CN117439958A

CN117439958A - AI network model interaction method and device and communication equipment

Info

Publication number: CN117439958A
Application number: CN202210822781.7A
Authority: CN
Inventors: 孙布勒
Original assignee: Vivo Mobile Communication Co Ltd
Current assignee: Vivo Mobile Communication Co Ltd
Priority date: 2022-07-12
Filing date: 2022-07-12
Publication date: 2024-01-23
Also published as: WO2024012303A1

Abstract

The application discloses an AI network model interaction method, an AI network model interaction device and communication equipment, which belong to the technical field of communication, and the AI network model interaction method comprises the following steps: the method comprises the steps that first equipment sends first information to second equipment, wherein the first information comprises information related to compression and/or model reasoning of an AI network model by the first equipment; the first device obtains relevant information of a target AI network model, and the target AI network model corresponds to the first information.

Description

AI network model interaction method and device and communication equipment

Technical Field

The application belongs to the technical field of communication, and particularly relates to an AI network model interaction method, an AI network model interaction device and communication equipment.

Background

In the related art, the 5 th generation (5) is promoted by means of an artificial intelligence (Artificial Intelligence, AI) network model ^th Generation, 5G) method of network performance of the communication system was studied.

The AI network model can realize the work of constructing, training and verifying the network model by means of the existing AI tool. And the AI network model obtained through training is deployed on the target equipment needing to use the AI network model by interacting with the AI network model in the wireless communication system, which involves the problem of transmission of the AI network model.

In the related art, the problems of high transmission overhead, large occupied computing resources during reasoning, high reasoning time delay and the like exist due to the fact that the size or complexity of an AI network model is large.

Disclosure of Invention

The embodiment of the application provides an AI network model interaction method, an AI network model interaction device and communication equipment, which can reduce transmission overhead and/or reduce occupied computing resources and reasoning time delay during reasoning.

In a first aspect, an artificial intelligence AI network model interaction method is provided, the method including:

the method comprises the steps that a first device sends first information to a second device, wherein the first information comprises information related to compression of an AI network model and/or model reasoning needed by the first device;

the first device obtains relevant information of a target AI network model, and the target AI network model corresponds to the first information.

In a second aspect, an artificial intelligence AI network model interaction apparatus is provided, for application to a first device, the apparatus comprising:

the first sending module is used for sending first information to the second equipment, wherein the first information comprises information related to compression of an AI network model and/or model reasoning needed by the first equipment;

and the first acquisition module is used for acquiring the related information of the target AI network model, and the target AI network model corresponds to the first information.

In a third aspect, an artificial intelligence AI network model interaction method is provided, including:

the method comprises the steps that a second device receives first information from a first device, wherein the first information comprises information related to compression of an AI network model and/or model reasoning needed by the first device;

the second device sends related information of a target AI network model to the first device, wherein the target AI network model corresponds to the first information, or the second device sends related information of a first AI network model according to the first information, and the first AI network model is used for compressing to obtain a second AI network model, and the second AI network model corresponds to the first information.

In a fourth aspect, an artificial intelligence AI network model interaction apparatus is provided for use with a second device, the apparatus comprising:

the first receiving module is used for receiving first information from first equipment, wherein the first information comprises information related to compression and/or model reasoning of an AI network model required by the first equipment;

the second sending module is configured to send related information of a target AI network model to the first device, where the target AI network model corresponds to the first information, or send related information of the first AI network model according to the first information, where the first AI network model is used to perform compression processing to obtain a second AI network model, and the second AI network model corresponds to the first information.

In a fifth aspect, there is provided a communication device comprising a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the method according to the first or third aspect.

In a sixth aspect, a communication device is provided, including a processor and a communication interface, where the communication interface is configured to send first information to a second device, where the first information includes information related to compression and/or model reasoning of an AI network model required by the first device; the communication interface or the processor is used for acquiring related information of a target AI network model, and the target AI network model corresponds to the first information; or,

the communication interface is configured to receive first information from a first device, and send related information of a target AI network model to the first device or send related information of the first AI network model according to the first information, where the first information includes information related to compression and/or model reasoning of the AI network model required by the first device, the target AI network model corresponds to the first information, and the first AI network model is configured to perform compression processing to obtain a second AI network model, and the second AI network model corresponds to the first information.

In a seventh aspect, a communication system is provided, comprising: a first device operable to perform the steps of the AI network model interaction method of the first aspect, and a second device operable to perform the steps of the AI network model interaction method of the third aspect.

In an eighth aspect, there is provided a readable storage medium having stored thereon a program or instructions which when executed by a processor, performs the steps of the method according to the first aspect, or performs the steps of the method according to the third aspect.

In a ninth aspect, there is provided a chip comprising a processor and a communication interface, the communication interface and the processor being coupled, the processor being for running a program or instructions to implement the method according to the first aspect or to implement the method according to the third aspect.

In a tenth aspect, there is provided a computer program/program product stored in a storage medium, the computer program/program product being executed by at least one processor to implement the steps of the AI network model interaction method as set forth in the first aspect, or the computer program/program product being executed by at least one processor to implement the steps of the AI network model interaction method as set forth in the third aspect.

In the embodiment of the application, a first device sends first information to a second device, wherein the first information comprises information related to compression and/or model reasoning of an AI network model required by the first device; the first device obtains relevant information of a target AI network model, and the target AI network model corresponds to the first information. Thus, in the case that the second device stores or trains to obtain the AI network model in advance, the first device may send, to the second device, information related to compression and/or model reasoning of the AI network model required by the first device in the process of obtaining the AI network model from the second device, so that the second device can determine at least one of the following according to the requirement of the first device: the type, the size, the function and the complexity of the AI network model required by the first equipment, and parameters, a compression method, compression nodes and the like when the determined AI network model is subjected to compression processing, so that the second equipment can compress the AI network model according to the requirement of the first equipment, and transmit the compressed AI network model, and the transmission cost of the AI network model can be reduced; in addition, the second device can select an AI network model matched with the model reasoning process of the first device according to the requirement of the first device, so that the calculation resources and the reasoning time delay occupied by the first device when reasoning the target AI network model can be reduced.

Drawings

Fig. 1 is a schematic structural diagram of a wireless communication system to which embodiments of the present application can be applied;

FIG. 2 is a flowchart of an AI network model interaction method provided in an embodiment of the present application;

fig. 3 is a schematic diagram of an embodiment of the present application applied to CSI feedback;

FIG. 4 is a schematic diagram of an interaction process between a first device and a second device in an embodiment of the present application;

FIG. 5 is a second schematic diagram of an interaction process between a first device and a second device according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an interaction process among a first device, a second device, and a third device in an embodiment of the present application;

FIG. 7 is a flowchart of another AI network model interaction method provided by an embodiment of the application;

fig. 8 is a schematic structural diagram of an AI network model interaction device according to an embodiment of the present application;

FIG. 9 is a schematic structural diagram of another AI network model interaction device provided in an embodiment of the disclosure;

fig. 10 is a schematic structural diagram of a communication device according to an embodiment of the present application.

Detailed Description

Technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application are within the scope of the protection of the present application.

The terms first, second and the like in the description and in the claims, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. 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 otherwise described herein, and that the terms "first" and "second" are generally intended to be used in a generic sense and not to limit the number of objects, for example, the first object may be one or more. Furthermore, in the description and claims, "and/or" means at least one of the connected objects, and the character "/" generally means a relationship in which the associated object is an "or" before and after.

It is noted that the techniques described in embodiments of the present application are not limited to long term evolution (Long Term Evolution, LTE)/LTE evolution (LTE-Advanced, LTE-a) systems, but may also be used in other wireless communication systems, such as code division multiple access (Code Division Multiple Access, CDMA), time division multiple access (Time Division Multiple Access, TDMA), frequency division multiple access (Frequency Division Multiple Access, FDMA), orthogonal frequency division multiple access (Orthogonal Frequency Division Multiple Access, OFDMA), single carrier frequency division multiple access (Single-carrier Frequency Division Multiple Access, SC-FDMA), and other systems. The terms "system" and "network" in embodiments of the present application are often used interchangeably, and the techniques described may be used for both the above-mentioned systems and radio technologies, as well as other systems and radio technologies. The following description describes a New air interface (NR) system for purposes of example and uses NR terminology in much of the description that follows, but these techniques are also applicable to applications other than NR system applications, such as generation 6 (6) ^th Generation, 6G) communication system.

Fig. 1 shows a block diagram of a wireless communication system to which embodiments of the present application are applicable. The wireless communication system includes a terminal 11 and a network device 12. The terminal 11 may be a mobile phone, a tablet (Tablet Personal Computer), a Laptop (Laptop Computer) or a terminal-side Device called a notebook, a personal digital assistant (Personal Digital Assistant, PDA), a palm top, a netbook, an ultra-mobile personal Computer (ultra-mobile personal Computer, UMPC), a mobile internet appliance (Mobile Internet Device, MID), an augmented reality (augmented reality, AR)/Virtual Reality (VR) Device, a robot, a Wearable Device (weather Device), a vehicle-mounted Device (VUE), a pedestrian terminal (PUE), a smart home (home Device with a wireless communication function, such as a refrigerator, a television, a washing machine, or a furniture), a game machine, a personal Computer (personal Computer, PC), a teller machine, or a self-service machine, and the Wearable Device includes: intelligent wrist-watch, intelligent bracelet, intelligent earphone, intelligent glasses, intelligent ornament (intelligent bracelet, intelligent ring, intelligent necklace, intelligent anklet, intelligent foot chain etc.), intelligent wrist strap, intelligent clothing etc.. Note that, the specific type of the terminal 11 is not limited in the embodiment of the present application. The network-side device 12 may comprise an access network device or a core network device, wherein the access network device 12 may also be referred to as a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function or a radio access network element. Access network device 12 may include a base station, a WLAN access point, a WiFi node, or the like, which may be referred to as a node B, an evolved node B (eNB), an access point, a base transceiver station (Base Transceiver Station, BTS), a radio base station, a radio transceiver, a basic service set (Basic Service Set, BSS), an extended service set (Extended Service Set, ESS), a home node B, a home evolved node B, a transmission and reception point (Transmitting Receiving Point, TRP), or some other suitable terminology in the art, and the base station is not limited to a particular technical vocabulary so long as the same technical effect is achieved, and it should be noted that in the embodiments of the present application, only a base station in an NR system is described as an example, and the specific type of the base station is not limited.

Artificial intelligence is currently in wide-spread use in various fields. There are various implementations of AI network models, such as neural networks, decision trees, support vector machines, bayesian classifiers, etc. The present application is described by way of example with respect to neural networks, but is not limited to a particular type of AI network model.

In general, the AI algorithm chosen and the network model employed will also vary according to the different types of problems that need to be solved. The main method for improving the performance of the 5G network by means of the AI network model is to enhance or replace the existing algorithm or processing module by using the algorithm and model based on the neural network. In certain scenarios, neural network-based algorithms and models may achieve better performance than deterministic-based algorithms. More common neural networks include deep neural networks, convolutional neural networks, recurrent neural networks, and the like. By means of the existing AI tool, the construction, training and verification work of the neural network can be realized.

In applications, the size and complexity of AI network models is a key issue for their deployment applications. The transmission problem of the AI network model is also involved when applying the AI scheme in a wireless communication system, and is also affected by the size and complexity of the AI network model. The large AI network model has large transmission cost, occupies more calculation resources during reasoning, and has high reasoning time delay.

In the embodiment of the present application, before acquiring an AI network model, a first device sends demand information to a second device to notify the second device of the size, compression scheme, model complexity, etc. of the AI network model required by the first device, so that the first device can acquire an AI network model that is more matched with the demand information, and in the case that the first device receives the compressed AI network model, resource overhead for transmitting the AI network model can be reduced; under the condition that the first equipment acquires the AI network model matched with the required model complexity, the computing resource and time delay when the first equipment performs model reasoning on the AI network model can be reduced.

The AI network model interaction method, the AI network model interaction device, the communication equipment and the like provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings through some embodiments and application scenarios thereof.

Referring to fig. 2, an execution subject of the AI network model interaction method provided in the embodiment of the present application is a first device, and as shown in fig. 2, the AI network model interaction method executed by the first device may include the following steps:

step 201, a first device sends first information to a second device, where the first information includes information related to compression and/or model reasoning of an AI network model required by the first device.

Wherein the first device may be a demander of the AI network model and the second device may be a sender of the AI network model, for example: the second device trains to obtain an AI network model, and sends the trained related information of the AI network model to the first device, wherein the related information of the AI network model can be parameters or model files of the AI network model and the like, and the data of the AI network model reasoning (namely, the AI network model is applied) can be provided for the first device. It is to be understood that "transmission AI network model" in the following embodiments may be interpreted as "parameters or model files of transmission AI network model". In implementation, the first device may be a terminal, for example: the various types of terminals 11 as listed in fig. 1, or the first device may also be a network-side device, for example: the network side device 12 or the core network device as illustrated in the embodiment shown in fig. 1, the second device may also be a terminal or a network side device, for example: an access network device or a core network device. For convenience of explanation, in the following embodiments, the first device is generally a terminal, and the second device is a base station, which is exemplified and not specifically limited herein.

It should be noted that in the embodiment of the present application, the information interaction between the first device, the second device, and the third device may use new signaling or information, or multiplex existing signaling or information in the related technology.

Specifically, the first device, the second device, and the third device may be a terminal or a network side device, and the related signaling or information in the related technology may be multiplexed based on information interaction among the first device, the second device, and the third device, where the related signaling sending end and the related signaling receiving end are the terminal or the network side device, and the related signaling or information is divided into the following 4 cases:

case 1) assume that in the information interaction process between the first device and the second device or the third device, the information sending end is a terminal, and the information receiving end is a network side device, then information in the interaction process (for example: at least one of the first information, the matching result, the first request information, information related to the AI network model received by the first device, and the third information) may be carried in at least one of the following signaling or information:

layer (layer) 1 signaling of the physical uplink control channel (Physical Uplink Control Channel, PUCCH);

MSG 1 of physical random access channel (Physical Random Access Channel, PRACH);

MSG 3 of PRACH;

MSG A of PRACH;

information of a physical uplink shared channel (Physical Uplink Shared Channel, PUSCH).

Case 2) assume that in the information interaction process between the first device and the second device or the third device, the information sending end is a network side device, and the information receiving end is a terminal, then information in the interaction process (for example: at least one of the first information, the matching result, the first request information, information related to the AI network model received by the first device, and the third information) may be carried in at least one of the following signaling or information:

a medium access control element (Medium Access Control Control Element, MAC CE);

a radio resource control (Radio Resource Control, RRC) message;

Non-Access Stratum (NAS) messages;

managing the orchestration message;

user plane data;

downlink control information (Downlink Control Information, DCI) information;

system information blocks (System Information Block, SIB);

layer 1 signaling of the physical downlink control channel (Physical Downlink Control Channel, PDCCH);

Information of a physical downlink shared channel (Physical Downlink Shared Channel, PDSCH);

MSG 2 of PRACH;

MSG 4 of PRACH;

MSG B of PRACH.

Case 3) assume that in the information interaction process between the first device and the second device or the third device, the information sending end and the information receiving end are different terminals, and then the information in the interaction process (for example: at least one of the first information, the matching result, the first request information, information related to the AI network model received by the first device, and the third information) may be carried in at least one of the following signaling or information:

xn interface signaling;

PC5 interface signaling;

information of the physical bypass control channel (Pysical Sidelink Control Channel, PSCCH);

information of the physical bypass shared channel (Physical Sidelink Shared Channel, PSSCH);

information of the physical through link broadcast channel (Physical Sidelink Broadcast Channel, PSBCH);

information of a physical through link discovery channel (Physical Sidelink Discovery Channel, PSDCH);

information of the physical bypass feedback channel (Physical Sidelink Feedback Channel, PSFCH).

Case 4) assuming that in the information interaction process between the first device and the second device or the third device, the information sending end and the information receiving end are different network side devices, the information in the interaction process (for example: at least one of the first information, the matching result, the first request information, information related to the AI network model received by the first device, and the third information) may be carried in at least one of the following signaling or information:

S1 interface signaling;

xn interface signaling (e.g., X2 interface signaling).

Step 202, the first device obtains related information of a target AI network model, where the target AI network model corresponds to the first information.

The acquiring the target AI network model in step 202 may be receiving the target AI network model, for example: receiving a target AI network model from a second device or a third device; alternatively, the acquiring the target AI network model in step 202 may be further processing to obtain the target AI network model according to feedback information of the second device on the first signal, for example: and the second equipment sends the first AI network model to the first equipment according to the first information, and the first equipment compresses the first AI network model to obtain the target AI network model.

In application, the first device may perform model reasoning on the AI network model after acquiring the related information of the target AI network model, i.e. apply the target AI network model, for example: the target AI network model is used to replace functional modules in existing communication systems. In implementation, the AI network model is used for replacing a module in the existing system, so that the system performance can be effectively improved.

For example: in the channel state information (Channel State Information Reference Signal, CSI) feedback process, as shown in fig. 3, an AI encoder (encoder) and an AI decoder (decoder) may be used instead of conventional CSI calculation, so that the corresponding system performance may be greatly improved under the same overhead condition. As shown in table 1 below, the AI scheme using AI encoder and AI encoder instead of conventional CSI computation can improve the spectral efficiency of the communication system by about 30% compared to the NR designation scheme (NR specified solution) in the related art.

TABLE 1

Scheme for the production of a semiconductor device	Spectral efficiency (bit/s/Hz)
		NR designation scheme (NR specified solution)	6.41
AI	8.28

The simulation conditions of table 1 above were: system level simulation (System Level Simulation, SLS), procedure size (Urban Micro, UMi) 38.901,7 cells (cells), 3sectors (3 sectors for each cell) per cell, UE speed (UE speed) 3km/h, carrier frequency (carrier frequency) 3.5ghz,32 gNB antennas (antenna) ([ Mg Ng M N P ] = [1 1 2 8 2 ]), 4 UE antenna ([ Mg Ng M N P ] = [1 1 2 2 2 ]), 52 Radio Bearers (RBs), precoding matrix indication (Precoding Matrix Indicator, PMI) overhead 58bits (the overhead of PMI is bits). Wherein Mg in [ Mg Ng M N P ] represents the number of antenna panels included in one column in the antenna panel array; ng represents the number of antenna panels contained in a row of the antenna panel array; m represents the number of antennas in a row on a panel; n represents the number of antennas in a row on a panel; p represents the number of planned directions of the antenna.

The first information includes information related to compression and/or model reasoning of an AI network model required by the first device, and may be at least one of the following:

the first information comprises the requirement information of the first device, so that the first device informs the second device of the function, the type, the size, the complexity degree of the AI network model required by the first device, whether the AI network model is compressed or not, and a compression method and the like are adopted when the AI network model is compressed, so that the second device provides a target AI network model which meets the requirement of the first device;

the first information includes capability information of the first device such that the first device informs the second device of its own capability information (e.g., available computing power supported by the first device, available computing power for performing model compression, model compression capabilities, model compression methods, etc.) to cause the second device to provide the first device with its supported target AI network model.

As an alternative embodiment, the first device obtains the relevant information of the target AI network model, including:

the first device receives related information of a target AI network model, wherein the target AI network model is a compressed AI network model or an uncompressed AI network model; or,

The first device receives related information of a first AI network model from the second device, compresses the first AI network model to obtain a second AI network model, and the target AI network model comprises the second AI network model.

In one embodiment, in a case that the second device has an AI network model matching the first information, the first device acquires related information of a target AI network model, including:

the first device receives information about an AI network model matching the first information from the second device, wherein the target AI network model includes the AI network model matching the first information.

The AI network model matched with the first information may be an AI network model that satisfies the capability of the first device, for example: the resources (such as electric quantity, computing resources, storage resources and the like) occupied by the model reasoning on the AI network model are smaller than or equal to the resources available to the first device, or the model complexity when the model reasoning on the AI network model is smaller than or equal to the maximum complexity which can be supported by the first device.

For example: as shown in fig. 4, 5 and 6, assuming that the first information includes parameters related to model compression, the second device may determine, according to the first information, a first model size of an AI network model required by the first device, and if an AI network model meeting the first model size requirement is found in the first model library, the target AI network model meeting the first model size requirement may be sent to the first device.

In this embodiment, in the case where the second device has the AI network model matching the first information, the first device may receive the AI network model matching the first information from the second device. In this way, the first device obtains the AI network model matched with the first information, so that the computational effort and the inference delay when the first device applies the AI network model can be reduced.

In another embodiment, when the second device does not have an AI network model matching the first information, such as the AI network model of the second device is larger than the AI network model indicated in the first information, the AI network model matching the first information may be obtained by compressing the AI network model of the second device. At this time, the device compressing the AI network model may be the first device, the second device, or the third device.

1) When the device that compresses the AI network model is a second device, the second device compresses the AI network model that it has according to the first information, and transmits the compressed AI network model to the first device.

For example: as shown in fig. 4, assuming that the first information includes parameters related to model compression, the second device may determine, according to the first information, a first model size of an AI network model required by the first device, and if an AI network model meeting the first model size requirement is not found in the first model library, perform compression processing on the first AI network model in the first model library, so as to obtain a second AI network model meeting the first model size requirement, and then send the second AI network model to the first device.

Therefore, resources occupied when the AI network model is transmitted between the second equipment and the first equipment can be reduced, the AI network model acquired by the first equipment is matched with the first information, and the computational effort and the reasoning time delay when the first equipment applies the AI network model can be reduced.

2) When the device for compressing the AI network model is a third device, the second device selects a first AI network model of the second device according to the first information, determines model compression information according to a difference between the first AI network model and the first information, and sends the first AI network model and the model compression information to the third device, so that the third device compresses the first AI network model according to the model compression information to obtain a compressed AI network model corresponding to the first information, and sends the compressed AI network model to the first device. In this way, the resources occupied by the first equipment when receiving the AI network model can be reduced, the AI network model acquired by the first equipment is matched with the first information, and the computational effort and the reasoning time delay when the first equipment applies the AI network model can be reduced;

For example: as shown in fig. 5, assuming that the first information includes parameters related to model compression, the second device may determine, according to the first information, a first model size of an AI network model required by the first device, and if an AI network model meeting the first model size requirement is not found in the first model library, the second device may send, to the third device, related information and third information of the first AI network model in the first model library, so that the third device performs compression processing on the received first AI network model according to the third information, to obtain a second AI network model, and send, by the third device, the second AI network model to the first device.

3) When the device for compressing the AI network model is a first device, the second device selects a first AI network model of the second device according to the first information, determines model compression information according to the difference between the first AI network model and the first information, and sends the first AI network model and the model compression information to the first device, so that the first device compresses the first AI network model according to the model compression information to obtain a compressed AI network model corresponding to the first information. In this way, although the resources occupied by the first device when receiving the AI network model are not reduced, the AI network model acquired by the first device is matched with the first information, so that the computational effort and the inference delay when the first device applies the AI network model can be reduced.

For example: as shown in fig. 6, assuming that the first information includes parameters related to model compression, the second device may determine, according to the first information, a first model size of an AI network model required by the first device, and if an AI network model meeting the first model size requirement is not found in the first model library, the second device may send, to the first device, related information and third information of the first AI network model in the first model library, so that the first device performs compression processing on the received first AI network model according to the third information to obtain a second AI network model.

As an alternative embodiment, the first information includes at least one of:

first capability information indicating compression capability of the first device on an AI network model and/or an AI network model compression method supported by the first device;

first requirement information indicating size information of an AI network model required by the first device;

first application information indicating function information of an AI network model required by the first device;

second information including information related to resource usage of the first device;

And first indication information indicating a device compressing the AI network model.

In one option, the first capability information may reflect a compression capability of the first device on the AI network model and/or an AI network model compression method supported by the first device, where the model compression capability or model compression method may include at least one of: knowledge distillation, pruning, low rank decomposition, tensor decomposition, and the like, are not exhaustive herein. In implementations, the first capability may be a field, such as: if the first capability is 0000, the first device supports the knowledge distillation method, if the first capability is 0001, the first device supports the pruning method, if the first capability is 0020, the first device supports the low rank decomposition method, and if the first capability is 0011, the first device supports the tensor decomposition method. The second device, after learning the compression capability of the first device and/or the AI network model compression method supported by the first device, may decide whether to compress the AI network model by the first device.

Option two, the first requirement information may reflect size information of the AI network model required by the first device (i.e., a requirement of the first device for a model size of the target AI network model), where the size information may include at least one of: an upper limit of a model size of the target AI network model, a lower limit of a model size of the target AI network model, a model size level of the target AI network model, an upper limit of a parameter amount of the target AI network model, a lower limit of a parameter amount of the target AI network model, a parameter amount level of the target AI network model, an upper limit of complexity of the target AI network model, a lower limit of complexity of the target AI network model or a complexity level of the target AI network model, an upper limit of performance of the target AI network model, a lower limit of performance of the target AI network model, a performance level of the target AI network model.

In the implementation, when the second device knows the first requirement information of the first device, the second device may select, from among AI network models provided in the second device, a target AI network model matching the first requirement information, for example: the target AI network model selected from the AI network models provided by the first device is less than or equal to the upper limit of the model size required by the first device.

Alternatively, when the second device does not have the target AI network model matching the first request information, the second device may perform compression processing on the AI network model provided to the second device based on the first information, to obtain the target AI network model matching the first request information, for example: assuming that the size of the AI network model possessed by the second device is larger than the upper limit of the model size of the AI network model required by the first device, it may be decided how to compress the AI network model possessed by the second device such that the compressed AI network model is smaller than or equal to the upper limit of the model size of the AI network model required by the first device, based on a difference between the AI network model possessed by the second device and the model size of the AI network model required by the first device.

Option three, the first application information may reflect function information of the AI network model required by the first device, for example: and the AI network model is used for realizing at least one of the functions of CSI prediction, CSI compression, beam prediction, flow prediction and the like.

In implementation, when knowing the function of the AI network model required by the first device, the second device may select, from the AI network models provided by itself, a target AI network model that matches the first application information, for example: the function of the target AI network model selected from among the AI network models possessed by the second device corresponds to the function information of the first application information, and then the second device may directly or indirectly provide the AI network model capable of realizing the function to the first device.

In option four, the second information may reflect a resource usage rate of the first device, where the resource usage rate may include: power usage, storage resource usage, computing resource usage, transmission resource usage, and the like. For example: the second information may include at least one of: the available computing power of the first device, the available computing power duty cycle of the first device, the available computing power level of the first device, the available power duty cycle of the first device, the available power level of the first device, the available storage duty cycle of the first device, the available storage level of the first device.

In implementation, when knowing the resource usage rate of the first device, the second device may select, from AI network models provided in itself, a target AI network model matching the resource usage rate of the first device, for example: the resource occupation amount of the selected target AI network model is smaller than or equal to the available resource of the first device or the resource which can be used for carrying out AI network model compression or reasoning, and then the second device can directly or indirectly provide the selected AI network model for the first device, so that the risk that the resource occupation amount of the AI network model is larger than the actual available resource of the first device can be reduced, the resource utilization rate of the first device in the model reasoning process is improved, and the time delay of the model reasoning process is reduced.

In the fifth aspect, the first instruction information indicates a device that compresses the AI network model, and may be configured to designate one or more devices to perform the model compression process when the second device does not have the AI network model with the model size required by the first device, and thus the AI network model that the second device has is required to perform the compression process, where the designated device may include at least one of the first device, the second device, and the third device may be any device other than the first device and the second device, for example: the function of the device may be an entity (such as a terminal, a base station, a core network device, or other network layer entity) in the communication network, or may be a third party device outside the communication network, where the function at least includes a model compression function.

In implementation, the second device may send the AI network model to be compressed and the model compression information to the device when knowing the device that compresses the AI network model, so that the AI network model is compressed by the device, and an AI network model with a model size that meets the requirement of the first device is obtained.

It should be noted that, the first information may include one or at least two of the above options one to five, for example: the second device may first select at least one AI network model matching the first application information from among AI network models provided in the second device, if it is determined that all of the at least one AI network model do not match a model size corresponding to the first requirement information, and the first capability information indicates that the first device supports the knowledge distillation method, the second device may select one of the at least one AI network model having a closest model size corresponding to the first requirement information, and transmit third information to the first device, where the third information may include parameter information, such as a compression method, a compression class, and the like, required when the AI network model transmitted to the first device is compressed to match the model size corresponding to the first requirement information.

In a first optional embodiment, in a case that the second device does not have an AI network model matching the first information, and the first indication information indicates that the first device compresses the AI network model, the first device acquires relevant information of a target AI network model, including:

the first device receives related information and third information of a first AI network model from the second device, wherein the third information is used for compressing the first AI network model into a second AI network model;

and the first equipment compresses the first AI network model according to the third information to obtain the second AI network model, wherein the target AI network model comprises the second AI network model.

The second device does not have an AI network model matched with the first information, and may be that the AI network model of the second device that satisfies the first portion of the first information does not satisfy the first requirement information, where the first portion of the first information may include information other than the first requirement information in the first information, such as: at least one of the first capability information, the first application information, and the second information.

The third information may include information required to compress the first AI network model into a second AI network model corresponding to the first information.

Optionally, the third information includes at least one of: and carrying out compression processing on the first AI network model, wherein the compression processing is carried out on the AI network model by adopting an AI network model compression method and the related limit information of AI network model compression.

Wherein the compression method may include at least one of: knowledge distillation, pruning, low rank decomposition, tensor decomposition, etc.

The AI network model compression-related constraint information may include at least one of: maximum compressible limit (e.g., compression ratio, compression level), upper limit of the number of parameters after compression, lower limit of the number of parameters after compression, upper limit of the model size after compression, lower limit of the model size after compression, etc.

In implementations, the third information may be determined from a difference between the AI network model required by the first information and the first AI network model.

Of course, the third information may comprise at least part of the first information, for example: the first requirement information and the like, so that the device that receives the third information can determine what compression processing is performed on the first AI network model based on the difference between the AI network model corresponding to the third information and the first AI network model.

In this embodiment, the second device selects the first AI network model that has been trained or stored based on the first information (for example, selects one closest to the first information as the first AI network model, or selects one whose model size does not match the model size required by the first information as the first AI network model), determines third information based on the difference between the first AI network model and the model size required in the first information, and information such as model compression capability supported by the first device, and transmits the third information and the first AI network model to the first device, so that the target AI network model can be obtained by the first device performing compression processing on the first AI network model based on the third information.

In a second optional implementation manner, in a case that the second device does not have an AI network model that matches the first information, and the first indication information indicates that the second device compresses the AI network model, the first device obtains relevant information of the target AI network model, including:

the first device receives related information of a second AI network model from the second device, wherein the target AI network model comprises the second AI network model, and the second AI network model is an AI network model obtained by compressing a first AI network model of the second device according to the first information.

The meaning of the AI network model that the second device does not have to match with the first information is the same as that of the first optional embodiment, and will not be described herein.

In this embodiment, the second device selects the trained or stored first AI network model according to the first information, compresses the first AI network model according to the difference between the first AI network model and the model size required in the first information, and the model compression capability supported by the second device to obtain a second AI network model, and then forwards the second AI network model to the first device.

In a third optional embodiment, in a case where the second device does not have an AI network model matching the first information, and the first indication information indicates that the third device compresses the AI network model, the first device obtains information about a target AI network model, including:

the first device receives related information of a second AI network model from the third device, wherein the target AI network model comprises the second AI network model, and the second AI network model is an AI network model obtained by compressing the first AI network model from the second device.

The second device does not have the meaning of the AI network model matched with the first information, and the meaning and the function of the third information are the same as those of the first optional embodiment, which are not described herein.

In this embodiment, the second device selects the first AI network model that has been trained or stored according to the first information, determines third information according to a difference between the first AI network model and a model size required in the first information, and information such as a model compression capability supported by the third device, and sends the third information and the first AI network model to the third device, so that the target AI network model may be obtained by the third device by compressing the first AI network model according to the third information and forwarding the first AI network model to the first device.

In the case where the third device performs compression processing on the first AI network model based on the third information to obtain the second AI network model, the first device may further transmit the first information, or a portion of the first information related to model compression, to the third device, so that the third device may determine what compression processing is performed on the first AI network model based on the first information or the portion of the first information related to model compression.

As an optional implementation manner, after the first device performs compression processing on the first AI network model according to the third information to obtain the second AI network model, the method further includes:

the first device sends information related to the second AI network model to the second device.

In this embodiment, after performing compression processing on the first AI network model according to the third information to obtain the second AI network model, the first device sends the second AI network model to the second device, so that in a subsequent model transmission process, the AI network model included in the second AI network model includes the second AI network model, and thus the second AI network model can be directly transmitted without performing compression processing on the first AI network model again.

Similarly, in the case where the second AI network model is obtained by compressing the first AI network model by the third device based on the third information, the second device may acquire the second AI network model from the third device.

In implementation, as shown in fig. 5 and fig. 6, for the second AI network model obtained by compressing the first AI network model according to the third information, the first device may further determine whether the second AI network model meets the requirement of the first information, for example: the second AI network model satisfies the model-size requirement in the first information, such that the first device may obtain information about the second AI network model from at least one of the first device and the third device if the first device determines that the second AI network model does satisfy the requirement for the first information.

As an alternative embodiment, in a case that the first device acquires a second AI network model compressed by the first device or a third device, the method further includes:

the first device obtains a matching result of the second AI network model and the first information;

the first device sends the matching result to the second device.

In this embodiment, when the target AI network model is obtained by compression processing by the first device or the third device, the first device further obtains a matching result between the target AI network model and the first information, for example: whether the target AI network model is consistent with the model size required by the first device or not, and feeding back the matching result to the second device. Thus, if the matching result indicates that the target AI network model does not match the first information, any of the following processes may be performed:

1) The first device changes the first information and re-requests the AI network model from the second device. The process is similar to the process of the AI network model interaction method provided in the embodiment of the present application, and will not be described herein.

2) The first device sends first request information, at this time, the first information is unchanged, and the first information may not be sent to the second device, and the second device may perform compression processing by using different third information according to the first request information, or perform compression processing on different first AI network models. For example: the first request information may carry advice information not to compress the first AI network model that was previously compressed and/or advice information to change the third information by the second device. Alternatively, the first request information may not carry the suggestion, but the second device decides which of the first AI network models to recompress, and whether to modify the third information.

3) The AI network model request of the first device is aborted.

In an alternative embodiment, in a case that the matching result indicates that the second AI network model does not match the first information, the AI network model interaction method further includes:

the first device sends first request information to the second device, the first request information being used to request the second device to update at least one of the third information and the first AI network model.

In this embodiment, when the first device acquires the AI network model of the model size that does not meet the requirement, the first device may send first request information to the second device, so that the second device updates at least one of the following according to the first request information: the compressed first AI network model and third information used in the compression process until the first device obtains an AI network model of a desired model size.

In another alternative embodiment, in a case that the matching result indicates that the second AI network model does not match the first information, the AI network model interaction method further includes:

the first device updates the first information and sends the updated first information to the second device;

And the first equipment acquires a target AI network model corresponding to the updated first information.

In this embodiment, when the first device acquires the AI network model of the undesirable model size, the first device may update at least one item of first information, for example: at least one of the first indication information, the first requirement information, the first capability information and the second information is updated, so that the second device updates at least one of the first AI network model and the third information according to the updated first information, and the target AI network model matched with the updated first information can be obtained by compressing the first AI network model according to the updated third information.

It should be noted that, as shown in fig. 5, when the third device performs compression processing on the first AI network model according to the third information to obtain the second AI network model, the first device may further send the matching result to the third device, so that the third device sends the compressed second AI network model to the second device according to the situation that the second AI network model is determined to match with the first information; and under the condition that the second AI network model is not matched with the first information, the compressed second AI network model is not sent to the second equipment, so that the resource waste caused by the transmission of the second AI network model when the second AI network model is not matched with the first information can be reduced.

In addition, as shown in fig. 5, when the third device compresses the first AI network model according to the third information to obtain the second AI network model, the first device may determine whether to transmit the second AI network model to the second device according to the matching result, and in this case, the first device may not transmit the matching result to the third device, and the third device may not need to determine whether to transmit the second AI network model to the second device according to the matching result.

In the embodiment of the application, a first device sends first information to a second device, wherein the first information comprises information related to compression and/or model reasoning of an AI network model required by the first device; the first device obtains relevant information of a target AI network model, and the target AI network model corresponds to the first information. Thus, in the case that the second device stores or trains to obtain the AI network model in advance, the first device may send, to the second device, information related to compression and/or model reasoning of the AI network model required by the first device in the process of obtaining the AI network model from the second device, so that the second device can determine at least one of the following according to the requirement of the first device: the type, the size, the function and the complexity of the AI network model required by the first equipment, and parameters, a compression method, compression nodes and the like when the determined AI network model is subjected to compression processing, so that the second equipment can compress the AI network model according to the requirement of the first equipment, and transmit the compressed AI network model, and the transmission cost of the AI network model can be reduced; in addition, the second device also selects an AI network model matched with the model reasoning process of the first device according to the requirement of the first device, so that the calculation resources and the reasoning time delay occupied by the first device when reasoning the target AI network model can be reduced.

Referring to fig. 7, another AI network model interaction method provided in the embodiment of the present application, an execution subject of which is a second device, and as shown in fig. 7, the AI network model interaction method executed by the second device may include the following steps:

step 701, the second device receives first information from the first device, where the first information includes information related to compression of an AI network model and/or model reasoning needed by the first device.

Step 702, the second device sends related information of a target AI network model to the first device, where the target AI network model corresponds to the first information, or the second device sends related information of a first AI network model according to the first information, where the first AI network model is used for performing compression processing to obtain a second AI network model, and the second AI network model corresponds to the first information.

In this embodiment of the present application, the first device, the second device, the first information, the related information of the first AI network model, and the meaning and the function of the second AI network model are the same as those of the first device, the second device, the first information, the related information of the first AI network model, and the meaning and the function of the second AI network model in the embodiment of the method shown in fig. 2, which are not described herein.

Optionally, the target AI network model is a compressed AI network model or an uncompressed AI network model.

Optionally, the first information includes at least one of:

Optionally, the first indication information indicates the first device, the second device, or the third device to compress the AI network model.

Optionally, the second device sends related information of the target AI network model, including:

the second device, with the AI network model matching the first information, sends the AI network model matching the first information to the first device, the target AI network model including the AI network model matching the first information.

Optionally, in a case that the second device does not have an AI network model that matches the first information, and the first indication information indicates that the second device compresses the AI network model, the method further includes:

the second equipment compresses the first AI network model according to the first information to obtain a second AI network model;

the second device sending information about the target AI network model, including:

the second device sends information about the second AI network model to the first device, the target AI network model including the second AI network model.

Optionally, the second device sends relevant parameters of the first AI network model according to the first information, including:

the second device sends related information and third information of a first AI network model to the first device under the condition that the first device does not have the AI network model matched with the first information and the first indication information indicates the first device to compress the AI network model, wherein the third information is used for compressing the first AI network model into a second AI network model, and the second AI network model corresponds to the first information; and/or the number of the groups of groups,

The second device sends related information of the first AI network model and the third information to the third device without the AI network model matching the first information, and the first indication information indicates the third device to compress the AI network model.

Optionally, the third information includes at least one of: and the AI network model compression method used when the first AI network model is compressed and the AI network model compression related limit information.

Optionally, after the second device sends the related information of the first AI network model and the third information to the first device or the third device, the method further includes:

the second device receives information regarding the second AI network model.

the second device receives a judgment result from the first device, where the judgment result is used to represent a matching result of the second AI network model and the first information.

Optionally, in a case where the matching result indicates that the second AI network model does not match the first information, the method further includes:

the second device receives first request information from the first device and updates at least one of the third information and the first AI network model according to the first request information;

and the second device sends the updated third information and/or the updated related information of the first AI network model.

the second device receives updated first information from the first device;

and the second device sends related information of a target AI network model matched with the updated first information to the first device, or the second device sends related information of a third AI network model according to the updated first information, and the third AI network model is used for compressing to obtain a fourth AI network model corresponding to the updated first information.

The third AI network model is similar to the first AI network model in the method embodiment shown in fig. 2, and may be an AI network model in a model base of the second device, and is different from the first AI network model in the method embodiment shown in fig. 2: the first AI network model corresponds to the first information before updating and the third AI network model corresponds to the first information after updating.

The fourth AI network model is similar to the second AI network model in the method embodiment shown in fig. 2, and may be an AI network model obtained by compressing the AI network model in the model base of the second device, where: the second AI network model is an AI network model which is obtained by compressing the first AI network model and corresponds to the first information before updating, and the fourth AI network model is an AI network model which is obtained by compressing the third AI network model and corresponds to the first information after updating.

The AI network model interaction method executed by the second device corresponds to the AI network model interaction method executed by the first device, and the first device and the second device execute the steps in the respective AI network model interaction methods respectively, so that the transmission cost of the AI network model can be reduced, and the calculation resources and the reasoning time delay occupied by the first device when reasoning the target AI network model can be reduced.

According to the AI network model interaction method provided by the embodiment of the application, the execution main body can be an AI network model interaction device. In the embodiment of the application, an AI network model interaction device executes an AI network model interaction method as an example, and the AI network model interaction device provided in the embodiment of the application is described.

Referring to fig. 8, an AI network model interaction device provided in the embodiment of the present application may be a device in a first apparatus, and as shown in fig. 8, the AI network model interaction device 800 may include the following modules:

a first sending module 801, configured to send first information to a second device, where the first information includes information related to compression and/or model reasoning of an AI network model required by the first device;

a first obtaining module 802, configured to obtain related information of a target AI network model, where the target AI network model corresponds to the first information.

Optionally, the first obtaining module 802 includes:

the first receiving unit is used for receiving related information of a target AI network model, wherein the target AI network model is a compressed AI network model or an uncompressed AI network model;

or,

a second receiving unit configured to receive related information of a first AI network model from the second device;

and the first processing unit is used for compressing the first AI network model to obtain a second AI network model, and the target AI network model comprises the second AI network model.

Optionally, the first information includes at least one of:

the first application information represents the function information of an AI network model required by the first equipment;

Optionally, in a case that the second device does not have an AI network model matching the first information, and the first indication information indicates that the first device compresses the AI network model, the first acquisition module 802 includes:

a third receiving unit configured to receive related information and third information of a first AI network model from the second device, where the third information is used to compress the first AI network model into a second AI network model;

The second processing unit is used for compressing the first AI network model according to the third information to obtain the second AI network model, wherein the target AI network model comprises the second AI network model;

and/or the number of the groups of groups,

in the case that the second device does not have an AI network model matching the first information, and the first indication information indicates that the second device compresses the AI network model, the first obtaining module 802 is specifically configured to:

receiving related information of a second AI network model from the second equipment, wherein the target AI network model comprises the second AI network model, and the second AI network model is an AI network model obtained by compressing a first AI network model of the second equipment according to the first information;

and/or the number of the groups of groups,

in the case that the second device does not have an AI network model matching the first information, and the first indication information indicates that the third device compresses the AI network model, the first obtaining module 802 is specifically configured to:

and receiving related information of a second AI network model from the third equipment, wherein the target AI network model comprises the second AI network model, and the second AI network model is an AI network model obtained by compressing the first AI network model from the second equipment.

Optionally, the AI network model interaction device 800 further includes:

and a third sending module, configured to send, to the second device, information related to the second AI network model.

Optionally, in the case that the second device has an AI network model matching the first information, the first obtaining module 802 is specifically configured to:

and receiving related information of an AI network model matched with the first information from the second device, wherein the target AI network model comprises the AI network model matched with the first information.

Optionally, in the case that the first device acquires the second AI network model compressed by the first device or the third device, the AI network model interaction apparatus 800 further includes:

the second acquisition module is used for acquiring a matching result of the second AI network model and the first information;

and the fourth sending module is used for sending the matching result to the second equipment.

Optionally, in a case where the matching result indicates that the second AI network model does not match the first information, the AI network model interaction apparatus 800 further includes:

And a fifth sending module, configured to send first request information to the second device, where the first request information is used to request the second device to update at least one of the third information and the first AI network model.

the updating module is used for updating the first information and sending the updated first information to the second equipment;

and a third acquisition module, configured to acquire a target AI network model corresponding to the updated first information.

The AI network model interaction device 800 provided in this embodiment of the present application may implement each process implemented by the first device in the method embodiment shown in fig. 2, and may obtain the same beneficial effects, so that repetition is avoided, and details are not repeated here.

Referring to fig. 9, another AI network model interaction device provided in the embodiment of the disclosure may be a device in a second apparatus, and as shown in fig. 9, the AI network model interaction device 900 may include the following modules:

a first receiving module 901, configured to receive first information from a first device, where the first information includes information related to compression and/or model reasoning of an AI network model required by the first device;

The second sending module 902 is configured to send, to the first device, relevant information of a target AI network model, where the target AI network model corresponds to the first information, or send, according to the first information, relevant information of a first AI network model, where the first AI network model is used to perform compression processing to obtain a second AI network model, and the second AI network model corresponds to the first information.

Optionally, the first information includes at least one of:

Optionally, the second sending module 902 is specifically configured to:

in the case that the second device has an AI network model matching the first information, the AI network model matching the first information is transmitted to the first device, and the target AI network model includes the AI network model matching the first information.

Optionally, in a case that the second device does not have an AI network model matching the first information, and the first indication information indicates that the second device compresses the AI network model, the AI network model interaction apparatus 900 further includes:

the first processing module is used for compressing the first AI network model according to the first information to obtain a second AI network model;

the second sending module 902 is specifically configured to:

and sending related information of the second AI network model to the first equipment, wherein the target AI network model comprises the second AI network model.

Optionally, the second sending module 902 is specifically configured to:

transmitting related information and third information of a first AI network model to the first device, wherein the third information is used for compressing the first AI network model into a second AI network model, and the second AI network model corresponds to the first information, under the condition that the second device does not have the AI network model matched with the first information, and the first indication information indicates the first device to compress the AI network model; and/or the number of the groups of groups,

And sending related information of the first AI network model and the third information to the third device under the condition that the second device does not have the AI network model matched with the first information and the first indication information indicates the third device to compress the AI network model.

Optionally, the AI network model interaction device 900 further includes:

and the second receiving module is used for receiving the related information of the second AI network model.

Optionally, the AI network model interaction device 900 further includes:

and the third receiving module is used for receiving a judging result from the first equipment, wherein the judging result is used for representing a matching result of the second AI network model and the first information.

Optionally, in a case where the matching result indicates that the second AI network model does not match the first information, the AI network model interaction apparatus 900 further includes:

a fourth receiving module, configured to receive first request information from the first device, and update at least one of the third information and the first AI network model according to the first request information;

And a sixth sending module, configured to send the updated third information and/or updated related information of the first AI network model.

a fifth receiving module, configured to receive updated first information from the first device;

a seventh sending module, configured to send, to the first device, relevant information of a target AI network model that matches the updated first information, or send, by the second device, relevant information of a third AI network model according to the updated first information, where the third AI network model is used to perform compression processing to obtain a fourth AI network model corresponding to the updated first information.

The AI network model interaction device 900 provided in this embodiment of the present application can implement each process implemented by the second device in the method embodiment shown in fig. 7, and can obtain the same beneficial effects, so that repetition is avoided, and details are not repeated here.

The AI network model interaction device in the embodiment of the present application may be an electronic device, for example, an electronic device with an operating system, or may be a component in the electronic device, for example, an integrated circuit or a chip. The electronic device may be a terminal, or may be other devices than a terminal. By way of example, terminals may include, but are not limited to, the types of terminals 11 listed above, other devices may be servers, network attached storage (Network Attached Storage, NAS), etc., and embodiments of the application are not specifically limited.

The AI network model interaction device provided in the embodiment of the present application can implement each process implemented by the method embodiment shown in fig. 2 or fig. 7, and achieve the same technical effects, so that repetition is avoided, and no further description is provided herein.

Optionally, as shown in fig. 10, the embodiment of the present application further provides a communication device 1000, including a processor 1001 and a memory 1002, where the memory 1002 stores a program or an instruction that can be executed on the processor 1001, for example, when the communication device 1000 is used as the first device, the program or the instruction is executed by the processor 1001 to implement the steps of the method embodiment shown in fig. 2, and achieve the same technical effect. When the communication device 1000 is used as a second device, the program or the instructions when executed by the processor 1001 implement the steps of the method embodiment shown in fig. 7, and achieve the same technical effects, and are not repeated here.

The embodiment of the application also provides communication equipment, which comprises a processor and a communication interface, wherein when the communication equipment is used as first equipment, the communication interface is used for sending first information to second equipment, and the first information comprises information related to compression and/or model reasoning of an AI network model required by the first equipment; the communication interface or the processor is used for acquiring related information of a target AI network model, and the target AI network model corresponds to the first information. When the communication device is used as a second device, the communication interface is used for receiving first information from a first device, and sending related information of a target AI network model to the first device or sending related information of the first AI network model according to the first information, wherein the first information comprises information related to compression and/or model reasoning of the AI network model required by the first device, the target AI network model corresponds to the first information, the first AI network model is used for carrying out compression processing to obtain a second AI network model, and the second AI network model corresponds to the first information.

The terminal embodiment corresponds to the method embodiment shown in fig. 2 or fig. 7, and each implementation procedure and implementation manner of the method embodiment shown in fig. 2 or fig. 7 are applicable to the communication device embodiment, and the same technical effects can be achieved.

The embodiment of the present application further provides a readable storage medium, where a program or an instruction is stored, and when the program or the instruction is executed by a processor, the program or the instruction implement each process of the method embodiment shown in fig. 2 or fig. 7, and the same technical effects can be achieved, so that repetition is avoided, and no further description is given here.

Wherein the processor is a processor in the terminal described in the above embodiment. The readable storage medium includes computer readable storage medium such as computer readable memory ROM, random access memory RAM, magnetic or optical disk, etc.

The embodiment of the present application further provides a chip, where the chip includes a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is configured to run a program or an instruction, to implement each process of the method embodiment shown in fig. 2 or fig. 7, and to achieve the same technical effect, so that repetition is avoided, and no further description is given here.

It should be understood that the chips referred to in the embodiments of the present application may also be referred to as system-on-chip chips, or the like.

The embodiments of the present application further provide a computer program/program product, which is stored in a storage medium, and executed by at least one processor to implement the respective processes of the method embodiments shown in fig. 2 or fig. 7, and achieve the same technical effects, and are not repeated herein.

The embodiment of the application also provides a communication system, which comprises: a first device operable to perform the steps of the AI network model interaction method, as shown in fig. 2, and a second device operable to perform the steps of the AI network model interaction method, as shown in fig. 7.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present application is not limited to performing the functions in the order shown or discussed, but may also include performing the functions in a substantially simultaneous manner or in an opposite order depending on the functions involved, e.g., 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.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solutions of the present application may be embodied essentially or in a part contributing to the prior art in the form of a computer software product stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk), comprising several instructions for causing a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present application.

The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those of ordinary skill in the art without departing from the spirit of the present application and the scope of the claims, which are also within the protection of the present application.

Claims

1. An artificial intelligence AI network model interaction method, comprising:

the method comprises the steps that first equipment sends first information to second equipment, wherein the first information comprises information related to compression and/or model reasoning of an AI network model required by the first equipment;

2. The method of claim 1, wherein the first device obtaining information regarding the target AI network model comprises:

3. The method of claim 1, wherein the first information comprises at least one of:

4. The method of claim 3, wherein the first indication information indicates that the first device, the second device, or a third device is to compress an AI network model.

5. The method of claim 4, wherein the first device obtaining information about a target AI network model if the second device does not have an AI network model that matches the first information, and the first indication information indicates that the first device is to compress AI network models, comprises:

The first device compresses the first AI network model according to the third information to obtain the second AI network model, wherein the target AI network model comprises the second AI network model;

and/or the number of the groups of groups,

in a case where the second device does not have an AI network model that matches the first information, and the first indication information indicates that the second device compresses the AI network model, the first device obtains information about a target AI network model, including:

the first device receives related information of a second AI network model from the second device, wherein the target AI network model comprises the second AI network model, and the second AI network model is an AI network model obtained by compressing a first AI network model of the second device according to the first information;

and/or the number of the groups of groups,

in a case where the second device does not have an AI network model matching the first information, and the first indication information indicates that the third device compresses the AI network model, the first device acquires information about a target AI network model, including:

6. The method of claim 5, wherein the third information comprises at least one of: and carrying out compression processing on the first AI network model, wherein the compression processing is carried out on the AI network model by adopting an AI network model compression method and the related limit information of AI network model compression.

7. The method of claim 5, wherein after the first device compresses the first AI network model according to the third information to obtain the second AI network model, the method further comprises:

8. The method of claim 3, wherein, in the case where the second device has an AI network model that matches the first information, the first device obtains information about a target AI network model, comprising:

9. The method of claim 5, wherein if the first device obtains a second AI network model compressed by the first device or a third device, the method further comprises:

the first device sends the matching result to the second device.

10. The method of claim 9, wherein if the matching result indicates that the second AI network model does not match the first information, the method further comprises:

11. The method of claim 9, wherein if the matching result indicates that the second AI network model does not match the first information, the method further comprises:

12. An artificial intelligence AI network model interaction method, comprising:

13. The method of claim 12, wherein the target AI network model is a compressed AI network model or an uncompressed AI network model.

14. The method of claim 12, wherein the first information comprises at least one of:

15. The method of claim 14, wherein the first indication information indicates that the first device, the second device, or a third device is to compress an AI network model.

16. The method of claim 14, wherein the second device transmitting information regarding the target AI network model comprises:

the second device, in the case of having an AI network model matching the first information, transmits, to the first device, information about the AI network model matching the first information, the target AI network model including the AI network model matching the first information.

17. The method of claim 15, wherein, if the second device does not have an AI network model that matches the first information, and the first indication information indicates that the second device is to compress the AI network model, the method further comprises:

18. The method of claim 15, wherein the second device transmitting relevant parameters of a first AI network model based on the first information, comprises:

19. The method of claim 18, wherein the third information comprises at least one of: and the AI network model compression method used when the first AI network model is compressed and the AI network model compression related limit information.

20. The method of claim 18, wherein after the second device transmits the related information of the first AI network model and the third information to the first device or the third device, the method further comprises:

the second device receives information regarding the second AI network model.

21. The method of claim 18, wherein after the second device transmits the related information of the first AI network model and the third information to the first device or the third device, the method further comprises:

22. The method of claim 21, wherein if the matching result indicates that the second AI network model does not match the first information, the method further comprises:

23. The method of claim 21, wherein if the matching result indicates that the second AI network model does not match the first information, the method further comprises:

the second device receives updated first information from the first device;

24. An artificial intelligence AI network model interaction apparatus for use with a first device, the apparatus comprising:

25. An artificial intelligence AI network model interaction apparatus for use with a second device, the apparatus comprising:

26. A communication device comprising a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the artificial intelligence AI network model interaction method of any of claims 1-23.

27. A readable storage medium having stored thereon a program or instructions which when executed by a processor performs the steps of the artificial intelligence AI network model interaction method of any of claims 1-23.