implement the first method described in the aspect.
- the first communication device generates AI information in the first AI model representation mode and sends it to the second communication device, because the AI information in the first AI model representation mode supports being converted into the second AI model representation
- the second communication device using the second AI model representation mode receives the AI information in the first AI model representation mode, and loads it into the second AI model representation mode to obtain the second AI model AI information in the representation mode.
- FIG. 1 is a schematic diagram of a wireless communication system according to an embodiment of the present application.
- FIG. 2 is a schematic flowchart of a method for transmitting AI information according to an embodiment of the present application
- FIG. 3 is a schematic structural diagram of an AI information transmission device according to an embodiment of the present application.
- FIG. 4 is a schematic structural diagram of a communication device according to an embodiment of the present application.
- FIG. 5 is a schematic structural diagram of a terminal according to an embodiment of the present application.
- Fig. 6 is a schematic structural diagram of a network side device according to an embodiment of the present application.
- first, second and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific sequence or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application are capable of operation in sequences other than those illustrated or described herein and that "first" and “second” distinguish objects. It is usually one category, and the number of objects is not limited. For example, there may be one or more first objects.
- “and/or” in the description and claims means at least one of the connected objects, and the character “/” generally means that the related objects are an "or” relationship.
- LTE Long Term Evolution
- LTE-Advanced LTE-Advanced
- LTE-A Long Term Evolution-Advanced
- CDMA Code Division Multiple Access
- TDMA Time Division Multiple Access
- FDMA Frequency Division Multiple Access
- OFDMA Orthogonal Frequency Division Multiple Access
- SC-FDMA Single carrier-Frequency Division Multiple Access
- system and “network” in the embodiments of the present application are often used interchangeably, and the described technology can be used for the above-mentioned system and radio technology, and can also be used for other systems and radio technologies.
- the following description describes the New Radio (New Radio, NR) system for illustrative purposes, and uses NR terminology in most of the following descriptions. These technologies can also be applied to applications other than NR system applications, such as the 6th Generation (6 th Generation , 6G) communication system.
- 6th Generation 6th Generation
- Fig. 1 shows a schematic diagram of a wireless communication system to which this embodiment of the present application is applicable.
- the wireless communication system includes a terminal 11 and a network side device 12 .
- the terminal 11 can also be called a terminal device or a user terminal (User Equipment, UE), and the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer) or a notebook computer, a personal digital Assistant (Personal Digital Assistant, PDA), handheld computer, netbook, ultra-mobile personal computer (ultra-mobile personal computer, UMPC), mobile internet device (Mobile Internet Device, MID), augmented reality (augmented reality, AR)/virtual reality (virtual reality, VR) equipment, robots, wearable devices (Wearable Device), vehicle-mounted equipment (VUE), pedestrian terminal (PUE), smart home (home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture etc.) and other terminal-side devices, wearable devices include: smart watches, smart bracelets, smart
- the network side device 12 may be a base station or a core network, where a base station may be called a node B, an evolved node B, an access point, a base transceiver station (Base Transceiver Station, BTS), a radio base station, a radio transceiver, a basic service Basic Service Set (BSS), Extended Service Set (ESS), Node B, Evolved Node B (eNB), Next Generation Node B (gNB), Home Node B, Home Evolved Node B, wireless A Wireless Local Area Network (WLAN) access point, a Wireless Fidelity (WiFi) node, a Transmitting Receiving Point (TRP), or some other appropriate term in the field in question, as long as the same Technical effect, the base station is not limited to a specific technical vocabulary, it should be noted that in the embodiment of the present application, only the base station in the NR system is taken as an example, but
- the embodiment of the present application provides a method 200 for transmitting AI information, which can be performed by the first communication device, in other words, the method can be performed by software or hardware installed on the first communication device, the
- the first communication device may be a network side device or a terminal, and the method includes the following steps.
- the first communication device generates AI information in a first AI model representation mode; wherein, the AI information in the first AI model representation mode supports being converted into AI information in a second AI model representation mode.
- the first communication device sends the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses the second AI model representation mode.
- the AI model representations mentioned in the various embodiments of this application can be specifically AI frameworks or independent AI representations
- the template or data structure of the model (such as Open Neural Network Exchange (ONNX), etc.), the above template or data structure is only responsible for representing the AI model, not responsible for training and inference.
- the aforementioned AI frameworks include, for example, TensorFlow, PyTorch, Keras, MXNet, Caffe2, etc. Each AI framework will use its own method to describe the AI model, and complete the building, training, inference and other operations of the AI model.
- the AI information mentioned in various embodiments of the present application may include AI models (or called AI networks), parameters, structures, etc., such AI models include neural network models, decision tree models, support vector machine models, and Bayesian classifiers model etc. Subsequent embodiments mostly use the AI model as an example to introduce this embodiment. It can be understood that the AI information is not limited to the AI model.
- the first communication device uses a third AI model representation
- the first communication device generating AI information in the first AI model representation includes: the first communication device uses the first AI model representation
- the AI information in the representation mode of the three AI models is converted or saved as the AI information in the representation mode of the first AI model.
- the "conversion" mentioned in this example, for example, the representation of the third AI model is TensorFlow
- the representation of the first AI model is PyTorch
- the first communication device saves the AI model under TensorFlow into information in TensorFlow format, and then converts it into Information in PyTorch format.
- the "save" mentioned in this example, for example, the third AI model is represented by TensorFlow
- the first AI model is represented by PyTorch
- the first communication device saves the AI model under TensorFlow as information in PyTorch format.
- the first AI model representation can be referred to as an intermediate AI model representation (such as an intermediate AI framework), that is, the first communication device and the second communication device select an intermediate AI model representation to transmit AI information , the intermediate AI model representation support is converted to AI information in the second AI model representation mode.
- the intermediate AI model representation manner may be Open Neural Network Exchange (ONNX).
- the first communication device uses the first AI model representation.
- the first communication device may generate AI information (such as an AI model) by using the first AI model representation manner (such as an AI framework).
- the first communication device may use the AI framework to train the AI model to obtain the trained AI model.
- the second communication device may know the type of the representation of the first AI model in a pre-agreed manner or according to an instruction of the first communication device.
- the second communication device may know that the representation of the first AI model is PyTorch, the AI information in the first AI model representation mode is the AI model and parameters under PyTorch.
- the second communication device can load the received AI information into the second AI model representation mode, combined with the above-mentioned type of the first AI model representation mode, the AI information in the second AI model representation mode can be obtained, for example , convert the MXNet module through PyTorch developed under MXNet to obtain the AI model under MXNet.
- the first communication device generates AI information in the first AI model representation mode and sends it to the second communication device, since the AI information support in the first AI model representation mode is converted into AI information in the second AI model representation mode, so that the second communication device using the second AI model representation mode receives the AI information in the first AI model representation mode and loads it into the second AI model representation mode,
- the AI information in the representation mode of the second AI model is obtained.
- Embodiment 200 mainly introduces the transmission process of AI information.
- the AI model representation mode for transmitting AI information can also be pre-configured; during or after the execution of embodiment 200, you can also Modify the AI model representation used to transfer AI information.
- the method provided in embodiment 200 may further include the following step: the first communication device sends configuration information, and the second communication device may also receive the configuration information, where the configuration information is used to configure or modify the configuration information for transmission AI model representation of AI information.
- the network side device may configure or modify an AI model representation manner for transmitting AI information through high-layer signaling.
- the network-side device may know the AI model representation used by the terminal side, and configure a corresponding AI model representation for each terminal. At this time, the network side may need to simultaneously train AI networks in multiple AI model representations.
- the network-side device may also instruct the terminal to use the same framework AI model representation manner as that of the network-side device.
- the network-side device may specify to use an intermediate AI model representation (such as an intermediate AI framework) to transmit AI information.
- the AI model representation configured by the configuration information for transmitting AI information is the first AI model representation
- the first communication device uses the first AI model representation
- the method further includes : the first communication device receives AI information from the second communication device and loads the received AI information into the first AI model representation; wherein the second communication device is further configured to The AI information in the second AI model representation mode is converted into the AI information in the first AI model representation mode.
- S200 introduces the process of sending AI information by the first communication device, and this embodiment introduces the process of receiving AI information by the first communication device.
- the first communication device is a network-side device
- the second communication device is a terminal
- the AI model is represented by an AI framework
- the AI framework configured by the network-side device for transmitting AI information is the same as the AI framework used by the network side.
- the terminal needs to convert the AI model under the AI framework used by itself into AI information under the configured AI framework according to the AI framework configured by the network-side device, and send it to the network-side device.
- the network-side device loads the received AI information into its own AI framework according to the pre-configured AI framework. The specific loading method is realized by the network-side device.
- the AI model representation configured by the configuration information for transmitting AI information is the second AI model representation
- the first communication device uses the first AI model representation
- the method further includes: The first communication device converts or saves the AI information in the first AI model representation mode as the AI information in the second AI model representation mode; the first communication device sends to the second communication device
- the second AI model represents AI information in a manner.
- the first communication device is a network-side device
- the second communication device is a terminal
- the AI model is represented by an AI framework
- the AI framework configured by the network-side device for transmitting AI information is the same as the AI framework used by the terminal.
- the network side device converts the AI model under its own AI framework into AI information under the configured AI framework and sends it to the terminal.
- the terminal side loads the received AI information into its own AI framework according to the configured AI framework.
- the specific loading method is implemented on the terminal side.
- generating the AI information in the first AI model representation mode by the first communication device includes: the first communication device according to the AI information used by each second communication device A model representation mode, respectively generating AI information in the first AI model representation mode.
- the first communication device is a network-side device
- the second communication device is a terminal.
- the network-side device may know the AI model representations used by multiple terminals. At this time, the network side may need to train multiple AI model representations at the same time. Under the AI network.
- the method further includes at least one of the following:
- the first communication device sends first instruction information to the second communication device, where the first instruction information is used to instruct the second communication device to use the same AI model representation mode as that of the first communication device .
- the network-side device may instruct the terminal to use the same frame AI model representation as that of the network-side device.
- the first communication device sends second indication information to the second communication device, where the second indication information is used to indicate the type of representation of the first AI model.
- the network-side device may specify to use an intermediate AI model representation (such as an intermediate AI framework) to transmit AI information.
- the first AI model representation is an intermediate AI model representation.
- S200 may also include the following steps: the first communication device receives configuration information from the third communication device, and the configuration information is used to configure or modify the AI model representation used to transmit AI information; wherein , the third communication device is further configured to send the configuration information to the second communication device.
- the first communication device and the second communication device may be two parallel nodes, for example, both are terminals; the third communication device is a third-party node, for example, the third communication device is a network-side device, and the network-side device is unified Configure or modify the AI model representation used to transmit AI information for the terminal.
- the method further includes: the first communication device sends third indication information, so that The third indication information is used to indicate the type of representation of the first AI model.
- the first communication device may indicate the type of the first AI model representation to the second communication device, so that the second communication device can load the received AI information into the second AI model representation , combined with the above-mentioned type of the first AI model representation, the AI information in the second AI model representation can be obtained.
- the first communication device sends the AI model representation used to the second communication device in advance
- the second communication device converts the AI model and parameters in the AI information according to the AI model representation corresponding to each AI information. and other content loaded into its own AI model representation.
- the network-side device may use radio resource control (Radio Resource Control, RRC) signaling (such as the third indication information above) , to configure the AI model representation mode used by the AI network of a certain function for a period of time when transmitting AI information each time, and then transmit the AI information under the AI model representation mode at the corresponding time.
- RRC Radio Resource Control
- the network side device indicates the AI model representation mode to be used through the downlink control information (Downlink Control Information, DCI) or the media access control control element (Media Access Control Control Element, MAC CE), Then pass the AI information in the AI model representation mode.
- DCI Downlink Control Information
- MAC CE Media Access Control Element
- the method further includes: the first communication device, according to the AI information to be sent, The first AI model representation is determined from multiple AI model representations; wherein, the AI information sent by the first communication device includes fourth indication information, and the fourth indication information is used to indicate the The type of the representation of the first AI model.
- the data volume of the AI information in the first AI model representation mode is the smallest among the multiple AI information in the multiple AI model representation modes.
- the AI information sent by the first communication device can use different AI model representations each time.
- the first communication device can adjust the applicable AI model representation according to the content of the AI information to be sent to ensure that the transmission bit As low as possible, and information indicating the AI model representation used is added to the transmitted information.
- the second communication device After receiving the AI information, the second communication device first determines the representation method to be used according to the information of the representation mode of the AI model, and then loads the received AI information into its own representation mode of the AI model.
- the network-side device when the first communication device is a network-side device and the second communication device is a terminal, the network-side device carries fourth indication information in the transmitted AI information, and the fourth indication information can be obtained through a special location. Indicates that the terminal first decodes the fourth indication information, and then uses a corresponding method to load the subsequent AI information.
- two nodes using different AI frameworks select an intermediate AI framework for AI information transmission.
- node A saves the AI model, parameters and other information trained under its own AI framework according to the structure of ONNX; node A passes the produced ONNX file to node B; node B according to the received ONNX file, in its own Load the ONNX file under the AI framework, and convert it to obtain the AI model and corresponding parameters under its own AI framework.
- node A and node B may be any two high-level nodes using AI functions, or may be a base station and a terminal or a terminal and a base station.
- the intermediate AI framework can adopt ONNX, or other defined AI structures.
- ONNX is a neural network interaction structure, which is mainly used for the transfer of neural networks between different AI frameworks.
- ONNX defines a general computing graph, and the computing graphs constructed by different neural network frameworks can be converted into it, so that the trained AI model can be transferred to other AI frameworks.
- This embodiment mainly introduces a channel state information (Channel State Information, CSI) feedback process based on an AI model (or AI network).
- CSI Channel State Information
- AI-based CSI feedback requires joint training of encoding and decoding.
- encoding is performed at the terminal and decoding is performed at the base station.
- the base station will continuously train its own encoding and decoding AI network.
- the base station will encode Part of the AI network structure and parameters are sent to a user. After the user receives it, they build their own AI network, and then use the AI network to encode and report the CSI information.
- the base station decodes the CSI encoding information reported by the terminal, and then restores the channel and schedules it. .
- the network side will save the AI network trained under TensorFlow as an ONNX file structure file, and pass this ONNX file to the terminal. After receiving it, load this ONNX file into your own PyTorch to get the same AI network with different description methods under PyTorch.
- the base station can send instruction information in advance, or directly configure the AI interaction model representation method for the terminal as TensorFlow when the terminal accesses, then the network side will directly save the trained AI network as a TensorFlow file structure, and pass a The user, after the user receives it, loads the TensorFlow file under PyTorch through his own method to obtain the AI network structure and parameters.
- the network side can inform the terminal that the model representation method used is PyTorch, then the network side saves the trained AI network as a PyTorch file structure, and then sends the file to the terminal, and the terminal directly loads the PyTorch file to obtain the AI network.
- the network side can interact with the terminal without prior interaction with the representation of the AI model.
- the network side saves the trained network as a TensorFlow file structure, and sends information such as TensorFlow and the corresponding version information together with the trained AI network to the The terminal, after the terminal receives it, first obtains the AI network representation as TensorFlow, then obtains the version information of TensorFlow, and then loads the received AI network into PyTorch according to the file structure of TensorFlow.
- TensorFlow and the corresponding version information can be directly transmitted in the form of a string, or can be an index (index) of some parameter combinations agreed by the protocol, and the index can be directly transmitted.
- This embodiment mainly introduces the positioning process based on the AI model (or AI network).
- AI-based positioning users can train the positioning AI network on the terminal side, and analyze the channel through the received positioning reference signal to obtain position information.
- the training of the AI network requires a large number of samples. If user A has been in a certain area for a long time, and user B has just entered this area, it will take a long time for user B to start training again. You can use the AI network that user A has trained Keep training and reduce training time.
- user A and user B use different AI frameworks, for example, user A uses Caffe2, and user B uses Keras, then user A can save the trained AI network as an ONNX file structure, and then pass it to user B. After user B receives the file, he loads the ONNX file into his own Keras structure and continues training.
- user A can save the trained AI network as the structure of caffe2, and then send information such as caffe2 and the corresponding version information to user B. After user B receives the information, he can solve the network structure as caffe2. Describe the method, and then load the caffe2 file into your own Keras.
- user A may send indication information to user B through a side link (sidelink) in advance, notifying user B that the network he uses is caffe2.
- sidelink sidelink
- user B can send an application to the network side, and the application network side will send user A's AI network to itself, and at the same time report the model representation that it can support, and the network side can support the model representation reported by user B And the known model representation that user A can support, if there is no same representation, the network side will send user A’s model representation to user B as Caffe2, and notify user A to send the AI network file saved based on the Caffe2 structure to User B, after user B receives the file, loads the Caffe2 file into its own Keras, or the network side wants user A to send user B's model representation as Keras, and user A converts his trained network into Keras representation file, send it to user B.
- the application network side will send user A's AI network to itself, and at the same time report the model representation that it can support, and the network side can support the model representation reported by user B
- the network side will send user A’s model representation to user B as Caffe2, and notify user A to send the AI network file saved based on
- the AI information transmission method provided in the embodiment of the present application may be executed by the AI information transmission device, or a control module in the AI information transmission device for executing the AI information transmission method.
- the method for transmitting the AI information performed by the AI information transmission device is taken as an example to describe the AI information transmission device provided in the embodiment of the present application.
- Fig. 3 is a schematic structural diagram of an apparatus for transmitting AI information according to an embodiment of the present application, and the apparatus may correspond to the first communication device in other embodiments.
- the device 300 includes the following modules.
- the generating module 302 may be configured to generate AI information in a first AI model representation mode; wherein, the AI information in the first AI model representation mode supports being converted into AI information in a second AI model representation mode.
- the sending module 304 may be configured to send the AI information in the first AI model representation mode to the second communication device; wherein the second communication device uses the second AI model representation mode.
- the AI information transmission device generates the AI information in the first AI model representation mode and sends it to the second communication device, because the AI information in the first AI model representation mode supports being converted into the second AI model AI information in the representation mode, so that the second communication device using the second AI model representation mode receives the AI information in the first AI model representation mode, and loads it into the second AI model representation mode to obtain the second AI AI information in model representation.
- communication devices using different AI model representations can maintain a consistent understanding of AI information, so that AI information can be transmitted between communication devices using different AI model representations, which is conducive to improving communication system performance.
- the device uses a third AI model representation
- the generation module 302 is configured to convert or save AI information in the third AI model representation as the first AI AI information in a model representation mode; or, the apparatus uses the first AI model representation mode.
- the sending module 304 is further configured to send configuration information, where the configuration information is used to configure or modify an AI model representation manner for transmitting AI information.
- the AI model representation used for transmitting AI information is the first AI model representation
- the device uses the first AI model representation
- the device further includes receiving A module, configured to receive AI information from the second communication device and load the received AI information into the first AI model representation; wherein, the second communication device is also configured to add the first AI information
- the AI information in the representation mode of the second AI model is converted into the AI information in the representation mode of the first AI model.
- the AI model representation used to transmit AI information is the second AI model representation, and the device uses the first AI model representation; the generating module 302 also uses Converting or saving the AI information in the first AI model representation mode to the AI information in the second AI model representation mode; the sending module 304 is further configured to send the AI information to the second communication device The AI information in the second AI model representation mode.
- the generating module 302 is configured to respectively generate AI information in the first AI model representation mode according to the AI model representation mode used by each of the second communication devices.
- the sending module 304 is further configured to at least one of the following: 1) Send first indication information to the second communication device, where the first indication information is used to indicate that the first The second communication device uses the same AI model representation as that of the apparatus; 2) sending second indication information to the second communication device, where the second indication information is used to indicate the type of the first AI model representation.
- the apparatus further includes a receiving module, configured to receive configuration information from a third communication device, where the configuration information is used to configure or modify an AI model representation for transmitting AI information;
- the third communication device is further configured to send the configuration information to the second communication device.
- the sending module 304 is further configured to send third indication information, where the third indication information is used to indicate the type of the first AI model representation.
- the device further includes a determining module, configured to determine the first AI model representation from multiple AI model representations according to the AI information to be sent; wherein, the sending The AI information sent by module 304 includes fourth indication information, where the fourth indication information is used to indicate the type of representation of the first AI model.
- the data volume of the AI information in the first AI model representation mode is the smallest among the multiple AI information in the multiple AI model representation modes.
- the device 300 can refer to the process of the method 200 corresponding to the embodiment of the present application, and each unit/module in the device 300 and the above-mentioned other operations and/or functions are respectively in order to realize the corresponding process in the method 200, And can achieve the same or equivalent technical effect, for the sake of brevity, no more details are given here.
- the AI information transmission device in the embodiment of the present application may be a device, a device with an operating system or an electronic device, or a component, an integrated circuit, or a chip in a terminal.
- the apparatus or electronic equipment may be a mobile terminal or a non-mobile terminal.
- the mobile terminal may include but not limited to the types of terminals 11 listed above, and the non-mobile terminal may be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television ( television, TV), teller machines or self-service machines, etc., are not specifically limited in this embodiment of the present application.
- the AI information transmission device provided by the embodiment of the present application can realize each process realized by the method embodiment in FIG. 2 and achieve the same technical effect. To avoid repetition, details are not repeated here.
- this embodiment of the present application further provides a communication device 400, including a processor 401, a memory 402, and programs or instructions stored in the memory 402 and operable on the processor 401,
- a communication device 400 including a processor 401, a memory 402, and programs or instructions stored in the memory 402 and operable on the processor 401
- the communication device 400 is a terminal
- the program or instruction is executed by the processor 401
- each process of the above-mentioned AI information transmission method embodiment can be realized, and the same technical effect can be achieved.
- the communication device 400 is a network-side device, when the program or instruction is executed by the processor 401, each process of the above-mentioned AI information transmission method embodiment can be achieved, and the same technical effect can be achieved. To avoid repetition, details are not repeated here.
- the embodiment of the present application also provides a terminal, including a processor and a communication interface, and the processor is used to generate AI information in a first AI model representation mode; wherein, the AI information support in the first AI model representation mode is converted into The AI information in the second AI model representation mode, the communication interface is used to send the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses the second AI model representation Way.
- This terminal embodiment corresponds to the above-mentioned terminal-side method embodiment, and each implementation process and implementation mode of the above-mentioned method embodiment can be applied to this terminal embodiment, and can achieve the same technical effect.
- FIG. 5 is a schematic diagram of a hardware structure of a terminal implementing an embodiment of the present application.
- the terminal 500 includes but is not limited to: a radio frequency unit 501, a network module 502, an audio output unit 503, an input unit 504, a sensor 505, a display unit 506, a user input unit 507, an interface unit 508, a memory 509, and a processor 510, etc. at least some of the components.
- the terminal 500 can also include a power supply (such as a battery) for supplying power to various components, and the power supply can be logically connected to the processor 510 through the power management system, so as to manage charging, discharging, and power consumption through the power management system. Management and other functions.
- a power supply such as a battery
- the terminal structure shown in FIG. 5 does not constitute a limitation on the terminal, and the terminal may include more or fewer components than shown in the figure, or combine certain components, or arrange different components, which will not be repeated here.
- the input unit 504 may include a graphics processing unit (Graphics Processing Unit, GPU) 5041 and a microphone 5042, and the graphics processing unit 5041 is used in a video capture mode or an image capture mode by an image capture device (such as the image data of the still picture or video obtained by the camera) for processing.
- the display unit 506 may include a display panel 5061, and the display panel 5061 may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like.
- the user input unit 507 includes a touch panel 5071 and other input devices 5072 .
- the touch panel 5071 is also called a touch screen.
- the touch panel 5071 may include two parts, a touch detection device and a touch controller.
- Other input devices 5072 may include, but are not limited to, physical keyboards, function keys (such as volume control keys, switch keys, etc.), trackballs, mice, and joysticks, which will not be repeated here.
- the radio frequency unit 501 receives the downlink data from the network side device, and processes it to the processor 510; in addition, sends the uplink data to the network side device.
- the radio frequency unit 501 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, and the like.
- the memory 509 can be used to store software programs or instructions as well as various data.
- the memory 509 may mainly include a program or instruction storage area and a data storage area, wherein the program or instruction storage area may store an operating system, at least one application or instruction required by a function (such as a sound playback function, an image playback function, etc.) and the like.
- the memory 509 may include a high-speed random access memory, and may also include a non-transitory memory, wherein the non-transitory memory may be a read-only memory (Read Only Memory, ROM), a programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electrically erasable programmable read-only memory (Electrically EPROM, EEPROM) or flash memory.
- ROM Read Only Memory
- PROM programmable read-only memory
- Erasable PROM Erasable PROM
- EPROM electrically erasable programmable read-only memory
- flash memory for example at least one disk storage device, flash memory device, or other non-transitory solid state storage device.
- the processor 510 may include one or more processing units; optionally, the processor 510 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, application programs or instructions, etc., Modem processors mainly handle wireless communications, such as baseband processors. It can be understood that the foregoing modem processor may not be integrated into the processor 510 .
- the processor 510 may be configured to generate AI information in a first AI model representation mode; wherein, the AI information in the first AI model representation mode supports being converted into AI information in a second AI model representation mode;
- the radio frequency unit 501 may be configured to send AI information in the first AI model representation mode to a second communication device; wherein the second communication device uses the second AI model representation mode.
- the terminal generates AI information in the first AI model representation mode and sends it to the second communication device. Since the AI information support in the first AI model representation mode is converted into the AI information in the second AI model representation mode AI information, in this way, the second communication device using the second AI model representation mode receives the AI information in the first AI model representation mode, and loads it into the second AI model representation mode, and obtains the AI information in the second AI model representation mode. AI information.
- communication devices using different AI model representations can maintain a consistent understanding of AI information, so that AI information can be transmitted between communication devices using different AI model representations, which is conducive to improving communication system performance.
- the terminal 500 provided in the embodiment of the present application can also implement the various processes in the above embodiment of the AI information transmission method, and can achieve the same technical effect. To avoid repetition, details are not repeated here.
- the embodiment of the present application also provides a network side device, including a processor and a communication interface, the processor is used to generate AI information in the first AI model representation mode; wherein, the AI information in the first AI model representation mode supports being Converted to AI information in the second AI model representation mode, the communication interface is used to send the AI information in the first AI model representation mode to the second communication device; wherein, the second communication device uses the second AI Model representation.
- the network-side device embodiment corresponds to the above-mentioned network-side device method embodiment, and each implementation process and implementation mode of the above-mentioned method embodiment can be applied to this network-side device embodiment, and can achieve the same technical effect.
- the embodiment of the present application also provides a network side device.
- the network side device 600 includes: an antenna 61 , a radio frequency device 62 , and a baseband device 63 .
- the antenna 61 is connected to the radio frequency device 62 .
- the radio frequency device 62 receives information through the antenna 61, and sends the received information to the baseband device 63 for processing.
- the baseband device 63 processes the information to be sent and sends it to the radio frequency device 62
- the radio frequency device 62 processes the received information and sends it out through the antenna 61 .
- the foregoing frequency band processing device may be located in the baseband device 63 , and the method performed by the network side device in the above embodiments may be implemented in the baseband device 63 , and the baseband device 63 includes a processor 64 and a memory 65 .
- the baseband device 63 can include at least one baseband board, for example, a plurality of chips are arranged on the baseband board, as shown in FIG. The operation of the network side device shown in the above method embodiments.
- the baseband device 63 may also include a network interface 66 for exchanging information with the radio frequency device 62, such as a Common Public Radio Interface (CPRI).
- CPRI Common Public Radio Interface
- the network-side device in the embodiment of the present application also includes: instructions or programs stored in the memory 65 and operable on the processor 64, and the processor 64 calls the instructions or programs in the memory 65 to execute the modules shown in FIG. 3 To avoid duplication, the method of implementation and to achieve the same technical effect will not be repeated here.
- the embodiment of the present application also provides a readable storage medium.
- the readable storage medium may be volatile or non-volatile.
- the readable storage medium may be transient or non-volatile. Transient, the readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the various processes of the above-mentioned AI information transmission method embodiments can be achieved, and the same technical effect can be achieved. In order to avoid Repeat, no more details here.
- the processor may be the processor in the terminal described in the foregoing embodiments.
- the readable storage medium includes computer readable storage medium, such as computer read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.
- the embodiment of the present application further provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, the processor is used to run programs or instructions, and realize the implementation of the above-mentioned AI information transmission method
- the chip includes a processor and a communication interface
- the communication interface is coupled to the processor
- the processor is used to run programs or instructions, and realize the implementation of the above-mentioned AI information transmission method
- the chip mentioned in the embodiment of the present application may also be called a system-on-chip, a system-on-chip, a system-on-a-chip, or a system-on-a-chip.
- An embodiment of the present application further provides a computer program product, the computer program product is stored in a non-transitory storage medium, and the computer program product is executed by at least one processor to implement the above-mentioned AI information transmission method embodiment.
- Each process can achieve the same technical effect, so in order to avoid repetition, it will not be repeated here.
- the embodiment of the present application further provides a communication device, which is configured to execute each process of the above-mentioned AI information transmission method embodiment, and can achieve the same technical effect. To avoid repetition, details are not repeated here.
- the term “comprising”, “comprising” or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or device. Without further limitations, an element defined by the phrase “comprising a " does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element.
- the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved. Functions are performed, for example, the described methods may be performed in an order different from that described, and various steps may also be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
- the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation.
- the technical solution of the present application can be embodied in the form of computer software products, which are stored in a storage medium (such as ROM/RAM, magnetic disk, etc.) , CD-ROM), including several instructions to enable a terminal (which may be a mobile phone, computer, server, air conditioner, or network-side device, etc.) to execute the methods described in various embodiments of the present application.
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本申请实施例公开了一种AI信息的传输方法和设备,属于通信技术领域。本申请实施例的AI信息的传输方法包括:第一通信设备生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;所述第一通信设备向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
Description
相关申请的交叉引用
本申请要求在2021年11月16日提交的中国专利申请第202111358543.7号的优先权,该中国专利申请的全部内容通过引用包含于此。
本申请属于通信技术领域,具体涉及一种人工智能(Artificial Intelligence,AI)信息的传输方法和设备,该设备可以包括AI信息的传输装置,终端或网络侧设备等通信设备。
将AI技术运用到通信系统中可以明显提升通信系统性能。然而,由于常见的AI模型表示方式(如AI框架)较多,不同的AI模型表示方式的侧重点、甚至支持的开发语言不尽相同,对AI模型的描述和功能的实现方式也不同,导致使用不同AI模型表示方式的两个或多个通信设备之间无法传递AI信息,影响通信系统的性能。
发明内容
本申请实施例提供一种AI信息的传输方法和设备,能够解决因使用不同AI表示方法的通信设备之间无法传递AI信息,影响通信系统性能的问题。
第一方面,提供了一种AI信息的传输方法,包括:第一通信设备生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;所述第一通信设备向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
第二方面,提供了一种AI信息的传输装置,包括:生成模块,用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;发送模块,用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
第三方面,提供了一种通信设备,该通信设备包括处理器、存储器及存储在所述存储器上并可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如第一方面所述的方法。
第四方面,提供了一种通信设备,包括处理器及通信接口,其中,所述处理器用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,所述通信接口用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
第五方面,提供了一种可读存储介质,所述可读存储介质上存储程序或指令,所述程序或指令被处理器执行时实现如第一方面所述的方法。
第六方面,提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如第一方面所述的方法。
第七方面,提供了一种计算机程序/程序产品,所述计算机程序/程序产品被存储在非瞬态的存储介质中,所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的方法。
在本申请实施例中,第一通信设备生成第一AI模型表示方式下的AI信息并发送给第二通信设备,由于第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,这样,使用第二AI模型表示方式的第二通信设备接收第一AI模型表示方式下的AI信息,并将其加载到第二AI模型表示方式下,得到第二AI模型表示方式下的AI信息。通过本申请实 施例,使用不同AI模型表示方式的通信设备可以对AI信息的理解保持一致,使得使用不同AI模型表示方式的通信设备之间可以进行AI信息的传输,有利于提高通信系统性能。
图1是根据本申请实施例的无线通信系统的示意图;
图2是根据本申请实施例的AI信息的传输方法的示意性流程图;
图3是根据本申请实施例的AI信息的传输装置的结构示意图;
图4是根据本申请实施例的通信设备的结构示意图;
图5是根据本申请实施例的终端的结构示意图;
图6是根据本申请实施例的网络侧设备的结构示意图。
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员所获得的所有其他实施例,都属于本申请保护的范围。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象,而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换,以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施,且“第一”、“第二”所区别的对象通常为一类,并不限定对象的个数,例如第一对象可以是一个,也可以是多个。此外,说明书以及权利要求中“和/或”表示所连接对象的至少其中之一,字符“/”一般表示前后关联对象是一种“或”的关系。
值得指出的是,本申请实施例所描述的技术不限于长期演进型(Long Term Evolution,LTE)/LTE的演进(LTE-Advanced,LTE-A)系统,还可用 于其他无线通信系统,诸如码分多址(Code Division Multiple Access,CDMA)、时分多址(Time Division Multiple Access,TDMA)、频分多址(Frequency Division Multiple Access,FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access,OFDMA)、单载波频分多址(Single carrier-Frequency Division Multiple Access,SC-FDMA)和其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用,所描述的技术既可用于以上提及的系统和无线电技术,也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio,NR)系统,并且在以下大部分描述中使用NR术语,这些技术也可应用于NR系统应用以外的应用,如第6代(6
th Generation,6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的示意图。无线通信系统包括终端11和网络侧设备12。其中,终端11也可以称作终端设备或者用户终端(User Equipment,UE),终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字助理(Personal Digital Assistant,PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer,UMPC)、移动上网装置(Mobile Internet Device,MID)、增强现实(augmented reality,AR)/虚拟现实(virtual reality,VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(VUE)、行人终端(PUE)、智能家居(具有无线通信功能的家居设备,如冰箱、电视、洗衣机或者家具等)等终端侧设备,可穿戴式设备包括:智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装、游戏机等。需要说明的是,在本申请实施例并不限定终端11的具体类型。网络侧设备12可以是基站或核心网,其中,基站可被称为节点B、演进节点B、接入点、基收发机站(Base Transceiver Station,BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set,BSS)、扩展服务集(Extended Service Set,ESS)、B节点、演进型B节点(eNB)、下一代节点B(gNB)、家用B节点、家用演进型B节点、无线局域网(Wireless Local Area Network,WLAN)接入点、无线保真(Wireless Fidelity,WiFi)节点、发送接收点(Transmitting Receiving Point,TRP)或所述领域中其他某个合适的术语,只要达到相同的技术效果,所述基站不限于特定技术词汇,需要说明的是,在本申请实施例中仅以NR系统中的基站为例,但是并不限定基站的具体类型。
下面结合附图,通过一些实施例及其应用场景对本申请实施例提供的AI信息的传输方法和设备进行详细地说明。
如图2所示,本申请实施例提供一种AI信息的传输方法200,该方法可以由第一通信设备执行,换言之,该方法可以由安装在第一通信设备的软件或硬件来执行,该第一通信设备可以是网络侧设备或终端等,该方法包括如下步骤。
S202:第一通信设备生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息。
S204:第一通信设备向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
本申请各个实施例中提到的AI模型表示方式,如第一AI模型表示方式,第二AI模型表示方式,第三AI模型表示方式等,具体可以是AI框架,也可以是独立的表示AI模型的模板或者数据结构(如开放式神经网络交换(Open Neural Network Exchange,ONNX)等),上述模板或者数据结构只负责表示AI模型,不负责训练推断等。上述AI框架例如包括TensorFlow,PyTorch,Keras,MXNet,Caffe2等,每一个AI框架都会使用自己的方法对AI模型进行描述,完成AI模型的搭建、训练、推断等操作。
本申请各个实施例中提到的AI信息可以包括AI模型(或称AI网络)、 参数、结构等,该AI模型例如包括神经网络模型、决策树模型、支持向量机模型、贝叶斯分类器模型等。后续实施例多以AI模型为例来对该实施例进行介绍,可以理解,该AI信息并不以AI模型为限。
在一个例子中,所述第一通信设备使用第三AI模型表示方式,S202中所述第一通信设备生成第一AI模型表示方式下的AI信息包括:所述第一通信设备将所述第三AI模型表示方式下的AI信息转换或保存为所述第一AI模型表示方式下的AI信息。该例子中提到的“转换”,例如,第三AI模型表示方式是TensorFlow,第一AI模型表示方式是PyTorch,第一通信设备将TensorFlow下的AI模型保存成TensorFlow格式的信息,然后转换成PyTorch格式的信息。该例子中提到的“保存”,例如,第三AI模型表示方式是TensorFlow,第一AI模型表示方式是PyTorch,第一通信设备将TensorFlow下的AI模型保存成PyTorch格式的信息。
该例子中,第一AI模型表示方式可以称作是中间AI模型表示方式(如中间AI框架),也即,第一通信设备和第二通信设备选择一个中间AI模型表示方式进行AI信息的传输,该中间AI模型表示方式支持被转换为第二AI模型表示方式下的AI信息。这样,使用第二AI模型表示方式的第二通信设备在接收到第一AI模型表示方式下的AI信息之后,即可将其加载到第二AI模型表示方式下,得到第二AI模型表示方式下的AI信息。可选地,该中间AI模型表示方式可以为开放式神经网络交换(Open Neural Network Exchange,ONNX)。
在另一个例子中,所述第一通信设备使用所述第一AI模型表示方式。该例子中,第一通信设备可以使用所述第一AI模型表示方式(如AI框架)生成AI信息(如AI模型)。例如,第一通信设备可以使用AI框架对AI模型进行训练,得到训练后的AI模型。
该例子中,第二通信设备可以通过预先约定的方式,或者是根据第一通信设备的指示,获知第一AI模型表示方式的类型,例如,第二通信设备可以 获知第一AI模型表示方式是PyTorch,第一AI模型表示方式下的AI信息是PyTorch下的AI模型以及参数等。这样,第二通信设备即可将接收到的AI信息加载到第二AI模型表示方式中,结合上述第一AI模型表示方式的类型,即可得到第二AI模型表示方式下的AI信息,例如,通过MXNet下开发的PyTorch转换MXNet模块得到MXNet下的AI模型。
本申请实施例提供的AI信息的传输方法,第一通信设备生成第一AI模型表示方式下的AI信息并发送给第二通信设备,由于第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,这样,使用第二AI模型表示方式的第二通信设备接收第一AI模型表示方式下的AI信息,并将其加载到第二AI模型表示方式下,得到第二AI模型表示方式下的AI信息。通过本申请实施例,使用不同AI模型表示方式的通信设备可以对AI信息的理解保持一致,使得使用不同AI模型表示方式的通信设备之间可以进行AI信息的传输,有利于提高通信系统性能。
实施例200主要介绍了AI信息的传输过程,实际上,实施例200之前,还可以预先配置用于传输AI信息的AI模型表示方式;在实施例200的执行过程中或执行完成后,还可以修改用于传输AI信息的AI模型表示方式。
可选地,实施例200提供的方法还可以包括如下步骤:所述第一通信设备发送配置信息,所述第二通信设备还可以接收该配置信息,该配置信息用于配置或修改用于传输AI信息的AI模型表示方式。
该实施例例如,网络侧设备可以通过高层信令配置或者修改用于传输AI信息的AI模型表示方式。可选地,网络侧设备可以已知终端侧使用的AI模型表示方式,给每个终端配置对应的AI模型表示方式,此时网络侧可能需要同时训练多个AI模型表示方式下的AI网络。可选地,网络侧设备还可以指示终端使用与网络侧设备相同的框架AI模型表示方式。可选地,网络侧设备可以指定使用中间AI模型表示方式(如中间AI框架)传输AI信息。
在一个例子中,配置信息配置的用于传输AI信息的AI模型表示方式为 所述第一AI模型表示方式,所述第一通信设备使用所述第一AI模型表示方式,所述方法还包括:所述第一通信设备接收来自于所述第二通信设备的AI信息并将接收到的AI信息加载到所述第一AI模型表示方式中;其中,所述第二通信设备还用于将所述第二AI模型表示方式下的AI信息转换为所述第一AI模型表示方式下的AI信息。需要说明的是,S200介绍的是第一通信设备发送AI信息的过程,该实施例介绍的是第一通信设备接收AI信息的过程。
该例子例如,第一通信设备是网络侧设备,第二通信设备是终端,AI模型表示方式为AI框架,网络侧设备配置的用于传输AI信息的AI框架与网络侧使用的AI框架相同。这样,终端需要根据网络侧设备配置的AI框架,将自己使用的AI框架下的AI模型转换成配置的AI框架下的AI信息,并发送给网络侧设备。网络侧设备接收到AI信息之后,按照预先配置的AI框架,将接收到的AI信息加载到自己的AI框架内,具体的加载方式为网络侧设备实现。
在另一个例子中,配置信息配置的用于传输AI信息的AI模型表示方式为所述第二AI模型表示方式,所述第一通信设备使用第一AI模型表示方式,所述方法还包括:所述第一通信设备将所述第一AI模型表示方式下的AI信息转换或保存为所述第二AI模型表示方式下的AI信息;所述第一通信设备向所述第二通信设备发送所述第二AI模型表示方式下的AI信息。
该例子例如,第一通信设备是网络侧设备,第二通信设备是终端,AI模型表示方式为AI框架,网络侧设备配置的用于传输AI信息的AI框架与终端使用的AI框架相同。这样,网络侧设备根据配置的AI框架,将自己的AI框架下的AI模型转换成配置的AI框架下的AI信息并发送给终端。终端侧接收到AI信息之后,按照配置的AI框架,将接收到的AI信息加载到自己的AI框架内,具体加载方式为终端侧实现。
可选地,在前文各个实施例的基础上,所述第一通信设备生成第一AI模型表示方式下的AI信息包括:所述第一通信设备根据每个所述第二通信设 备使用的AI模型表示方式,分别生成所述第一AI模型表示方式下的AI信息。
该例子例如,第一通信设备是网络侧设备,第二通信设备是终端,网络侧设备可以已知多个终端分别使用的AI模型表示方式,此时网络侧可能需要同时训练多个AI模型表示方式下的AI网络。
可选地,在前文各个实施例的基础上,所述方法还包括如下至少之一:
1)所述第一通信设备向所述第二通信设备发送第一指示信息,所述第一指示信息用于指示所述第二通信设备使用与所述第一通信设备相同的AI模型表示方式。例如网络侧设备可以指示终端使用与网络侧设备相同的框架AI模型表示方式。
2)所述第一通信设备向所述第二通信设备发送第二指示信息,所述第二指示信息用于指示所述第一AI模型表示方式的类型。网络侧设备可以指定使用中间AI模型表示方式(如中间AI框架)传输AI信息,此时,第一AI模型表示方式为中间AI模型表示方式。
前文多个实施例多以第一通信设备是网络侧设备,第二通信设备是终端为例进行介绍,实际上,第一通信设备与第二通信设备还可以是两个平行的节点,例如都是终端,这样,S200还可以包括如下步骤:所述第一通信设备接收来自于第三通信设备的配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式;其中,所述第三通信设备还用于向所述第二通信设备发送所述配置信息。
该实施例中,第一通信设备与第二通信设备可以是两个平行的节点,例如都是终端;第三通信设备是第三方节点,如第三通信设备是网络侧设备,网络侧设备统一为终端配置或修改用于传输AI信息的AI模型表示方式。
可选地,在实施例200的基础上,所述第一通信设备生成第一AI模型表示方式下的AI信息之前,所述方法还包括:所述第一通信设备发送第三指示信息,所述第三指示信息用于指示所述第一AI模型表示方式的类型。
该实施例中,所述第一通信设备可以向第二通信设备指示第一AI模型表示方式的类型,这样,第二通信设备即可将接收到的AI信息加载到第二AI模型表示方式中,结合上述第一AI模型表示方式的类型,即可得到第二AI模型表示方式下的AI信息。
该实施例具体例如,第一通信设备将使用的AI模型表示方式提前发送给第二通信设备,第二通信设备根据每个AI信息对应的AI模型表示方式,将AI信息中的AI模型、参数等内容加载到自己的AI模型表示方式内。
该实施例中,在第一通信设备是网络侧设备,第二通信设备是终端的情况下,网络侧设备可以通过无线资源控制(Radio Resource Control,RRC)信令(如上述第三指示信息),来配置一段时间内某个功能的AI网络在每次传递AI信息的时候使用的AI模型表示方式,然后到了对应时刻传递AI模型表示方式下的AI信息。或者,网络侧设备在每次传递AI信息之前,通过下行控制信息(Downlink Control Information,DCI)或者媒体接入控制控制单元(Media Access Control Control Element,MAC CE)来指示使用的AI模型表示方式,然后传递AI模型表示方式下的AI信息。
可选地,在实施例200的基础上,所述第一通信设备生成第一AI模型表示方式下的AI信息之前,所述方法还包括:所述第一通信设备根据待发送的AI信息,从多个AI模型表示方式中确定出所述第一AI模型表示方式;其中,所述第一通信设备发送的AI信息中包括有第四指示信息,所述第四指示信息用于指示所述第一AI模型表示方式的类型。可选地,所述第一AI模型表示方式下的AI信息的数据量,是所述多个AI模型表示方式下的多个AI信息中数据量最小的。
该实施例具体例如,第一通信设备每次发送的AI信息可以使用不同的AI模型表示方式,例如,第一通信设备可以根据待发送的AI信息内容调整适用的AI模型表示方式,保证传输比特尽量低,并且在传输的信息中加入了指示使用的AI模型表示方式的信息。第二通信设备接收到AI信息之后,先 根据AI模型表示方式的信息确定使用的表示方法,然后将接收到的AI信息加载到自己的AI模型表示方式内。
该实施例中,在第一通信设备是网络侧设备,第二通信设备是终端的情况下,网络侧设备在传递的AI信息中携带第四指示信息,第四指示信息可以通过特殊的位置来表示,终端先解出来第四指示信息,然后使用对应的方法加载后面的AI信息。
为详细说明本申请实施例提供的AI信息的传输方法,以下将结合几个具体的实施例进行说明。
实施例一
该实施例中,两个使用不同AI框架的节点(如第一通信设备和第二通信设备)选择一个中间AI框架进行AI信息传递。
例如,节点A将在自己AI框架下训练好的AI模型、参数等信息按照ONNX的结构进行保存;节点A将生产的ONNX文件传递给节点B;节点B根据接收到的ONNX文件,在自己的AI框架下加载ONNX文件,转换得到自己AI框架下的AI模型及对应的参数。
该实施例中,节点A和节点B可以是任意的两个使用AI功能的高层节点,也可以是基站和终端或者终端和基站。中间AI框架可以采用ONNX,也可以采用其他定义的AI结构。其中,ONNX是一种神经网络交互结构,主要用于不同AI框架之间的神经网络的传递。ONNX定义了一种通用的计算图,不同神经网络框架构建的计算图都能转化为它,从而可以实现将训练好的AI模型传递给其他AI框架。
实施例二
该实施例主要介绍基于AI模型(或称AI网络)的信道状态信息(Channel State Information,CSI)反馈过程。
基于AI的CSI反馈需要将编码和解码联合训练,通常编码在终端执行,解码在基站执行,基站会不断地训练自己的编码解码AI网络,当有终端接入 到小区的时候,基站会将编码部分的AI网络结构和参数发送个用户,用户接收到之后,搭建自己的AI网络,然后利用AI网络进行CSI信息的编码并上报,基站根据终端上报的CSI编码信息进行解码,然后恢复信道并调度。
如果基站和用户使用的AI框架不同,例如终端使用的PyTorch,网络侧使用的TensorFlow,网络侧将在TensorFlow下训练好的AI网络保存成ONNX文件结构的文件,将这个ONNX文件传递给终端,终端接收到之后,将这个ONNX文件加载到自己的PyTorch中,得到PyTorch下不同描述方法的相同的AI网络。
可选的,基站可以提前发送指示信息,或者在终端接入的时候直接给终端配置AI交互的模型表示方法为TensorFlow,则网络侧直接将训练好的AI网络保存成TensorFlow的文件结构,传递个用户,用户接收到之后,通过自己的方法,在PyTorch下加载TensorFlow的文件,得到AI网络结构和参数。
或者,网络侧可以通知终端使用的模型表示方法为PyTorch,则网络侧自己将训练好的AI网络保存成PyTorch的文件结构,然后将文件发送给终端,终端直接加载PyTorch文件,得到AI网络。
可选的,网络侧可以和终端不事先交互AI模型的表示方式,网络侧将训练好的网络保存成TensorFlow的文件结构,将TensorFlow以及对应的版本信息等信息和训练好的AI网络一起发送给终端,终端接收之后,首先获取AI网络表示方式为TensorFlow,然后获取TensorFlow的版本信息,然后按照TensorFlow的文件结构将接收到的AI网络加载到PyTorch中。
具体地,TensorFlow和对应的版本信息可以是直接以字符串的方式传递,也可以是协议约定好的一些参数组合的索引(index),直接传递这个索引。
实施例三
该实施例主要介绍基于AI模型(或称AI网络)的定位过程。
基于AI的定位有很多方法,例如,用户可以在终端侧进行定位AI网络的训练,通过接收到的定位参考信号,对信道进行分析,得到位置信息。
AI网络的训练需要大量的样本,如果用户A长时间处于某个区域内,用户B刚刚进入这个区域,此时用户B重新开始训练需要很长的时间,可以使用用户A已经训练好的AI网络继续训练,减少训练时间。
如果用户A和用户B使用的AI框架不同,例如用户A使用的Caffe2,用户B使用的是Keras,此时用户A可以将训练好的AI网络保存为ONNX的文件结构,然后传递给用户B,用户B收到文件之后,将ONNX文件加载到自己的Keras结构中,继续训练。
可选的,用户A可以将训练好的AI网络保存为caffe2的结构,然后将caffe2和对应的版本信息等信息一并发送个用户B,用户B收到信息之后,解出网络结构为caffe2的描述方法,然后将caffe2的文件加载到自己的Keras中。
或者用户A可以提前通过侧链路(sidelink)向用户B发送指示信息,通知用户B自己使用的网络为caffe2。
可选的,用户B可以向网络侧发送申请,申请网络侧将用户A的AI网络发送给自己,并同时上报自己可以支持的模型表示方式,网络侧根据用户B上报的可以支持的模型表示方式和已知的用户A可以支持的模型表示方式,发现没有相同的表示方式,则网络侧向用户B发送用户A的模型表示方式为Caffe2,并且通知用户A发送基于Caffe2结构保存的AI网络文件给用户B,用户B收到文件之后,将Caffe2的文件加载到自己的Keras中,或者网络侧想用户A发送用户B的模型表示方式为Keras,用户A将自己训练好的网络转成Keras表示方法的文件,发送给用户B。
需要说明的是,本申请实施例提供的AI信息的传输方法,执行主体可以为AI信息的传输装置,或者,该AI信息的传输装置中的用于执行AI信息的传输方法的控制模块。本申请实施例中以AI信息的传输装置执行AI信息的传输方法为例,说明本申请实施例提供的AI信息的传输装置。
图3是根据本申请实施例的AI信息的传输装置的结构示意图,该装置可 以对应于其他实施例中的第一通信设备。如图3所示,装置300包括如下模块。
生成模块302,可以用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息。
发送模块304,可以用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
在本申请实施例中,AI信息的传输装置生成第一AI模型表示方式下的AI信息并发送给第二通信设备,由于第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,这样,使用第二AI模型表示方式的第二通信设备接收第一AI模型表示方式下的AI信息,并将其加载到第二AI模型表示方式下,得到第二AI模型表示方式下的AI信息。通过本申请实施例,使用不同AI模型表示方式的通信设备可以对AI信息的理解保持一致,使得使用不同AI模型表示方式的通信设备之间可以进行AI信息的传输,有利于提高通信系统性能。
可选地,作为一个实施例,所述装置使用第三AI模型表示方式,所述生成模块302,用于将所述第三AI模型表示方式下的AI信息转换或保存为所述第一AI模型表示方式下的AI信息;或者,所述装置使用所述第一AI模型表示方式。
可选地,作为一个实施例,所述发送模块304,还用于发送配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式。
可选地,作为一个实施例,所述用于传输AI信息的AI模型表示方式为所述第一AI模型表示方式,所述装置使用所述第一AI模型表示方式,所述装置还包括接收模块,用于接收来自于所述第二通信设备的AI信息并将接收到的AI信息加载到所述第一AI模型表示方式中;其中,所述第二通信设备还用于将所述第二AI模型表示方式下的AI信息转换为所述第一AI模型表 示方式下的AI信息。
可选地,作为一个实施例,所述用于传输AI信息的AI模型表示方式为所述第二AI模型表示方式,所述装置使用第一AI模型表示方式;所述生成模块302,还用于将所述第一AI模型表示方式下的AI信息转换或保存为所述第二AI模型表示方式下的AI信息;所述发送模块304,还用于向所述第二通信设备发送所述第二AI模型表示方式下的AI信息。
可选地,作为一个实施例,所述生成模块302,用于根据每个所述第二通信设备使用的AI模型表示方式,分别生成所述第一AI模型表示方式下的AI信息。
可选地,作为一个实施例,所述发送模块304,还用于如下至少之一:1)向所述第二通信设备发送第一指示信息,所述第一指示信息用于指示所述第二通信设备使用与所述装置相同的AI模型表示方式;2)向所述第二通信设备发送第二指示信息,所述第二指示信息用于指示所述第一AI模型表示方式的类型。
可选地,作为一个实施例,所述装置还包括接收模块,用于接收来自于第三通信设备的配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式;其中,所述第三通信设备还用于向所述第二通信设备发送所述配置信息。
可选地,作为一个实施例,所述发送模块304,还用于发送第三指示信息,所述第三指示信息用于指示所述第一AI模型表示方式的类型。
可选地,作为一个实施例,所述装置还包括确定模块,用于根据待发送的AI信息,从多个AI模型表示方式中确定出所述第一AI模型表示方式;其中,所述发送模块304发送的AI信息中包括有第四指示信息,所述第四指示信息用于指示所述第一AI模型表示方式的类型。
可选地,作为一个实施例,所述第一AI模型表示方式下的AI信息的数据量,是所述多个AI模型表示方式下的多个AI信息中数据量最小的。
根据本申请实施例的装置300可以参照对应本申请实施例的方法200的流程,并且,该装置300中的各个单元/模块和上述其他操作和/或功能分别为了实现方法200中的相应流程,并且能够达到相同或等同的技术效果,为了简洁,在此不再赘述。
本申请实施例中的AI信息的传输装置可以是装置,具有操作系统的装置或电子设备,也可以是终端中的部件、集成电路、或芯片。该装置或电子设备可以是移动终端,也可以为非移动终端。示例性的,移动终端可以包括但不限于上述所列举的终端11的类型,非移动终端可以为服务器、网络附属存储器(Network Attached Storage,NAS)、个人计算机(personal computer,PC)、电视机(television,TV)、柜员机或者自助机等,本申请实施例不作具体限定。
本申请实施例提供的AI信息的传输装置能够实现图2的方法实施例实现的各个过程,并达到相同的技术效果,为避免重复,这里不再赘述。
可选的,如图4所示,本申请实施例还提供一种通信设备400,包括处理器401,存储器402,存储在存储器402上并可在所述处理器401上运行的程序或指令,例如,该通信设备400为终端时,该程序或指令被处理器401执行时实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果。该通信设备400为网络侧设备时,该程序或指令被处理器401执行时实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供一种终端,包括处理器和通信接口,处理器用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,通信接口用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。该终端实施例是与上述终端侧方法实施例对应的,上述方法实施例的各个实施过程和实现方式均可适用于该 终端实施例中,且能达到相同的技术效果。具体地,图5为实现本申请实施例的一种终端的硬件结构示意图。
该终端500包括但不限于:射频单元501、网络模块502、音频输出单元503、输入单元504、传感器505、显示单元506、用户输入单元507、接口单元508、存储器509、以及处理器510等中的至少部分部件。
本领域技术人员可以理解,终端500还可以包括给各个部件供电的电源(比如电池),电源可以通过电源管理系统与处理器510逻辑相连,从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。图5中示出的终端结构并不构成对终端的限定,终端可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置,在此不再赘述。
应理解的是,本申请实施例中,输入单元504可以包括图形处理器(Graphics Processing Unit,GPU)5041和麦克风5042,图形处理器5041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元506可包括显示面板5061,可以采用液晶显示器、有机发光二极管等形式来配置显示面板5061。用户输入单元507包括触控面板5071以及其他输入设备5072。触控面板5071,也称为触摸屏。触控面板5071可包括触摸检测装置和触摸控制器两个部分。其他输入设备5072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆,在此不再赘述。
本申请实施例中,射频单元501将来自网络侧设备的下行数据接收后,给处理器510处理;另外,将上行的数据发送给网络侧设备。通常,射频单元501包括但不限于天线、至少一个放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器509可用于存储软件程序或指令以及各种数据。存储器509可主要包括存储程序或指令区和存储数据区,其中,存储程序或指令区可存储操作系统、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播 放功能等)等。此外,存储器509可以包括高速随机存取存储器,还可以包括非瞬态性存储器,其中,非瞬态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 ROM,PROM)、可擦除可编程只读存储器(Erasable PROM,EPROM)、电可擦除可编程只读存储器(Electrically EPROM,EEPROM)或闪存。例如至少一个磁盘存储器件、闪存器件、或其他非瞬态性固态存储器件。
处理器510可包括一个或多个处理单元;可选的,处理器510可集成应用处理器和调制解调处理器,其中,应用处理器主要处理操作系统、用户界面和应用程序或指令等,调制解调处理器主要处理无线通信,如基带处理器。可以理解的是,上述调制解调处理器也可以不集成到处理器510中。
其中,处理器510,可以用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;射频单元501,可以用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
在本申请实施例中,终端生成第一AI模型表示方式下的AI信息并发送给第二通信设备,由于第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,这样,使用第二AI模型表示方式的第二通信设备接收第一AI模型表示方式下的AI信息,并将其加载到第二AI模型表示方式下,得到第二AI模型表示方式下的AI信息。通过本申请实施例,使用不同AI模型表示方式的通信设备可以对AI信息的理解保持一致,使得使用不同AI模型表示方式的通信设备之间可以进行AI信息的传输,有利于提高通信系统性能。
本申请实施例提供的终端500还可以实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供一种网络侧设备,包括处理器和通信接口,处理器 用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息,通信接口用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。该网络侧设备实施例是与上述网络侧设备方法实施例对应的,上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中,且能达到相同的技术效果。
具体地,本申请实施例还提供了一种网络侧设备。如图6所示,该网络侧设备600包括:天线61、射频装置62、基带装置63。天线61与射频装置62连接。在上行方向上,射频装置62通过天线61接收信息,将接收的信息发送给基带装置63进行处理。在下行方向上,基带装置63对要发送的信息进行处理,并发送给射频装置62,射频装置62对收到的信息进行处理后经过天线61发送出去。
上述频带处理装置可以位于基带装置63中,以上实施例中网络侧设备执行的方法可以在基带装置63中实现,该基带装置63包括处理器64和存储器65。
基带装置63例如可以包括至少一个基带板,该基带板上设置有多个芯片,如图6所示,其中一个芯片例如为处理器64,与存储器65连接,以调用存储器65中的程序,执行以上方法实施例中所示的网络侧设备操作。
该基带装置63还可以包括网络接口66,用于与射频装置62交互信息,该接口例如为通用公共无线接口(Common Public Radio Interface,CPRI)。
具体地,本申请实施例的网络侧设备还包括:存储在存储器65上并可在处理器64上运行的指令或程序,处理器64调用存储器65中的指令或程序执行图3所示各模块执行的方法,并达到相同的技术效果,为避免重复,故不在此赘述。
本申请实施例还提供一种可读存储介质,所述可读存储介质可以是易失性的,也可以是非易失性的,所述可读存储介质可以是瞬态性的,也可以是 非瞬态性的,所述可读存储介质上存储有程序或指令,该程序或指令被处理器执行时实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
其中,所述处理器可以为上述实施例中所述的终端中的处理器。所述可读存储介质,包括计算机可读存储介质,如计算机只读存储器(Read-Only Memory,ROM)、随机存取存储器(Random Access Memory,RAM)、磁碟或者光盘等。
本申请实施例另提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
应理解,本申请实施例提到的芯片还可以称为系统级芯片,系统芯片,芯片系统或片上系统芯片等。
本申请实施例另提供了一种计算机程序产品,所述计算机程序产品存储于非瞬态的存储介质中,所述计算机程序产品被至少一个处理器执行以实现上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例另提供了一种通信设备,被配置成用于执行上述AI信息的传输方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
需要说明的是,在本文中,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外,需要指出的是, 本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能,还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能,例如,可以按不同于所描述的次序来执行所描述的方法,并且还可以添加、省去、或组合各种步骤。另外,参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请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计算机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(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端(可以是手机,计算机,服务器,空调器,或者网络侧设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述,但是本申请并不局限于上述的具体实施方式,上述的具体实施方式仅仅是示意性的,而不是限制性的,本领域的普通技术人员在本申请的启示下,在不脱离本申请宗旨和权利要求所保护的范围情况下,还可做出很多形式,均属于本申请的保护之内。
Claims (24)
- 一种人工智能AI信息的传输方法,包括:第一通信设备生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;所述第一通信设备向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
- 根据权利要求1所述的方法,其中,所述第一通信设备使用第三AI模型表示方式,所述第一通信设备生成第一AI模型表示方式下的AI信息包括:所述第一通信设备将所述第三AI模型表示方式下的AI信息转换或保存为所述第一AI模型表示方式下的AI信息;或者,所述第一通信设备使用所述第一AI模型表示方式。
- 根据权利要求1所述的方法,其中,所述方法还包括:所述第一通信设备发送配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式。
- 根据权利要求3所述的方法,其中,所述用于传输AI信息的AI模型表示方式为所述第一AI模型表示方式,所述第一通信设备使用所述第一AI模型表示方式,所述方法还包括:所述第一通信设备接收来自于所述第二通信设备的AI信息并将接收到的AI信息加载到所述第一AI模型表示方式中;其中,所述第二通信设备还用于将所述第二AI模型表示方式下的AI信 息转换为所述第一AI模型表示方式下的AI信息。
- 根据权利要求3所述的方法,其中,所述用于传输AI信息的AI模型表示方式为所述第二AI模型表示方式,所述第一通信设备使用第一AI模型表示方式,所述方法还包括:所述第一通信设备将所述第一AI模型表示方式下的AI信息转换或保存为所述第二AI模型表示方式下的AI信息;所述第一通信设备向所述第二通信设备发送所述第二AI模型表示方式下的AI信息。
- 根据权利要求1至3任一项所述的方法,其中,所述第一通信设备生成第一AI模型表示方式下的AI信息包括:所述第一通信设备根据每个所述第二通信设备使用的AI模型表示方式,分别生成所述第一AI模型表示方式下的AI信息。
- 根据权利要求1至3任一项所述的方法,其中,所述方法还包括如下至少之一:所述第一通信设备向所述第二通信设备发送第一指示信息,所述第一指示信息用于指示所述第二通信设备使用与所述第一通信设备相同的AI模型表示方式;所述第一通信设备向所述第二通信设备发送第二指示信息,所述第二指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求1所述的方法,其中,所述方法还包括:所述第一通信设备接收来自于第三通信设备的配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式;其中,所述第三通信设备还用于向所述第二通信设备发送所述配置信息。
- 根据权利要求1所述的方法,其中,所述第一通信设备生成第一AI模型表示方式下的AI信息之前,所述方法还包括:所述第一通信设备发送第三指示信息,所述第三指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求1所述的方法,其中,所述第一通信设备生成第一AI模型表示方式下的AI信息之前,所述方法还包括:所述第一通信设备根据待发送的AI信息,从多个AI模型表示方式中确定出所述第一AI模型表示方式;其中,所述第一通信设备发送的AI信息中包括有第四指示信息,所述第四指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求10所述的方法,其中,所述第一AI模型表示方式下的AI信息的数据量,是所述多个AI模型表示方式下的多个AI信息中数据量最小的。
- 一种AI信息的传输装置,包括:生成模块,用于生成第一AI模型表示方式下的AI信息;其中,所述第一AI模型表示方式下的AI信息支持被转换为第二AI模型表示方式下的AI信息;发送模块,用于向第二通信设备发送所述第一AI模型表示方式下的AI信息;其中,所述第二通信设备使用所述第二AI模型表示方式。
- 根据权利要求12所述的装置,其中,所述装置使用第三AI模型表示方式,所述生成模块,用于将所述第三 AI模型表示方式下的AI信息转换或保存为所述第一AI模型表示方式下的AI信息;或者,所述装置使用所述第一AI模型表示方式。
- 根据权利要求12所述的装置,其中,所述发送模块,还用于发送配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式。
- 根据权利要求14所述的装置,其中,所述用于传输AI信息的AI模型表示方式为所述第一AI模型表示方式,所述装置使用所述第一AI模型表示方式,所述装置还包括接收模块,用于接收来自于所述第二通信设备的AI信息并将接收到的AI信息加载到所述第一AI模型表示方式中;其中,所述第二通信设备还用于将所述第二AI模型表示方式下的AI信息转换为所述第一AI模型表示方式下的AI信息。
- 根据权利要求14所述的装置,其中,所述用于传输AI信息的AI模型表示方式为所述第二AI模型表示方式,所述装置使用第一AI模型表示方式;所述生成模块,还用于将所述第一AI模型表示方式下的AI信息转换或保存为所述第二AI模型表示方式下的AI信息;所述发送模块,还用于向所述第二通信设备发送所述第二AI模型表示方式下的AI信息。
- 根据权利要求12至14任一项所述的装置,其中,所述生成模块,用于根据每个所述第二通信设备使用的AI模型表示方式,分别生成所述第一AI模型表示方式下的AI信息。
- 根据权利要求12至14任一项所述的装置,其中,所述发送模块, 还用于如下至少之一:向所述第二通信设备发送第一指示信息,所述第一指示信息用于指示所述第二通信设备使用与所述装置相同的AI模型表示方式;向所述第二通信设备发送第二指示信息,所述第二指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求12所述的装置,其中,所述装置还包括接收模块,用于接收来自于第三通信设备的配置信息,所述配置信息用于配置或修改用于传输AI信息的AI模型表示方式;其中,所述第三通信设备还用于向所述第二通信设备发送所述配置信息。
- 根据权利要求12所述的装置,其中,所述发送模块,还用于发送第三指示信息,所述第三指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求12所述的装置,其中,所述装置还包括确定模块,用于根据待发送的AI信息,从多个AI模型表示方式中确定出所述第一AI模型表示方式;其中,所述发送模块发送的AI信息中包括有第四指示信息,所述第四指示信息用于指示所述第一AI模型表示方式的类型。
- 根据权利要求21所述的装置,其中,所述第一AI模型表示方式下的AI信息的数据量,是所述多个AI模型表示方式下的多个AI信息中数据量最小的。
- 一种通信设备,包括处理器,存储器及存储在所述存储器上并可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如权利要求1至11任一项所述的AI信息的传输方法。
- 一种可读存储介质,所述可读存储介质上存储程序或指令,所述程序或指令被处理器执行时实现如权利要求1至11任一项所述的AI信息的传输方法。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111358543.7 | 2021-11-16 | ||
CN202111358543.7A CN116137596A (zh) | 2021-11-16 | 2021-11-16 | Ai信息的传输方法和设备 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023088268A1 true WO2023088268A1 (zh) | 2023-05-25 |
Family
ID=86333008
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/132085 WO2023088268A1 (zh) | 2021-11-16 | 2022-11-15 | Ai信息的传输方法和设备 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN116137596A (zh) |
WO (1) | WO2023088268A1 (zh) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112132219A (zh) * | 2020-09-24 | 2020-12-25 | 天津锋物科技有限公司 | 一种基于移动端的深度学习检测模型的通用部署方案 |
CN112947899A (zh) * | 2019-12-11 | 2021-06-11 | 杭州海康威视数字技术股份有限公司 | 深度学习模型转换方法、系统及装置 |
CN113599802A (zh) * | 2021-07-23 | 2021-11-05 | 安徽三七极域网络科技有限公司 | 数据处理方法、装置和系统 |
US20210349943A1 (en) * | 2018-09-20 | 2021-11-11 | Nokia Technologies Oy | An apparatus and a method for artificial intelligence |
-
2021
- 2021-11-16 CN CN202111358543.7A patent/CN116137596A/zh active Pending
-
2022
- 2022-11-15 WO PCT/CN2022/132085 patent/WO2023088268A1/zh unknown
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210349943A1 (en) * | 2018-09-20 | 2021-11-11 | Nokia Technologies Oy | An apparatus and a method for artificial intelligence |
CN112947899A (zh) * | 2019-12-11 | 2021-06-11 | 杭州海康威视数字技术股份有限公司 | 深度学习模型转换方法、系统及装置 |
CN112132219A (zh) * | 2020-09-24 | 2020-12-25 | 天津锋物科技有限公司 | 一种基于移动端的深度学习检测模型的通用部署方案 |
CN113599802A (zh) * | 2021-07-23 | 2021-11-05 | 安徽三七极域网络科技有限公司 | 数据处理方法、装置和系统 |
Also Published As
Publication number | Publication date |
---|---|
CN116137596A (zh) | 2023-05-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022083634A1 (zh) | 资源配置方法、装置、设备及可读存储介质 | |
US20240188047A1 (en) | Computing session update method and apparatus, and communication device | |
WO2023088268A1 (zh) | Ai信息的传输方法和设备 | |
WO2023143581A1 (zh) | 信息传输方法、装置、终端及网络侧设备 | |
WO2023284796A1 (zh) | Tci状态的指示方法、装置、终端和网络侧设备 | |
CN115460698A (zh) | Pdcch监听方法、终端及网络侧设备 | |
WO2023098535A1 (zh) | 信息交互方法、装置及通信设备 | |
WO2023125934A1 (zh) | Ai网络信息传输方法、装置及通信设备 | |
WO2023151649A1 (zh) | 信息激活方法、终端及网络侧设备 | |
WO2023088269A1 (zh) | Ai信息的传输方法和设备 | |
WO2024027582A1 (zh) | 模型更新方法、装置、通信设备及可读存储介质 | |
WO2023151650A1 (zh) | 信息激活方法、终端及网络侧设备 | |
WO2023165393A1 (zh) | 传输方法、装置、终端和存储介质 | |
WO2023051609A1 (zh) | 下行控制信道监测方法、装置及通信设备 | |
WO2023109828A1 (zh) | 数据收集方法及装置、第一设备、第二设备 | |
WO2024037409A1 (zh) | 定位消息传输方法、终端及网络侧设备 | |
WO2024149290A1 (zh) | 侧链路定位的消息交互方法、终端及网络侧设备 | |
WO2024032489A1 (zh) | Prs接收方法及装置、终端 | |
WO2024012279A1 (zh) | 信息传输方法、装置及设备 | |
WO2024022398A1 (zh) | 托管网络的选网信息的获取方法、终端及网络侧设备 | |
WO2023098586A1 (zh) | 信息交互方法、装置及通信设备 | |
WO2023104025A1 (zh) | Srs资源配置方法、装置、终端及网络侧设备 | |
WO2023125869A1 (zh) | Ai网络信息传输方法、装置及通信设备 | |
WO2024027618A1 (zh) | gap配置方法、装置、网络侧设备及存储介质 | |
WO2023207842A1 (zh) | 波束信息确定方法、终端及网络侧设备 |
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: 22894801 Country of ref document: EP Kind code of ref document: A1 |