Communication method, device, system, storage medium and program product

Abstract

The present disclosure relates to a communication method, device, system, storage medium and program product. The communication method comprises: a network device sending first information to a terminal device at a first time, the first information being used to instruct the terminal device to use a model, a time interval between the first time and a second time being greater than or equal to a first length, the second time being a time when the network device completes sending the model to the terminal device. The present disclosure can avoid communication errors as much as possible and improve the efficiency of the communication system.

Description

Communication methods, devices, systems, storage media and software products Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to communication methods, devices, systems, storage media, and program products. Background Technology

[0002] In recent years, artificial intelligence (AI) and machine learning (ML) technologies have made continuous breakthroughs in many fields. Summary of the Invention

[0003] This disclosure provides communication methods, devices, systems, storage media, and program products.

[0004] According to a first aspect of the present disclosure, a communication method is proposed, the method comprising: a network device sending first information to a terminal device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, and the second moment being the moment when the network device completes sending the model to the terminal device.

[0005] According to a second aspect of the present disclosure, a communication method is proposed, the method comprising: a terminal device receiving first information sent by a network device, the first information being sent by the network device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, and the second moment being the moment when the network device completes sending the model to the terminal device.

[0006] According to a third aspect of the present disclosure, a network device is provided, comprising: a transceiver module, configured to send first information to a terminal device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, the second moment being the moment when the network device completes sending the model to the terminal device.

[0007] According to a fourth aspect of the present disclosure, a terminal device is provided, comprising: a transceiver module, configured to receive first information sent by a network device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, and the second moment being the moment when the network device completes sending the model to the terminal device.

[0008] According to a fifth aspect of the present disclosure, a network device is provided, comprising: one or more processors; wherein the network device is configured to perform the first aspect and any one of the communication methods in the first aspect.

[0009] According to a sixth aspect of the present disclosure, a terminal device is provided, comprising: one or more processors; wherein the terminal device is configured to execute the second aspect and any one of the communication methods in the second aspect.

[0010] According to a seventh aspect of the present disclosure, a communication system is provided, including a network device and a terminal device, wherein the network device is configured to implement the first aspect and any one of the communication methods in the first aspect, and the terminal device is configured to implement the second aspect and any one of the communication methods in the second aspect.

[0011] According to an eighth aspect of the present disclosure, a storage medium is provided that stores instructions which, when executed on a communication device, cause the communication device to perform a communication method as described in the first aspect and any one thereof, or the second aspect and any one thereof.

[0012] According to a ninth aspect of the present disclosure, a program product is provided, comprising: a computer program, which, when executed by a communication device, causes the communication device to perform a communication method as described in the first aspect and any one of the first aspects or the second aspect and the second aspect.

[0013] This disclosure involves a network device sending first information to a terminal device at a first moment to instruct the terminal device to use a model, such as performing inference, but not limited to this. The time interval between the first moment and the moment when the network device completes sending the model to the terminal device (i.e., the second moment) is greater than or equal to the first duration, thereby avoiding communication errors and improving the efficiency of the communication system. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings required for the description of the embodiments are introduced below. The following drawings are only some embodiments of this disclosure and do not impose specific limitations on the protection scope of this disclosure.

[0015] Figure 1a is a schematic diagram of a communication system architecture according to an embodiment of the present disclosure.

[0016] Figure 1b is a schematic diagram of a network device instructing a terminal device to use a model.

[0017] Figure 2 is a schematic diagram of a communication method interaction according to an embodiment of the present disclosure.

[0018] Figure 3 is a flowchart illustrating a communication method according to an embodiment of the present disclosure.

[0019] Figure 4 is a flowchart illustrating a communication method according to an embodiment of the present disclosure.

[0020] Figure 5a is a schematic diagram of the communication method interaction according to an embodiment of the present disclosure.

[0021] Figure 5b is a schematic diagram illustrating the interaction of the model transfer method according to an embodiment of the present disclosure.

[0022] Figure 6a is a schematic diagram of the structure of the terminal device proposed in an embodiment of this disclosure.

[0023] Figure 6b is a schematic diagram of the structure of the network device proposed in an embodiment of this disclosure.

[0024] Figure 7a is a schematic diagram of the structure of a communication device proposed in an embodiment of this disclosure.

[0025] Figure 7b is a schematic diagram of the chip structure proposed in an embodiment of this disclosure. Detailed Implementation

[0026] This disclosure provides communication methods, devices, systems, storage media, and program products.

[0027] In a first aspect, embodiments of this disclosure propose a communication method, the method comprising: a network device sending first information to a terminal device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, and the second moment being the moment when the network device completes sending the model to the terminal device.

[0028] In some alternative embodiments of the first aspect, the first duration is determined based on second information, the second information including at least one of the following: processing capability information of the terminal device; type information of the model; structural information of the model; parameter information of the model; size information of the model; and format information of the model.

[0029] In some alternative embodiments of the first aspect, the parameter information of the model includes at least one of the following: the number of parameters of the model; the data type of the model.

[0030] In some alternative embodiments of the first aspect, the first duration is determined based on the second duration corresponding to each item in the second information.

[0031] In some alternative embodiments of the first aspect, the value or range of a piece of information in the second information corresponds to a value of the second duration.

[0032] In some alternative embodiments of the first aspect, the first duration is equal to the sum of the second durations corresponding to each item in the second information.

[0033] In some alternative embodiments of the first aspect, the first duration is determined based on the third duration and a first scaling factor corresponding to each item in the second information.

[0034] In some alternative embodiments of the first aspect, the third duration is predefined by the protocol, or the third duration is sent by the terminal device to the network device.

[0035] In some alternative embodiments of the first aspect, the value or range of a piece of information in the second information corresponds to the value of a first scaling factor.

[0036] In some alternative embodiments of the first aspect, the first duration is equal to the third duration multiplied by a first scaling factor corresponding to each item in the second information.

[0037] In some alternative embodiments of the first aspect, the first duration is determined based on the fourth duration corresponding to the first part of the second information and the second scaling factor corresponding to the second part of the second information.

[0038] In some alternative embodiments of the first aspect, the value or range of a piece of information in the first part of the information corresponds to a value of the fourth duration.

[0039] In some alternative embodiments of the first aspect, the value or range of a piece of information in the second part of the information corresponds to the value of a second scaling factor.

[0040] In some alternative embodiments of the first aspect, the first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

[0041] In some alternative embodiments of the first aspect, the first part of the information includes at least one of the following: processing capability information of the terminal device; format information of the model; and type information of the model.

[0042] In some alternative embodiments of the first aspect, the second part of the information includes at least one of the following: parameter information of the model; size information of the model; and structural information of the model.

[0043] In some alternative embodiments of the first aspect, different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

[0044] In some alternative embodiments of the first aspect, the model includes at least one of the following: an artificial intelligence (AI) model; a machine learning (ML) model.

[0045] In a second aspect, a communication method is provided, the method comprising: a terminal device receiving first information sent by a network device, the first information being sent by the network device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, and the second moment being the moment when the network device completes sending the model to the terminal device.

[0046] In some alternative embodiments of the second aspect, the first duration is determined based on second information, which includes at least one of the following: processing capability information of the terminal device; type information of the model; structural information of the model; parameter information of the model; size information of the model; and format information of the model.

[0047] In some alternative embodiments of the second aspect, the parameter information of the model includes at least one of the following: the number of parameters of the model; the data type of the model.

[0048] In some alternative embodiments of the second aspect, the first duration is determined based on the second duration corresponding to each item in the second information.

[0049] In some alternative embodiments of the second aspect, the value or range of a piece of information in the second information corresponds to a second duration.

[0050] In some alternative embodiments of the second aspect, the first duration is equal to the sum of the second durations corresponding to each item in the second information.

[0051] In some alternative embodiments of the second aspect, the first duration is determined based on the third duration and the first scaling factor corresponding to each item in the second information.

[0052] In some alternative embodiments of the second aspect, the third duration is predefined by the protocol, or the third duration is sent by the terminal device to the network device.

[0053] In some alternative embodiments of the second aspect, the value or range of a piece of information in the second information corresponds to a first scaling factor.

[0054] In some alternative embodiments of the second aspect, the first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information.

[0055] In some alternative embodiments of the second aspect, the first duration is determined based on the fourth duration corresponding to the first part of the information in the second information and the second scaling factor corresponding to the second part of the information in the second information.

[0056] In some alternative embodiments of the second aspect, the value or range of a piece of information in the first part of the information corresponds to a fourth duration.

[0057] In some alternative embodiments of the second aspect, the value or range of a piece of information in the second part of the information corresponds to a second scaling factor.

[0058] In some alternative embodiments of the second aspect, the first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

[0059] In some alternative embodiments of the second aspect, the first part of the information includes at least one of the following: the processing capability information of the terminal device; the format information of the model; and the type information of the model.

[0060] In some alternative embodiments of the second aspect, the second part of the information includes at least one of the following: parameter information of the model; size information of the model; and structural information of the model.

[0061] In some alternative embodiments of the second aspect, different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

[0062] In some alternative embodiments of the second aspect, the model includes at least one of the following: an artificial intelligence (AI) model; a machine learning (ML) model.

[0063] Thirdly, a network device is provided, comprising: a transceiver module, configured to send first information to a terminal device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, the second moment being the moment when the network device completes sending the model to the terminal device.

[0064] Fourthly, a terminal device is provided, comprising: a transceiver module, configured to receive first information sent by a network device at a first moment, the first information being used to instruct the terminal device to use a model, the time interval between the first moment and a second moment being greater than or equal to a first duration, the second moment being the moment when the network device completes sending the model to the terminal device.

[0065] Fifthly, a network device is provided, including one or more processors; wherein the network device is configured to perform the first aspect and any one of the communication methods in the first aspect.

[0066] A sixth aspect provides a terminal device, comprising: one or more processors; wherein the terminal device is configured to execute the second aspect and any one of the communication methods in the second aspect.

[0067] A seventh aspect provides a communication system, including a network device and a terminal device, wherein the network device is configured to implement the first aspect and any one of the communication methods in the first aspect, and the terminal device is configured to implement the second aspect and any one of the communication methods in the second aspect.

[0068] Eighthly, a storage medium is provided that stores instructions, which, when executed on a communication device, cause the communication device to perform a communication method as described in the first aspect and any one thereof, or the second aspect and any one thereof.

[0069] Ninthly, embodiments of this disclosure provide a program product that, when executed by a communication device, causes the communication device to perform the method as described in the optional implementations of the first or second aspect.

[0070] In a tenth aspect, embodiments of this disclosure provide a computer program that, when run on a computer, causes the computer to perform the methods described in an optional implementation of the first or second aspect.

[0071] Eleventhly, embodiments of this disclosure provide a chip or chip system. The chip or chip system includes processing circuitry configured to perform the methods described in the optional implementations of the first or second aspect above.

[0072] It is understood that the terminal devices, access network devices, first network elements, other network elements, core network devices, communication systems, storage media, program products, computer programs, chips, or chip systems involved in the embodiments of this disclosure are all used to execute the methods proposed in the embodiments of this disclosure. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0073] This disclosure provides communication methods, devices, systems, storage media, and program products. In some embodiments, the terms "communication method" and "information processing method" can be used interchangeably, as can the terms "communication device" and "information processing device" and "communication device," and the terms "information processing system" and "communication system."

[0074] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0075] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. The technical environments of different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0076] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0077] In this embodiment of the disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular expression or a plural expression.

[0078] In the embodiments disclosed herein, "multiple" refers to two or more.

[0079] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0080] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "in response to one case A, in response to another case B", etc., may include the following technical solutions depending on the situation: in some embodiments, A (execute A regardless of B); in some embodiments, B (execute B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). The same applies when there are more branches such as A, B, C, etc.

[0081] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, C, etc.

[0082] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, quantity, or content of the descriptive objects. The description of the descriptive objects is found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers and can be one or more. For example, in "first device," the number of "devices" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the object being described is "information", then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0083] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0084] In some embodiments, the terms “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “if…”, “if…”, etc., can be used interchangeably.

[0085] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0086] In some embodiments, the apparatus and device may be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they may also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "body", etc.

[0087] In some embodiments, "network" can be interpreted as devices included in a network, such as access network devices, core network devices, etc.

[0088] In some embodiments, "access network device (AN device)" may also be referred to as "radio access network device (RAN device)," "base station (BS)," "radio base station," or "fixed station." In some embodiments, it may also be understood as "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," or "bandwidth part (BWP)."

[0089] In some embodiments, the terms "terminal", "terminal device", "user equipment (UE)", "user terminal", "mobile station (MS)", "mobile terminal (MT)", "subscriber station", "mobile unit", "subscriber unit", "wireless unit", "remote unit", "mobile device", "wireless device", "wireless communication device", "remote device", "mobile subscriber station", "access terminal", "mobile terminal", "wireless terminal", "remote terminal", "handset", "user agent", "mobile client", and "client" can be used interchangeably.

[0090] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0091] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0092] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0093] Figure 1a is a schematic diagram of a communication system architecture according to an embodiment of the present disclosure.

[0094] As shown in Figure 1a, the communication system 100 includes a terminal device 101 and a network device 102.

[0095] In some embodiments, terminal device 101 includes, but is not limited to, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home.

[0096] In some embodiments, network device 102 may include at least one of access network device and core network device.

[0097] In some embodiments, the access network device is, for example, a node or device that connects a terminal device to a wireless network. The access network device may include, but is not limited to, at least one of the following in a 5G communication system: evolved Node B (eNB), next-generation eNB (ng-eNB), next-generation Node B (gNB), node B (NB), home node B (HNB), home evolved node B (HeNB), radio backhaul device, radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), base band unit (BBU), mobile switching center, base station in a 6G communication system, open RAN, cloud RAN, base station in other communication systems, and access node in a Wi-Fi system.

[0098] In some embodiments, the technical solutions of this disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within access network devices involved in the embodiments of this disclosure can be transformed into internal interfaces of Open RAN. The processes and information interactions between these internal interfaces can be implemented by software or programs.

[0099] In some embodiments, the access network device may be composed of a central unit (CU) and a distributed unit (DU). The CU may also be called a control unit. The CU-DU structure can separate the protocol layer of the access network device. Some of the protocol layer functions are centrally controlled by the CU, while the remaining part or all of the protocol layer functions are distributed in the DU and centrally controlled by the CU. However, this is not the only possibility.

[0100] In some embodiments, a core network device may be a single device comprising one or more network elements, or it may be multiple devices or a group of devices, each comprising all or part of the aforementioned one or more network elements. Network elements may be virtual or physical. The core network may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), or a Next Generation Core (NGC).

[0101] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions proposed in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions proposed in this disclosure are also applicable to similar technical problems.

[0102] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1a, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1a are illustrative. The communication system may include all or some of the main bodies in FIG1a, or it may include other main bodies outside of FIG1a. The number and form of each main body are arbitrary. Each main body may be physical or virtual. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection may be in any way, such as direct connection or indirect connection, wired connection or wireless connection.

[0103] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 6th generation mobile communication system (6G), 5G New Radio (NR), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New Radio Access (NX), Future Generation Radio Access (FX), Global System for Mobile Communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0104] The development of communication technologies has brought tremendous changes to all aspects of people's lives, such as 5G (NR, 6G, etc.). According to the vision of the International Telecommunication Union (ITU), 5G will permeate all areas of future society, building a comprehensive information ecosystem centered on the user. Specifically, 5G user experience speeds can reach 100 Mbit / s to 1 Gbit / s, supporting ultimate service experiences such as mobile virtual reality; 5G peak speeds can reach 10 Gbit / s to 20 Gbit / s, with a traffic density of 10 Mbit / s / m², supporting more than a thousandfold increase in mobile traffic; 5G connection density can reach 1 million / m², effectively supporting massive numbers of IoT devices; 5G transmission latency can be down to the millisecond level, meeting the stringent requirements of vehicle-to-everything (V2X) and industrial control; 5G can support mobile speeds of 500 km / h, providing a good user experience even in high-speed rail environments. It is conceivable that 5G, as a representative of new infrastructure, will reshape the future information society.

[0105] In recent years, artificial intelligence (AI) and machine learning (ML) technologies have achieved continuous breakthroughs in multiple fields. The ongoing development of fields such as intelligent voice and computer vision has not only brought a wide variety of applications to smart terminal devices, but has also found widespread use in education, transportation, home, healthcare, retail, security, and many other sectors, bringing convenience to people's lives while promoting industrial upgrading across various industries. AI technology is also accelerating its cross-disciplinary integration with other disciplines, combining knowledge from different fields while providing new directions and methods for the development of various disciplines.

[0106] In the 3rd generation partnership project (3GPP) Release 18, a research project on the application of artificial intelligence (AI) technology in the radio access network (RAN) working group (WG) 1 was established. This project aimed to investigate how to introduce AI technology into the RAN and explore how AI technology can assist in improving RAN transmission technology.

[0107] In research geared towards 6G, 6G systems can provide AI services across more dimensions. This mainly includes the following three aspects:

[0108] AI-enabled connectivity. This means using AI to improve communication performance, such as using AI for beam management.

[0109] Computing power services. This refers to the network side providing computing power to the terminal device side, such as assisting the terminal device in model training, model inference, and other advanced AI services. Essentially, it involves enhancing the network's transmission pipeline to improve the user experience of AI application services.

[0110] Currently, research has been conducted on AI / ML model transfer / delivery. AI / ML model transfer / delivery is typically initiated by network devices, which send pre-trained AI / ML models to terminal devices. The terminal devices then use the received AI / ML models for inference. However, before performing inference using the AI / ML models sent by the network devices, the terminal devices need to perform operations such as AI / ML model reading and AL / ML model compilation. These operations generate significant computational loads and consume considerable time. In other words, after sending the AI / ML model to the terminal device, the network device cannot immediately schedule the terminal device to use the AI / ML model or perform any associated operations related to it; otherwise, the terminal device will be unable to execute operations as instructed by the network device, leading to communication errors.

[0111] Therefore, this disclosure provides a communication method in which a network device sends first information to a terminal device at a first moment to instruct the terminal device to use a model, such as performing inference using the model, but not limited thereto. The time interval between the first moment and the moment when the network device completes sending the model to the terminal device (i.e., the second moment) is greater than or equal to the first duration, thereby avoiding communication errors and improving the efficiency of the communication system.

[0112] For example, Figure 1b is a schematic diagram of a network device instructing a terminal device to use a model. As shown in Figure 1b, t represents time, and t1, t2, t3, etc., represent different times. The network device starts sending the model to the terminal device from t1 and completes the model transmission at t2, that is, the model is sent within T1. The network device instructs the terminal device to use the model at t3. In this disclosure, the first time can be t3 in Figure 1b, and the second time can be t2 in Figure 1b.

[0113] Figure 2 is a schematic diagram of a communication method interaction according to an embodiment of the present disclosure. As shown in Figure 2, this embodiment of the present disclosure relates to a communication method for a communication system 100, the method including:

[0114] In step S2101, network device 102 sends the model to terminal device 101.

[0115] In some embodiments, terminal device 101 receives a model sent by network device 102.

[0116] In some embodiments, the model may be an AI model, an ML model, etc., but is not limited to this.

[0117] It is understood that the model in this embodiment is not a physical model, but a logical model. That is, sending a model from a network device to a terminal device can be understood as the network device sending relevant information about the model to the terminal device. Receiving a model from a terminal device can be understood as the terminal device receiving relevant information about the model. The terminal device can construct a model based on this relevant information. This relevant information includes, but is not limited to, at least one of the following: model type information, model structure information, model parameter information, model size information, and model format information.

[0118] In step S2102, network device 102 sends first information to terminal device 101 at the first moment.

[0119] In some embodiments, terminal device 101 receives first information sent by network device 102 at a first moment.

[0120] In some embodiments, the first information is used to instruct the terminal device to use the model, the time interval between the first moment and the second moment is greater than or equal to the first duration, and the second moment is the moment when the network device completes sending the model to the terminal device. Using the model can be performing inference using the model, or performing associated operations of the model, but is not limited to these. Since the terminal device may need to perform operations such as model reading and model compilation before using the model, and these operations generate a large amount of computation and consume time, the network device, after completing sending the model to the terminal device, does not immediately schedule the terminal device to use the model to avoid communication errors. Instead, it waits for the first duration before instructing the terminal device to use the model, thereby avoiding communication errors.

[0121] In some embodiments, the network device considers the terminal device ready to perform inference using the model after it has finished sending the model to the terminal device and a first period of time has elapsed. Therefore, the network device instructs the terminal device to perform inference using the AI / ML model at the first moment.

[0122] In some embodiments, the first duration can be understood as the duration required for the terminal device to prepare the model for inference, that is, the time interval from the moment the network device finishes sending the model to the moment the terminal device is ready to use the model for inference.

[0123] It is understood that, for ease of description, this disclosure refers to the moment when the network device sends the first information as the first moment and the moment when the network device completes sending the model to the terminal device as the second moment, but this disclosure does not limit the names of the first moment and the second moment. For example, the second moment can be referred to as the moment when the network device completes sending the model to the terminal device. As another example, the first moment can be referred to as the moment when the network device completes sending the model to the terminal device and has experienced a duration greater than or equal to the first time.

[0124] Understandably, the name of the first duration is not limited; it can be, for example, the first time period.

[0125] To further avoid communication errors and reduce latency, this disclosure proposes a method for designing a first duration, or a method for determining a first duration, as illustrated in the following embodiments.

[0126] In some embodiments, the first duration is determined based on second information, which includes at least one of the following: processing capability information of the terminal device; model type information; model structure information; model parameter information; model size information; and model format information.

[0127] Optionally, the second information may include the processing capability information of the terminal device. For example, the processing capability information of the terminal device may depend on the processor type, number of processors, processor clock speed, processor bit width, and memory read / write frequency of the terminal device, but is not limited to these. Since the processing capability of the terminal device affects the time required for the terminal device to read and compile the model, the processing capability information of the terminal device can be used as one of the factors in determining the first duration. That is, the second information may include the processing capability information of the terminal device. The network device can determine the first duration based on the processing capability information of the terminal device, and determine the first moment based on the first duration and the second moment, sending the first information at the first moment to instruct the terminal device to use the model.

[0128] Optionally, the second information may include model type information. For example, model type information can also be referred to as the model's backbone. For instance, model type information may include deep neural networks (DNN), convolutional neural networks (CNN), residual networks (ResNet), transformers, etc. Since different model types have different computational requirements, affecting the time it takes for the terminal device to read and compile the model, the model type information can be used as one of the factors in determining the first duration. That is, the second information may include model type information. The network device can determine the first duration based on the model type information, and determine the first time step based on the first duration and the second time step. At the first time step, the first information is sent to instruct the terminal device to use the model.

[0129] Optionally, the second information may include the model's structural information. For example, the model's structural information includes the modules that make up the model and the dimensions of each module. Different model types may correspond to different structural information. Taking a DNN as an example, the model's structural information includes the number of fully connected layers in the DNN and the number of neurons in each layer. Taking a CNN as an example, the model's structural information includes the dimensions of the DNN's input layer, the number and dimensions of convolutional layers, pooling layers, fully connected layers, and the dimensions of the output layer. Taking a Transformer as an example, the model's structural information includes the dimensions of the Transformer's input layer, the number of encoders, the number of decoders, the dimensions of self-attention layers and fully connected network layers included in the decoder and encoder, and the dimensions of the output layer. It is understood that the above examples are merely illustrative, and the type of model is not limited to these examples; the structural information corresponding to a particular model type is not limited to these specific examples. Since the structural information of a model affects the time required for terminal devices to read and compile the model—for example, the more complex the model structure, the longer the time required for model reading and compilation; conversely, the simpler the model structure, the shorter the time required for model reading and compilation—the structural information of the model can be used as one of the factors in determining the first duration. That is, the second information can include the structural information of the model. The network device can determine the first duration based on the structural information of the model, and determine the first moment based on the first duration and the second moment. At the first moment, the first information is sent to instruct the terminal device to use the model.

[0130] Optionally, the second information may include model parameter information. For example, model parameter information includes the number of model parameters and the model's data type. For instance, the number of model parameters affects the time required for the terminal device to read and compile the model. More parameters result in a larger model size, leading to longer reading and compilation times; conversely, fewer parameters result in a smaller model size, leading to shorter reading and compilation times. As another example, the model's data type can refer to the data type of the model parameters. Information such as the number of bits, numerical range, and numerical precision of various data types affects the time required for the terminal device to read and compile the model. For example, higher bit counts for a data type result in longer reading and compilation times; conversely, lower bit counts result in shorter reading and compilation times for the AI / ML model. The number of bits can also be referred to as the quantization bit depth. For example, data types may include, but are not limited to, FP32, FP16, TF32, BF16, Int32, Int16, and Int8. Information regarding the number of bits, numerical range, and numerical precision of these data types can be found in Table 1 below. Here, FP represents a floating-point number, FP32 represents a 32-bit floating-point number, FP16 represents a 16-bit floating-point number, TF represents a tensor flow, TF32 represents a 32-bit tensor, BF represents a brain floating-point format (bflot), BF16 represents a 16-bit bit field, and Int represents an integer, Int32 represents a 32-bit integer, Int16 represents a 16-bit integer, and Int8 represents an 8-bit integer.

[0131] Table 1

[0132] Optionally, the second information may include model size information. For example, model size information may refer to the total number of bits in the model, the storage size of the model, or the file size of the model. Here, the model file refers to the carrier of the model sent by the network device to the terminal device. The size of the model is usually determined by both the model's structural information and its parameter information. For example, the more complex the model's structure, the more parameters it has, and the higher the quantization precision of its data types, the larger the model will be, and the longer the model reading and compilation time will be; conversely, the simpler the model's structure, the fewer the parameters it has, and the lower the quantization precision of its data types, the smaller the model will be, and the shorter the AI / ML model reading and compilation time will be. Since model size information affects the time required for the terminal device to read and compile the model, it can be used as one of the factors in determining the first duration. That is, the second information may include model size information. The network device can determine the first duration based on the model size information, and determine the first moment based on the first duration and the second moment, sending the first information at the first moment to instruct the terminal device to use the model.

[0133] Optionally, the second information may include the model's format information. For example, the model's format may refer to the model's storage format or the model file format. For example, the model's format may include, but is not limited to, Open Neural Network Exchange (ONNX) and Intermediate Representation (IR). Since the model's format information affects the time it takes for the terminal device to read, compile, and perform other operations on the model, for example, the terminal device may optimize different model formats differently. Therefore, the model's format information can be used as one of the factors in determining the first duration. That is, the second information may include the model's format information. The network device can determine the first duration based on the model's format information, and determine the first time step based on the first duration and the second time step. At the first time step, the first information is sent to instruct the terminal device to use the model.

[0134] It is understood that the above optional examples can be combined arbitrarily, that is, the second information may include at least one of the above information. For example, if the second information includes the processing capability information of the terminal device, then the first duration can be determined based on the processing capability information of the terminal device. If the second information includes the processing capability information of the terminal device and the type information of the model, then the first duration can be determined based on the processing capability information of the terminal device and the type information of the model. This disclosure does not provide examples one by one, but is not limited to the examples given herein.

[0135] In some embodiments, the first duration is determined based on the second information, which may be based on the second duration corresponding to each item in the second information. For example, each item in the second information corresponds to a second duration, and the first duration is determined based on these second durations. For example, if the second information includes the processing capability information of the terminal device and the type information of the model, where the processing capability information of the terminal device corresponds to the second duration A and the type information of the model corresponds to the second duration B, then the first duration can be determined based on the second duration A and the second duration B. As another example, if the second information includes the type information of the model, the structure information of the model, and the number of parameters of the model, where the type information of the model corresponds to the second duration B, the structure information of the model corresponds to the second duration C, and the number of parameters of the model corresponds to the second duration D, then the first duration can be determined based on the second duration B, the second duration C, and the second duration D. This disclosure does not provide a complete list of examples, but is not limited to these.

[0136] In some embodiments, the value or range of a single piece of information in the second information corresponds to a value of a second duration. For example, each piece of information corresponds to a second duration, and the magnitude of the second duration corresponding to each piece of information depends on the value or range of that information. For example, if the value or range of a certain piece of information is different, then the value of the second duration corresponding to that piece of information will also be different.

[0137] Optionally, the second information includes the processing capability information of the terminal device. The value or range of the terminal device's processing capability information corresponds to a value of a second duration. It is understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the terminal device's processing capability information corresponds to a value of a second duration" refers to the second duration corresponding to the processing capability information of the terminal device among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the processing capability information of the terminal device among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the processing capability information of the terminal device as second duration A.

[0138] For example, the value corresponding to the processing capability information of the terminal device remains the same for different models; that is, the processing capability information of the terminal device can correspond to a fixed second duration A. It is understood that "value" and "value range" in this disclosure do not only refer to numerical values ​​and ranges. For example, the processing capability information of the terminal device can be the processor type of the terminal device, and its value can be "CPU" or "GPU." Its value range can be a set including CPU and GPU, as well as a set consisting of other processor types; a set is a value range. Taking processor type as an example, if the processor type of the terminal device is CPU, then regardless of which model the terminal device receives from the network device, and regardless of which model the network device indicates the terminal device is using through the first information, the processor type of the terminal device is always CPU. In other words, the value corresponding to the processing capability information of the terminal device is fixed, and the second duration A corresponding to the processing capability information of the terminal device is also fixed. For ease of description, the value of the second duration A in this disclosure can be a first value. If the second information includes the processing capability information of the terminal device, then the first duration A can be determined based on the first value.

[0139] For example, the value or range of the processing capability information of the terminal device may differ for different models; that is, different values ​​of the processing capability information of the terminal device will correspond to different values ​​of the second duration A. For instance, when the network device sends different models, the terminal device can activate different processor types. For example, if the network device sends model 1, the terminal device can activate the CPU; if the network device sends model 2, the terminal device can activate the GPU. The CPU, as a value of the processing capability information of the terminal device, can correspond to a value of the second duration A, let's assume it's a second duration A1. Similarly, the GPU, as a value of the processing capability information of the terminal device, can correspond to a value of the second duration A, let's assume it's a second duration A2. If, after sending model 1, the network device needs to instruct the terminal device to use model 1 through the first information at the first moment, the first moment can be determined based on the first duration, and the first duration can be determined based on the second duration A corresponding to the processing capability information of the terminal device. The value of the second duration A depends on the value corresponding to the processing capability information of the terminal device. For Model 1, assuming the terminal device has activated the CPU (meaning the terminal device's processing power information corresponds to the CPU), the second duration A is set to the second duration A1. Therefore, the first duration can be determined based on the second duration A1, and the first moment can be determined based on the first duration. At the first moment, the first message is sent to instruct the terminal device to use Model 1. Similarly, if the network device, after sending Model 2, wants to instruct the terminal device to use Model 2 via the first message at the first moment, the first moment can be determined based on the first duration. The first duration can be determined based on the second duration A corresponding to the terminal device's processing power information, and the value of the second duration A depends on the value corresponding to the terminal device's processing power information. For Model 2, assuming the terminal device has activated the GPU (meaning the terminal device's processing power information corresponds to the GPU), the second duration A is set to the second duration A2. Therefore, the first duration can be determined based on the second duration A2, and the first moment can be determined based on the first duration. At the first moment, the first message is sent to instruct the terminal device to use Model 2.

[0140] Optionally, the second information includes model type information. The value or range of the model type information corresponds to a value of a second duration. It is understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the model type information corresponds to a value of a second duration" refers to the second duration corresponding to the model type information among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the model type information among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the model type information as second duration B.

[0141] For example, taking "DNN", "CNN", "ResNet", and "Transformer" as examples of model type information, the value of the second duration B can be as shown in Table 2. For instance, when the model type information is DNN, the second duration B takes the second value. If the network device instructs the terminal device to use a model of type DNN through the first information, then the second duration B takes the second value. The first duration can be determined based on the second value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. As another example, when the model type information is CNN, the second duration B takes the third value. If the network device instructs the terminal device to use a model of type CNN through the first information, then the second duration B takes the third value. The first duration can be determined based on the third value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not exemplify every item in Table 2, but is not limited to the examples given. It is understood that Table 2 is merely an example, and this disclosure does not limit which values ​​the model type information includes, nor does it limit the size of the second to fifth values.

[0142] Table 2

[0143] Optionally, the second information includes the model's structural information. The value or range of the model's structural information corresponds to a value of a second duration. It is understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the model's structural information corresponds to a value of a second duration" refers to the second duration corresponding to the model's structural information among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the model's structural information among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the model's structural information as the second duration C.

[0144] For example, taking a model of type DNN and its structural information as the number of neurons in a DNN, the value of the second duration C can be as shown in Table 3. For instance, when the number of neurons falls within the first range, the second duration C is the sixth value. If the network device instructs the terminal device to use a model of type DNN with the first information, and the number of neurons falls within the first range, then the second duration C is the sixth value. The first duration can be determined based on the sixth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. As another example, when the number of neurons falls within the second range, the second duration C is the seventh value. If the network device instructs the terminal device to use a model of type DNN with the first information, and the number of neurons falls within the second range, then the second duration C is the seventh value. The first duration can be determined based on the seventh value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not provide examples for every item in Table 3, but is not limited to the examples given. It is understood that Table 3 is merely an example, and this disclosure does not limit what the structural information of the model is, nor does it limit which values ​​the structural information of the model includes, nor does it limit the size of the sixth to eighth values.

[0145] For example, there are no overlapping values ​​between the first and third value ranges in Table 3.

[0146] Table 3

[0147] Optionally, the second information includes model parameter information, and the model parameter information includes the number of model parameters. That is, the second information includes the number of model parameters. The value or range of the number of model parameters corresponds to a value of a second duration. It can be understood that the second information includes at least one piece of information, each of which corresponds to a second duration, that is, the second information corresponds to at least one second duration. The "a second duration" in "the value or range of the number of model parameters corresponds to a value of a second duration" refers to the second duration corresponding to the number of model parameters in the at least one second duration corresponding to the second information, and also refers to the second duration corresponding to the number of model parameters in the at least one second duration used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the number of model parameters as the second duration D.

[0148] For example, the range of values ​​for the number of model parameters and the value of the second duration D can be as shown in Table 4. For instance, when the range of values ​​for the number of model parameters is the fourth range, the value of the second duration D is the ninth value. If the network device instructs the terminal device to use a model whose number of parameters is within the fourth range via the first information, then the value of the second duration D is the ninth value. The first duration can be determined based on the ninth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. As another example, when the range of values ​​for the number of model parameters is the fifth range, the value of the second duration D is the tenth value. If the network device instructs the terminal device to use a model whose number of parameters is within the fifth range via the first information, then the value of the second duration D is the tenth value. The first duration can be determined based on the tenth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not exemplify every item in Table 4, but is not limited to the examples given. It is understood that Table 4 is merely an example, and this disclosure does not limit the number of parameters of the model, nor does it limit the value range of the ninth to twelfth values.

[0149] For example, there are no overlapping values ​​between any two of the fourth to seventh value ranges in Table 4.

[0150] Table 4

[0151] Optionally, the second information includes the model's parameter information, and the model's parameter information includes the model's data type. That is, the second information includes the model's data type. The value or range of the model's data type corresponds to a value of a second duration. It can be understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the model's data type corresponds to a value of a second duration" refers to the second duration corresponding to the model's data type among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the model's data type among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the model's data type as the second duration E.

[0152] For example, the values ​​of the model's data type and the second duration E can be as shown in Table 5. For instance, when the model's data type is FP32, the second duration E is the thirteenth value. If the network device instructs the terminal device to use a model with data type FP32 via the first information, then the second duration E is the thirteenth value. The first duration can be determined based on the thirteenth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. As another example, when the model's data type is FP16, the second duration E is the fourteenth value. If the network device instructs the terminal device to use a model with data type FP16 via the first information, then the second duration E is the fourteenth value. The first duration can be determined based on the fourteenth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not exemplify every item in Table 5, but is not limited to the examples given. It is understood that Table 5 is merely exemplary, and this disclosure does not limit the data types of the model to include any range of values, nor does it limit the size of the thirteenth to nineteenth values.

[0153] Table 5

[0154] Optionally, the second information includes model size information. The value or range of the model size information corresponds to a value of a second duration. It is understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the model size information corresponds to a value of a second duration" refers to the second duration corresponding to the model size information among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the model size information among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the model size information as the second duration F.

[0155] For example, the range of values ​​for the model size information and the value of the second duration F can be as shown in Table 6. For instance, when the range of values ​​for the model size information is the eighth range, the value of the second duration F is the twentieth value. If the network device instructs the terminal device to use a model whose size is within the eighth range via the first information, then the value of the second duration F is the twentieth value. The first duration can be determined based on the twentieth value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. As another example, when the range of values ​​for the model size information is the ninth range, the value of the second duration F is the twenty-first value. If the network device instructs the terminal device to use a model whose size is within the ninth range via the first information, then the value of the second duration F is the twenty-first value. The first duration can be determined based on the twenty-first value, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not exemplify every item in Table 6, but is not limited to the examples given. It is understood that Table 6 is merely an example, and this disclosure does not limit which value ranges the size information of the model includes, nor does it limit the size of the twentieth to twenty-second values.

[0156] For example, there are no overlapping values ​​between any two values ​​in the eighth to tenth value ranges in Table 6.

[0157] Table 6

[0158] Optionally, the second information includes the model's format information. The value or range of the model's format information corresponds to a value of a second duration. It is understood that the second information includes at least one piece of information, each of which corresponds to a second duration; that is, the second information corresponds to at least one second duration. The "second duration" in "the value or range of the model's format information corresponds to a value of a second duration" refers to the second duration corresponding to the model's format information among the at least one second durations corresponding to the second information, and also refers to the second duration corresponding to the model's format information among the at least one second durations used to determine the first duration. For ease of description, this disclosure refers to the second duration corresponding to the model's format information as the second duration G.

[0159] For example, the values ​​of the model format information and the second duration G can be as shown in Table 7. For instance, when the model format information is ONNX, the second duration G is the twenty-third value. If the network device instructs the terminal device to use a model with an ONNX format via the first information, then the second duration G is the twenty-third value. The first duration can be determined based on the twenty-third value, the first time can be determined based on the first duration and the second time, and the first information can be sent to the terminal device at the first time. As another example, when the model format information is IR, the second duration G is the twenty-fourth value. If the network device instructs the terminal device to use a model with an IR format via the first information, then the second duration G is the twenty-fourth value. The first duration can be determined based on the twenty-fourth value, the first time can be determined based on the first duration and the second time, and the first information can be sent to the terminal device at the first time. This disclosure does not exemplify every item in Table 7, but is not limited to the examples given. It is understood that Table 7 is merely an example, and this disclosure does not limit the range of values ​​to be included in the format information of the model, nor does it limit the size of the twenty-third to twenty-fifth values.

[0160] Table 7

[0161] In some embodiments, the above optional examples can be combined arbitrarily. For example, if the second information includes the capability information of the terminal device and the type information of the model, the first duration can be determined based on the second duration A and the second duration B. If the value corresponding to the capability information of the terminal device is fixed as the first value, then when the model type is DNN, the value of the second duration B is the second value, and the first duration can be determined based on the first and second values. If the model type is CNN, the value of the second duration B is the third value, and the first duration can be determined based on the first and third values. For another example, if the second information includes the model's structural information, the number of model parameters, and the model's data type, the first duration can be determined based on the second duration C, the second duration D, and the second duration E. If the number of neurons in the model is within the first value range, and the number of model parameters is within the fourth value range, and the model's data type is FP32, then the value of the second duration C is the sixth value, the value of the second duration D is the ninth value, and the value of the second duration E is the thirteenth value, and the first duration is determined based on the sixth, ninth, and thirteenth values. If the number of neurons in the model is within the third range, the number of parameters in the model is within the sixth range, and the data type of the model is Int32, then the value of the second duration C is the eighth value, the value of the second duration D is the eleventh value, and the value of the second duration E is the seventeenth value. The first duration can be determined based on the eighth, eleventh, and seventeenth values. It is understood that this disclosure does not provide examples for all combinations, but is not limited to this.

[0162] In some embodiments, the first duration is equal to the sum of the second durations corresponding to each item in the second information. Taking Tables 2 to 7 above as examples, if the second information includes the capability information of the terminal device and the type information of the model, and the value corresponding to the capability information of the terminal device is fixed as the first value, and the type of the model is DNN, then the first duration can be equal to the sum of the first value and the second value. For another example, if the second information includes the structural information of the model, the number of parameters of the model, and the data type of the model, and the number of neurons in the model is within the first value range, the number of parameters of the model is within the fourth value range, and the data type of the model is FP32, then the first duration can be equal to the sum of the sixth value, the ninth value, and the thirteenth value. It is understood that this disclosure does not provide examples of all combinations, but is not limited to this.

[0163] In some embodiments, the first duration is determined based on the second information, which can be based on a third duration and a first scaling factor corresponding to each item in the second information. For example, each item in the second information corresponds to a first scaling factor, and the first duration can be determined based on these first scaling factors and the third duration. The third duration can be understood as a base value or a reference value.

[0164] In some embodiments, the third duration may be predefined by the protocol. Alternatively, the third duration may be sent by the terminal device to the network device. For example, a third duration may be fixed in the protocol, allowing the network device to determine the third duration from the protocol and determine the first duration based on the third duration and a first scaling factor. As another example, the third duration may be determined by the terminal device, as it serves as the base duration or reference duration for model reading and compilation. The terminal device may send the third duration to the network device, enabling the network device to determine the first duration based on the third duration and the first scaling factor, determine the first time based on the first duration and a second time point, and send first information to the terminal device at the first time point to instruct the terminal device to use the model.

[0165] In some embodiments, the value or range of a single piece of information in the second information corresponds to the value of a first scaling factor. For example, each piece of information corresponds to a first scaling factor, and the magnitude of the first scaling factor corresponding to each piece of information depends on the value or range of that information. That is, if the value or range of a certain piece of information is different, then the value of the first scaling factor corresponding to that piece of information is also different. The first scaling factor may also be referred to as a first weight, etc., and this disclosure does not limit the name of the first scaling factor.

[0166] It is understandable that the principle of "the value or range of a piece of information in the second information corresponds to the value of a first scaling factor" is similar to the principle of "the value or range of a piece of information in the second information corresponds to the value of a second duration". The following embodiments are only briefly described, and you can refer to the above embodiments for details.

[0167] Optionally, the second information includes the processing capability information of the terminal device. The value or range of the processing capability information of the terminal device corresponds to the value of a first scaling factor. For ease of description, this disclosure refers to the first scaling factor corresponding to the processing capability information of the terminal device as the first scaling factor A.

[0168] For example, for different models, the value corresponding to the processing capability information of the terminal device is always the same; that is, the processing capability information of the terminal device can correspond to a fixed first scaling factor A. For example, the first scaling factor A can be a twenty-sixth value. If the second information includes the processing capability information of the terminal device, then the first duration A can be determined based on the twenty-sixth value.

[0169] For example, for different models, the value or range of the processing capability information of the terminal device may be different. That is, if the value of the processing capability information of the terminal device is different, the value of the first scaling factor A may be different.

[0170] Optionally, the second information includes model type information. For ease of description, this disclosure refers to the first scaling factor corresponding to the model type information as the first scaling factor B.

[0171] For example, the values ​​of the model type information and the first scaling factor B can be as shown in Table 8. For instance, when the model type information is DNN, the first scaling factor B is the twenty-sixth value. If the network device indicates to the terminal device via the first information that the model used is DNN, then the first scaling factor B is the twenty-sixth value. The first duration can be determined based on the twenty-sixth value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment. This disclosure does not exemplify every item in Table 8, but is not limited to the examples given. It is understood that Table 8 is merely exemplary, and this disclosure does not limit which values ​​the model type information includes, nor does it limit the magnitude of the twenty-sixth to twenty-ninth values.

[0172] Table 8

[0173] Optionally, the second information includes the structural information of the model. For ease of description, this disclosure refers to the first scaling factor corresponding to the structural information of the model as the first scaling factor C.

[0174] For example, taking a model of type DNN and a model structure information of the number of neurons in a DNN as an example, the value of the first scaling factor B can be as shown in Table 9. For example, when the number of neurons is within the first range, the value of the first scaling factor C is the thirtieth value. If the network device instructs the terminal device to use a model of type DNN and the number of neurons is within the first range through the first information, then the value of the first scaling factor C is the thirtieth value. The first duration can be determined based on the thirtieth value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment.

[0175] Table 9

[0176] Optionally, the second information includes the model's parameter information, and the model's parameter information includes the number of model parameters. That is, the second information includes the number of model parameters. For ease of description, this disclosure refers to the first scaling factor corresponding to the number of model parameters as the first scaling factor D.

[0177] For example, the range of values ​​for the number of model parameters and the value of the first scaling factor D can be as shown in Table 10. For instance, when the range of values ​​for the number of model parameters is the fourth range, the value of the first scaling factor D is the thirty-third value. If the network device instructs the terminal device to use a model whose number of parameters is within the fourth range via the first information, then the value of the first scaling factor D is the thirty-third value. The first duration can be determined based on the thirty-third value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment.

[0178] Table 10

[0179] Optionally, the second information includes the model's parameter information, and the model's parameter information includes the model's data type. That is, the second information includes the model's data type. For ease of description, this disclosure refers to the first scaling factor corresponding to the model's data type as the first scaling factor E.

[0180] For example, the values ​​of the model's data type and the first scaling factor E can be as shown in Table 11. For instance, when the model's data type is FP32, the first scaling factor E is the thirty-seventh value. If the network device instructs the terminal device to use a model whose data type is FP32 via the first information, then the first scaling factor E is the thirty-seventh value. The first duration can be determined based on the thirty-seventh value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment.

[0181] Table 11

[0182] Optionally, the second information includes the model size information. For ease of description, this disclosure refers to the first scaling factor corresponding to the model size information as the first scaling factor F.

[0183] For example, the range of values ​​for the model size information and the value of the first scaling factor F can be as shown in Table 12. For instance, when the range of values ​​for the model size information is the eighth range, the value of the first scaling factor F is the forty-fourth value. If the network device instructs the terminal device to use a model whose size is within the eighth range via the first information, then the value of the first scaling factor F is the forty-fourth value. The first duration can be determined based on the forty-fourth value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment.

[0184] Table 12

[0185] Optionally, the second information includes the model's format information. For ease of description, this disclosure refers to the first scaling factor corresponding to the model's format information as the first scaling factor G.

[0186] For example, the values ​​of the model format information and the first scaling factor G can be as shown in Table 13. For instance, when the model format information is ONNX, the first scaling factor G is the forty-seventh value. If the network device instructs the terminal device to use a model with an ONNX format via the first information, then the first scaling factor G is the forty-seventh value. The first duration can be determined based on the forty-seventh value and the third duration, the first moment can be determined based on the first duration and the second moment, and the first information can be sent to the terminal device at the first moment.

[0187] Table 13

[0188] In some embodiments, the above-mentioned optional examples can be combined arbitrarily. For example, if the second information includes the capability information of the terminal device and the type information of the model, the first duration can be determined based on the third duration, the first scaling factor A, and the first scaling factor B. If the first scaling factor A is fixed at the fiftieth value, then when the model type is DNN, the first scaling factor B takes the twenty-sixth value, and the first duration can be determined based on the third duration, the fiftieth value, and the twenty-sixth value. If the model type is CNN, the first scaling factor B takes the twenty-seventh value, and the first duration can be determined based on the third duration, the fiftieth value, and the twenty-seventh value. As another example, if the second information includes the model's structural information, the number of model parameters, and the model's data type, then the first duration can be determined based on the third duration, the first scaling factor C, the first scaling factor D, and the first scaling factor E. If the number of neurons in the model is within the first range, the number of parameters in the model is within the fourth range, and the model data type is FP32, then the first scaling factor C is the thirtieth value, the first scaling factor D is the thirty-third value, and the first scaling factor E is the thirty-seventh value. The first duration is then determined based on the third duration, the thirtieth value, the thirty-third value, and the thirty-seventh value. It is understood that this disclosure does not provide examples for all combinations, but is not limited to this.

[0189] In some embodiments, the first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information. For example, the first duration is determined based on the third duration, the fiftieth value, and the twenty-sixth value, and can be equal to the product of the third duration, the fiftieth value, and the twenty-sixth value. Assuming the first duration is T1 and the third duration is T3, T1 = T3 × the fiftieth value × the twenty-sixth value. As another example, the first duration is determined based on the third duration, the thirtieth value, the thirty-third value, and the thirty-seventh value, and can be equal to the product of the third duration, the thirtieth value, the thirty-third value, and the thirty-seventh value. That is, T1 = T3 × the thirtieth value × the thirty-third value × the thirty-seventh value. This disclosure does not provide all examples, but is not limited to these.

[0190] In some embodiments, the first duration is determined based on the second information. Specifically, the first duration can be determined based on a fourth duration corresponding to a first portion of the second information and a second scaling factor corresponding to a second portion of the second information. The first or second portion of the information can be considered a subset of the second information; that is, the first portion of the information includes one or more pieces of information from the second information, and the second portion of the information includes one or more pieces of information from the second information other than the first portion. For example, each piece of information in the first portion corresponds to a fourth duration, and each piece of information in the second portion corresponds to a second scaling factor. The first duration can be determined based on these fourth durations and these second scaling factors.

[0191] In some embodiments, the first part of the information includes at least one of the following: processing capability information of the terminal device; format information of the model; and type information of the model, but is not limited thereto. It is understood that the first part of the information may include any one or more of the second part of the information, and this disclosure does not limit which information is included in the first part of the information.

[0192] In some embodiments, the second part of the information includes at least one of the following: parameter information of the model; size information of the model; and structural information of the model, but is not limited thereto. It is understood that the second part of the information may include any one or more of the information in the second part of the information, and this disclosure does not limit which information is included in the second part of the information.

[0193] For example, the first part of the information may include the processing capability information of the terminal device and the format information of the model. The processing capability information of the terminal device corresponds to the fourth duration A, and the format information of the model corresponds to the fourth duration G. The second part of the information may include the structural information of the model, the number of parameters of the model, and the data type of the model. The structural information of the model corresponds to the second scaling factor C, the number of parameters of the model corresponds to the second scaling factor D, and the data type of the model corresponds to the second scaling factor E. Then the first duration is determined based on the fourth duration A, the fourth duration G, the second scaling factor C, the second scaling factor D, and the second scaling factor E.

[0194] It is understood that there are any possible combinations of information in the first part and information in the second part. This disclosure will not provide examples of each combination, but is not limited to the examples provided.

[0195] It is understood that, similar to the second duration, this disclosure, for the purpose of distinction, refers to the duration corresponding to each item in the second information as the second duration, and the duration corresponding to the first part of the second information as the fourth duration. The second duration and the fourth duration can be the same; for example, the second duration corresponding to the first part of the information is its corresponding fourth duration. The method for determining the fourth duration can refer to the above embodiments, and the second duration in Tables 2 to 7 above can be replaced with the fourth duration.

[0196] It is understood that the second scaling factor is similar to the first scaling factor. For the purpose of distinction, this disclosure refers to a scaling factor corresponding to each item in the second information as the first scaling factor, and a scaling factor corresponding to the second part of the second information as the second first scaling factor. The first scaling factor and the second scaling factor can be the same; for example, the first scaling factor corresponding to the second part of the information is its corresponding second scaling factor. The method for determining the second scaling factor can refer to the above embodiments, and the first scaling factor in Tables 8 to 13 can be replaced with the second scaling factor.

[0197] In some embodiments, the value or range of a single piece of information in the first part of the information corresponds to a value of the fourth duration. For example, each piece of information in the first part of the information corresponds to a fourth duration, and the magnitude of the fourth duration corresponding to each piece of information depends on the value or range of that piece of information. That is, if the value or range of a certain piece of information in the first part of the information is different, then the fourth duration corresponding to that piece of information is different.

[0198] It is understood that the fourth duration is similar to the second duration. The principle that "the value or range of a piece of information in the first part of the information corresponds to a value of the fourth duration" is similar to the principle that "the value or range of a piece of information in the second part of the information corresponds to a value of the second duration". Please refer to the above embodiments. This disclosure will not repeat it again.

[0199] In some embodiments, the value or range of a single item in the second part of the information corresponds to the value of a second scaling factor. For example, each item in the second part of the information corresponds to a second scaling factor, and the magnitude of the second scaling factor corresponding to each item depends on the value or range of that item. That is, if the value or range of a certain item in the second part of the information is different, then the second scaling factor corresponding to that item is different.

[0200] It is understood that the second scaling factor is similar to the first scaling factor. The principle that "the value or range of a piece of information in the second part of the information corresponds to the value of the second scaling factor" is similar to the principle that "the value or range of a piece of information in the second part of the information corresponds to the value of the first scaling factor". Please refer to the above embodiments, and this disclosure will not repeat it.

[0201] In some embodiments, the first duration is determined based on the fourth duration corresponding to the first part of the information in the second information and the second scaling factor corresponding to the second part of the information in the second information. It can be that the first duration is equal to the sum of the fourth durations corresponding to each item in the first part of the information multiplied by the second scaling factor corresponding to each item in the second part of the information.

[0202] For example, the first duration is determined based on the fourth duration A, the fourth duration G, the second scaling factor C, the second scaling factor D, and the second scaling factor E. It can be that the first duration equals the sum of the fourth durations A and G multiplied by the second scaling factors C, D, and E. Assuming the first duration is T1, T1 = (fourth duration A + fourth duration G) × second scaling factor C × second scaling factor D × second scaling factor E. Taking Tables 2 to 13 above as examples, if the fourth duration corresponding to the terminal device's capability information is a fixed first value, and if the number of model parameters is within the fourth value range, then the fourth duration G is the ninth value. If the model structure is a DNN, and the number of neurons is within the first value range, then the second scaling factor C is the thirtieth value. If the number of model parameters is within the fifth value range, then the second scaling factor D is the thirty-fourth value. If the model's data type is TF32, then the second scaling factor E is the thirty-ninth value. Based on the above assumptions, the first duration is equal to the sum of the first and ninth values ​​multiplied by the thirtieth, thirty-fourth, and thirty-ninth values. That is, T1 = (first value + ninth value) × thirtieth value × thirty-fourth value × thirty-ninth value. The specific examples in this embodiment are merely illustrative, and this disclosure does not list all examples, but is not limited thereto.

[0203] In some embodiments, different pieces of the second information correspond to different first durations, and at least one piece of the different pieces of the second information has a different value or range. That is, the first duration is determined based on the second information, and there may be a one-to-one correspondence between the first duration and the second information. Different first durations correspond to different pieces of the second information, and the different pieces of the second information are reflected in the different values ​​or ranges of at least one piece of the second information.

[0204] In some embodiments, taking the second information as an example, which includes the terminal device's capability information, the model's type information, and the model's structure information, the second information is different if any one of these three pieces of information has a different value or range of values.

[0205] For example, suppose that the terminal device capability information (e.g., processor type) in second information 1 corresponds to CPU, the model type information is DNN, and the model structure information (e.g., number of neurons) is within a first value range. In second information 2, the terminal device capability information corresponds to CPU, the model type information is DNN, and the model structure information is within a second value range. Under this assumption, if one item in second information 1 and second information 2 has a different value range (i.e., the model structure information), then second information 1 and second information 2 are different, and their corresponding first durations are different. For example, second information 1 corresponds to first duration 1, and second information 2 corresponds to first duration 2. If the network device instructs the terminal device to use a model corresponding to second information 1 through the first information—for example, if the terminal device's processor type is CPU, the model type is DNN, and the number of neurons in the model is within a first value range—then the model corresponds to second information 1, and the first duration can be determined as first duration 1. Based on first duration 1 and the second time point, the first time point is determined, and the first information is sent to the terminal device at the first time point. If the network device instructs the terminal device to use a model corresponding to the second information 2 through the first information, for example, if the terminal device's processor type is CPU, the model type is DNN, and the number of neurons in the model is within the second value range, then the model corresponds to the second information 2, the first duration can be determined as the first duration 2, and the first time can be determined based on the first duration 2 and the second time, and the first information is sent to the terminal device at the first time.

[0206] For example, suppose that the terminal device capability information (e.g., processor type) in second information 1 corresponds to CPU, the model type information is DNN, and the model structure information (e.g., number of neurons) is within a first value range. In second information 2, the terminal device capability information corresponds to GPU, the model type information is DNN, and the model structure information is within a second value range. Under this assumption, two items in second information 1 and second information 2 have different value ranges: the terminal device capability information and the model structure information. Therefore, second information 1 and second information 2 are different, and their corresponding first durations are different. If the network device instructs the terminal device to use a model corresponding to second information 1 via first information, it can determine the first duration as first duration 1, and determine the first time based on first duration 1 and the second time point, sending the first information to the terminal device at the first time point. If the network device instructs the terminal device to use a model corresponding to second information 2 via first information, it can determine the first duration as first duration 2, and determine the first time based on first duration 2 and the second time point, sending the first information to the terminal device at the first time point.

[0207] For example, suppose that the terminal device capability information (e.g., processor type) in second information 1 corresponds to CPU, the model type information is DNN, and the model structure information (e.g., number of neurons) is within a first value range. In second information 2, the terminal device capability information corresponds to GPU, the model type information is CNN, and the model structure information is within a second value range. Under this assumption, each piece of information in second information 1 and second information 2 is different, therefore the first durations corresponding to second information 1 and second information 2 are different. If the network device instructs the terminal device to use a model corresponding to second information 1 through the first information, it can determine the first duration as first duration 1, and determine the first time based on first duration 1 and the second time point, sending the first information to the terminal device at the first time point. If the network device instructs the terminal device to use a model corresponding to second information 2 through the first information, it can determine the first duration as first duration 2, and determine the first time based on first duration 2 and the second time point, sending the first information to the terminal device at the first time point.

[0208] For example, assuming that the terminal device capability information (such as processor type) in the second information 1 and the second information 2 both correspond to CPU, the model type information both correspond to DNN, and the model structure information (such as the number of neurons) both correspond to the first value range, then the second information 1 and the second information 2 are the same second information and correspond to the same first duration.

[0209] In some embodiments, taking the second information including the processing capability information of the terminal device, the number of parameters of the model, and the data type of the model as an example, the correspondence between the second information and the first duration can be as shown in Table 14. The value corresponding to the processing capability information of the terminal device in different pieces of second information can be the same. Taking the processor type of the terminal device as an example, for different models, the processor of the terminal device is always a CPU. Therefore, the value corresponding to the processing capability information of the terminal device in different pieces of second information is the same. "--" indicates the value corresponding to the processing capability information of the terminal device.

[0210] For example, if the number of parameters in the model is within the first range and the model's data type is FP32, then the first duration is the fifty-first value. If the network device instructs the terminal device to use a model via the first information, and the number of parameters in the model is within the first range and the model's data type is FP32, then the first duration is the fifty-first value. The first time can be determined based on the fifty-first value and the second time, and the first information can be sent to the terminal device at the first time to instruct the terminal device to use the model.

[0211] For example, if the number of parameters in the model is within the first range and the data type of the model is FP16, then the first duration is the fifty-second value. If the network device instructs the terminal device to use a model through the first information, and the number of parameters in the model is within the first range and the data type of the model is FP16, then the first duration is the fifty-second value. The first time can be determined based on the fifty-second value and the second time, and the first information can be sent to the terminal device at the first time to instruct the terminal device to use the model.

[0212] For example, if the number of parameters in the model falls within the second range, and the data type of the model is FP16, then the first duration is the fifty-ninth value. If the network device instructs the terminal device to use a model via the first information, and the number of parameters in the model falls within the second range, and the data type of the model is FP16, then the first duration is the fifty-ninth value. The first time can be determined based on the fifty-ninth value and the second time, and the first information can be sent to the terminal device at the first time to instruct the terminal device to use the model. This disclosure does not provide examples for every item in Table 14, but is not limited to the examples given.

[0213] Table 14

[0214] It is understood that Table 14 is merely exemplary, and this disclosure does not limit what information the second information includes, nor does it limit the values ​​or ranges of values ​​for each piece of information. For example, if the second information includes model type information, model size information, and model format information, it can be shown in Table 15; if the second information includes the number of model parameters and model format information, it can be shown in Table 16. This disclosure is not limited to these.

[0215] Table 15

[0216] Table 16

[0217] In some embodiments, the terminal device may passively receive the first information, or it may determine the first moment and actively receive the first information at the first moment. Alternatively, the terminal device may determine the first moment, and if it does not receive the first information at the first moment, it may not use the model, or it may report the failure to receive the first information to the network device, etc. The terminal device may determine the first moment by having the network device indicate the first moment to the terminal device. Alternatively, the network device may indicate the second moment and the first duration to the terminal device, and the terminal device may determine the first moment based on the second moment and the first duration. Alternatively, the network device may indicate the second moment to the terminal device, and the terminal device may determine the first duration itself, and determine the first moment based on the first duration and the second moment. Alternatively, the terminal device may determine the second moment and the first duration itself, and determine the first moment based on the second moment and the first duration. The method by which the terminal device determines the first duration can refer to the method by which the network device determines the first duration, i.e., the above embodiments, and will not be repeated here.

[0218] In some embodiments, a network device may send multiple models to the same terminal device. For different models, the time at which the network device sends the first information (i.e., the first moment) may be the same or different.

[0219] For example, if a network device completes the transmission of multiple models simultaneously, the second time corresponding to the multiple models is the same; if the first time corresponding to the multiple models is also the same, the first time is the same.

[0220] For example, since multiple models are sent to the same terminal device, the processing capability information of the terminal devices corresponding to the multiple models is the same. If the first duration is determined based on the second information, and the second information only includes the processing capability information of the terminal device, then the first duration corresponding to the multiple models is the same. In this case, since the second time is the same, the first time is also the same. That is, the network device can send the first information at the same time to instruct the terminal device to use these multiple models.

[0221] For example, multiple models may be of the same type, meaning that the type information corresponding to the multiple models is the same. If the first duration is determined based on the second information, and the second information includes the model type information, or if the second information includes the processing capability information of the terminal device and the model type information, then the first duration corresponding to the multiple models is the same. In this case, since the second time is the same, the first time is also the same. That is, the network device can send the first information at the same time to instruct the terminal device to use these multiple models. Of course, it is also valid if the model type information is replaced by one or more of the model structure information, model parameter information, model size information, and model information, but this disclosure will not provide examples of each.

[0222] For example, if a network device completes the transmission of multiple models simultaneously, the second time points corresponding to the multiple models will be the same; if the first time points corresponding to the multiple models are different, the first time points will be different.

[0223] For example, different models may have different types, meaning that different models correspond to different model type information. If the first duration is determined based on the second information, and the second information includes the model type information, then different models will have different first durations. In this case, although the second time is the same, the first time is different. That is, the network device can send the first information for different models at different times to instruct the terminal device to use different models at different times. Of course, it is also valid to replace the model type information with one or more of the model structure information, model parameter information, model size information, and model input information; this disclosure will not provide examples of all of them.

[0224] For example, if the network device does not simultaneously complete the transmission of multiple models, the second time points corresponding to the multiple models will be different; if the first durations corresponding to the multiple models are the same, the first time points will be different. Exemplary embodiments with the same first duration can be found in the above embodiments, and will not be repeated here.

[0225] For example, if the network device does not simultaneously complete the transmission of multiple models, the second time points corresponding to the multiple models will be different. If the first durations corresponding to the multiple models are also different, the first time points may be the same or different. Exemplary embodiments with different first durations can be found in the above embodiments, and will not be repeated here.

[0226] In some embodiments, the units for the first duration, the second duration, and the third duration may be one of the following: seconds, milliseconds, microseconds, time slots, symbols, half-frames, frames, etc., which are not limited in this disclosure.

[0227] In some embodiments, the name of the first information is not limited, and it may be, for example, "instruction information", "activation information", etc.

[0228] Step S2103, terminal device 101 uses the model.

[0229] In some embodiments, after receiving the first information, the terminal device 101 uses the model based on the instructions of the first information. For example, the model can be used to perform inference, or to perform associated operations of the model.

[0230] The communication method involved in the embodiments of this disclosure may include at least one of steps S2101 to S2103. For example, step S2102 may be implemented as a separate embodiment, but is not limited thereto.

[0231] In some embodiments, step S2101 is optional and may be omitted or replaced in different embodiments.

[0232] In some embodiments, step S2103 is optional and may be omitted or replaced in different embodiments.

[0233] In some embodiments, other optional implementations described before or after the specification corresponding to FIG2 may be referred to.

[0234] Figure 3 is a flowchart illustrating a communication method according to an embodiment of the present disclosure. As shown in Figure 3, this embodiment of the present disclosure relates to a communication method executed by a terminal device 101, the method including:

[0235] Step S3101: Obtain the model.

[0236] The optional implementation of step S3101 can be found in the optional implementation of step S2101 in Figure 2, as well as other related parts in the embodiments involved in Figure 2, which will not be repeated here.

[0237] In some embodiments, terminal device 101 receives a model sent by network device 102, but is not limited thereto; it may also receive a model sent by other entities.

[0238] In some embodiments, terminal device 101 acquires the model defined by the protocol.

[0239] In some embodiments, the terminal device 101 obtains the model from the upper layer(s).

[0240] In some embodiments, the terminal device 101 processes the data to obtain the model.

[0241] In some embodiments, step S3101 is omitted, and the terminal device 101 autonomously implements the function indicated by the model, or the above function is default or default.

[0242] Step S3102: Obtain the first information.

[0243] The optional implementation of step S3102 can be found in the optional implementation of step S2102 in Figure 2, as well as other related parts in the embodiments involved in Figure 2, which will not be repeated here.

[0244] In some embodiments, terminal device 101 receives first information sent by network device 102, but is not limited thereto; it may also receive first information sent by other entities.

[0245] In some embodiments, the terminal device 101 obtains first information as defined by the protocol.

[0246] In some embodiments, the terminal device 101 obtains first information from the upper layer(s).

[0247] In some embodiments, the terminal device 101 processes the information to obtain the first information.

[0248] In some embodiments, step S3102 is omitted, and the terminal device 101 autonomously implements the function indicated by the first information, or the above function is the default or default.

[0249] Step S3103, use the model.

[0250] The optional implementation of step S3103 can be found in the optional implementation of step S2103 in Figure 2, as well as other related parts in the embodiments involved in Figure 2, which will not be repeated here.

[0251] The communication method involved in the embodiments of this disclosure may include at least one of steps S3101 to S3103. For example, step S3102 may be implemented as a separate embodiment, but is not limited thereto.

[0252] In some embodiments, step S3101 is optional and may be omitted or replaced in different embodiments.

[0253] In some embodiments, step S3103 is optional and may be omitted or replaced in different embodiments.

[0254] In some embodiments, other optional implementations may be described before or after the specification corresponding to FIG3.

[0255] Figure 4 is a flowchart illustrating a communication method according to an embodiment of the present disclosure. As shown in Figure 4, this embodiment of the present disclosure relates to a communication method executed by a network device 102, the method comprising:

[0256] Step S4101: Send the model.

[0257] The optional implementation of step S4101 can be found in the optional implementation of step S2101 in Figure 2, as well as other related parts in the embodiments involved in Figure 2, which will not be repeated here.

[0258] In some embodiments, network device 102 sends the model to terminal device 101, but is not limited thereto; it may also send the model to other entities.

[0259] Step S4102: Send the first message.

[0260] The optional implementation of step S4102 can be found in the optional implementation of step S2102 in Figure 2, as well as other related parts in the embodiments involved in Figure 2, which will not be repeated here.

[0261] In some embodiments, network device 102 sends first information to terminal device 101, but is not limited thereto; it may also send first information to other entities.

[0262] The communication method involved in the embodiments of this disclosure may include at least one of steps S4101 to S4102. For example, step S4102 may be implemented as a standalone embodiment, but is not limited thereto.

[0263] In some embodiments, step S4101 is optional and may be omitted or replaced in different embodiments.

[0264] In some embodiments, other optional implementations may be described before or after the specification corresponding to Figure 4.

[0265] Figure 5a is a schematic diagram of a communication method interaction according to an embodiment of the present disclosure. As shown in Figure 5a, the present disclosure relates to a communication method, which includes:

[0266] In step S5101, network device 102 sends first information to terminal device 101 at the first moment.

[0267] In some embodiments, the above methods may include the methods of the embodiments related to the communication system 100, terminal device 101, and network device 102, which will not be described again here.

[0268] This disclosure provides a model transfer method, as follows:

[0269] In some embodiments, the network device considers the terminal device ready to perform inference using the AI / ML model, starting from a first duration after completing the transmission of the AI / ML model to the terminal device; wherein the first duration is determined based on at least one of the following parameters:

[0270] The processing power of the terminal equipment;

[0271] Type information of AI / ML models;

[0272] Structural information of AI / ML models;

[0273] Parameter information of AI / ML models;

[0274] The number of parameters in an AI / ML model;

[0275] Data types of AI / ML models;

[0276] Size information of AI / ML models;

[0277] Format information of AI / ML models;

[0278] In some embodiments, "the first duration is determined according to at least one parameter" includes: the first duration is determined according to a second duration corresponding to at least one parameter, wherein each of the at least one parameter corresponds to a second duration.

[0279] Optionally, for any one of the at least one parameters, different values ​​or ranges of values ​​correspond to different values ​​of the second duration.

[0280] For example, the first duration is the sum of the second durations corresponding to each of the at least one parameter.

[0281] In some embodiments, "the first duration is determined according to at least one parameter" includes: the first duration is determined according to a third duration and a first scaling factor corresponding to at least one parameter, wherein each of the at least one parameters corresponds to a first scaling factor.

[0282] Optionally, for any one of the at least one parameters, different values ​​or ranges of values ​​correspond to different values ​​of the first scaling factor.

[0283] Optionally, the third duration is predefined by the protocol, or it is sent by the terminal device to the network device.

[0284] For example, the first duration is the product of the third duration and the first scaling factor corresponding to each of the at least one parameter.

[0285] In some embodiments, "the first duration is determined according to at least one parameter" includes: the first duration is determined according to a second duration and / or a first scaling factor corresponding to at least one parameter, wherein some of the parameters in the at least one parameter correspond to the second duration and other parameters correspond to the first scaling factor.

[0286] Optionally, for at least one parameter, there are parameters corresponding to a second duration, where each parameter corresponds to a second duration.

[0287] Optionally, for any parameter in at least one of the parameters corresponding to the second duration, different values ​​or ranges of values ​​correspond to different values ​​of the second duration.

[0288] Optionally, for at least one parameter, a portion of the parameters corresponding to the first scaling factor, wherein each parameter corresponds to a first scaling factor.

[0289] Optionally, for any parameter in at least one of the parameters corresponding to the first scaling factor, different values ​​or ranges of values ​​correspond to different values ​​of the first scaling factor.

[0290] For example, the first duration is the sum of the second durations corresponding to the partial parameters of the second duration in at least one parameter, and the product of the first scaling factor corresponding to the partial parameters of the first scaling factor in at least one parameter.

[0291] For example, the parameters corresponding to the second duration include: the processing power of the terminal device and / or the format information of the AI / ML model; some parameters corresponding to the first scaling factor include: the size information of the AI / ML model and / or the parameter information of the AI / ML model.

[0292] In some embodiments, "the first duration is determined according to at least one parameter" includes: different values ​​or combinations of value ranges of all parameters in the at least one parameter, corresponding to different values ​​of the first duration.

[0293] In some embodiments, FIG5b is a schematic diagram of a model transfer method according to an embodiment of the present disclosure. As shown in FIG5b, the model transfer method described above may include the following steps.

[0294] In step S5201, network device 102 sends the AI / ML model to terminal device 101.

[0295] In some embodiments, the terminal device receives an AI / ML model sent by the network device.

[0296] It should be understood that sending an AI / ML model from a network device to a terminal device includes sending information such as the AI / ML model type, structure, and / or parameters. Therefore, sending an AI / ML model from a network device to a terminal device can also be replaced by sending information such as the AI / ML model type, structure, and / or parameters; the technical essence remains the same.

[0297] In step S5202, starting from the first time interval after completing step S5101, network device 102 sends the first information to terminal device 101.

[0298] In some embodiments, the first information is used to instruct the terminal device to perform inference using an AI / ML model.

[0299] In some embodiments, terminal device 101 receives first information sent by network device.

[0300] In step S5203, the terminal device 101 performs inference using an AI / ML model based on the first information.

[0301] It should be understood that the network device considers the terminal device ready to perform inference using the AI / ML model starting from the first time interval after sending the AI / ML model to the terminal device. Therefore, in step 3, the network device will only instruct the terminal device to perform inference using the AI / ML model starting from the first time interval after completing step 1.

[0302] It should be understood that the first duration can be interpreted as the time required for the terminal device to prepare the AI / ML model for inference, that is, the time interval from the moment the network device completes sending the AI / ML model to the moment the terminal device is ready to use the AI / ML model for inference. The first duration can be reported by the terminal device to the network device, or it can be predefined by the protocol, etc., and this invention does not impose any limitations.

[0303] Optionally, the first duration is determined based on at least one of the following parameters:

[0304] Processing capacity of terminal devices:

[0305] It should be understood that the processing power of a terminal device is determined by multiple factors, such as: the type of processor (Central Processing Unit (CPU) or Graphics Processing Unit (GPU)), the number of processors, the processor's clock speed, the processor's computational bit width, the size of the memory, and the memory read / write frequency. These factors collectively determine the time required to read and compile AI / ML models.

[0306] AI / ML model type information:

[0307] It should be understood that the type information of an AI / ML model is also called the backbone of the AI / ML model; for example: Deep neural network (DNN), Convolutional neural network (CNN), Residual network (ResNet), Transformer, etc. Because different types of AI / ML models have different computational requirements, the time required for reading and compiling AI / ML models also varies.

[0308] Structural information of AI / ML models:

[0309] It should be understood that the structural information of an AI / ML model includes the modules that make up the model and the dimensions of each module. Different types of AI / ML models may have different structural information. For example, for a DNN, the structural information includes the number of fully connected layers and the number of neurons in each layer. For a CNN, the structural information includes the dimensions of the DNN's input layer, the number and dimensions of convolutional layers, pooling layers, fully connected layers, and the dimension of the output layer. For a Transformer, the structural information includes the dimensions of the Transformer's input layer, the number of encoders, the number of decoders, the dimensions of the self-attention layers and fully connected layers included in the encoder and decoder, and the dimension of the output layer. The structural information of an AI / ML model directly affects its compilation time. For example, the more complex the structure of an AI / ML model, the longer it takes to read and compile it; conversely, the simpler the structure of an AI / ML model, the shorter it takes to read and compile it.

[0310] Parameter information of AI / ML models:

[0311] It should be understood that the parameter information of an AI / ML model includes the number of parameters and the data type of the AI / ML model.

[0312] Number of parameters in AI / ML models:

[0313] The number of parameters in an AI / ML model directly affects the time required to read and compile the AI / ML model. For example, the more parameters an AI / ML model has, the larger its size, and the longer the reading and compilation time. Conversely, the fewer parameters an AI / ML model has, the smaller its size, and the shorter the reading and compilation time.

[0314] Data types of AI / ML models:

[0315] The data type in an AI / ML model refers to the data type of the AI / ML model parameters. Commonly used data types include FP32, FP16, TF32, BF16, Int32, Int16, and Int8. Information such as the number of bits, numerical range, and numerical precision for each data type is shown in Table 1. The quantization precision of the data type directly affects the AI / ML model reading and compilation time. For example, the higher the quantization precision, the longer the AI / ML model reading and compilation time; conversely, the lower the quantization precision, the shorter the AI / ML model reading and compilation time.

[0316] Size information of AI / ML models:

[0317] It should be understood that the size information of an AI / ML model refers to the total number of bits in the AI / ML model, or its storage size, or file size. The AI / ML model file is the carrier of the AI / ML model sent from the network device to the terminal device. The size of an AI / ML model is usually determined by both its structural and parameter information. For example, the more complex the structure, the more parameters, and the higher the quantization precision of the data type, the larger the AI / ML model will be, and the longer the reading and compilation time will be. Conversely, the simpler the structure, the fewer the parameters, and the lower the quantization precision of the data type, the smaller the AI / ML model will be, and the shorter the reading and compilation time will be.

[0318] AI / ML model format information:

[0319] It should be understood that the format of an AI / ML model refers to its storage format, or the format of the AI / ML model file. Currently, commonly used AI / ML model formats include Open Neural Network Exchange (ONNX) and Intermediate Representation (IR). It is also possible that 3GPP will define new AI / ML model formats in the future. Terminal devices optimize for different AI / ML model formats differently, which leads to variations in the time required for reading and compiling AI / ML models with different formats.

[0320] Optionally, "the first duration is determined based on at least one of the following parameters" includes several possible methods:

[0321] Method 1: The first duration is determined based on the second duration corresponding to at least one parameter, where each of the at least one parameter corresponds to a second duration.

[0322] Optionally, for any one of the at least one parameters, different values ​​or ranges of values ​​correspond to different values ​​for the second duration. Example:

[0323] The processing capacity of the terminal device: the corresponding second duration A = the first value. It should be understood that the processing capacity of the terminal device does not have a corresponding value or range of values, so its corresponding second duration has only one value.

[0324] AI / ML model type information: The corresponding second duration is the second duration B; the value of the second duration B is shown in Table 2.

[0325] The structural information of the AI / ML model: the corresponding second duration is the second duration C; taking DNN as an example, the value of the second duration C is shown in Table 3.

[0326] Among them, there are no overlapping values ​​between the first value range and the third value range.

[0327] The number of parameters in the AI / ML model: the corresponding second duration is the second duration D; the value of the second duration D is shown in Table 4.

[0328] Among them, there is no overlap between any two values ​​in the fourth to seventh value ranges.

[0329] Data type of AI / ML model: The corresponding second duration is the second duration E; the value of the second duration E is shown in Table 5.

[0330] The size information of the AI / ML model: the corresponding second duration is the second duration F; the value of the second duration F is shown in Table 6.

[0331] Among them, there is no overlap between any two values ​​in the range from the eighth to the tenth.

[0332] The format information of the AI / ML model: the corresponding second duration is the second duration G; the value of the second duration G is shown in Table 7.

[0333] For example, the first duration is the sum of the second durations corresponding to each of the at least one parameter. For instance, if the at least one parameter includes: the processing power of the terminal device, the number of parameters in the AI / ML model, and the data type of the AI / ML model, then the first duration = second duration A + second duration D + second duration E. As another example, if the at least one parameter includes: the processing power of the terminal device, the type information of the AI / ML model, and the size information of the AI / ML model, then the first duration T_1 = second duration A + second duration B + second duration F. As yet another example, if the at least one parameter includes: the processing power of the terminal device, the structural information of the AI / ML model, and the format information of the AI / ML model, then the first duration T_1 = second duration A + second duration C + second duration G.

[0334] Method 2: The first duration is determined based on the third duration and a first scaling factor corresponding to at least one parameter, wherein each of the at least one parameters corresponds to a first scaling factor.

[0335] Optionally, for any one of the at least one parameters, different values ​​or ranges of values ​​correspond to different values ​​of the first scaling factor.

[0336] For example:

[0337] The processing capacity of the terminal device: the corresponding second duration is the first scaling factor A; the first scaling factor A = the fiftieth value. It should be understood that the processing capacity of the terminal device does not have a corresponding value or range of values, so its corresponding second duration has only one value.

[0338] AI / ML model type information: The corresponding second duration is the first scaling factor B; the value of the first scaling factor B is shown in Table 8;

[0339] The structural information of the AI / ML model: the corresponding second duration is the first scaling factor C; taking DNN as an example, the value of the first scaling factor C is shown in Table 9.

[0340] Among them, there are no overlapping values ​​between the first value range and the third value range.

[0341] The number of parameters in the AI / ML model: the corresponding second duration is the first scaling factor D; the value of the first scaling factor D is shown in Table 10.

[0342] Among them, there is no overlap between any two values ​​in the fourth to seventh value ranges.

[0343] Data type of AI / ML model: The corresponding second duration is the first scaling factor E; the value of the first scaling factor E is shown in Table 11.

[0344] Size information for AI / ML models: The corresponding second duration is the first scaling factor F; the values ​​of the first scaling factor F are shown in Table 12. There are no overlapping values.

[0345] AI / ML model format information: The corresponding second duration is the first scaling factor G; the value of the first scaling factor G is shown in Table 13.

[0346] Optionally, the third duration is predefined by the protocol, or it is sent by the terminal device to the network device.

[0347] For example, the first duration can be T1, and the third duration can be T3.

[0348] For example, the first duration is the product of the third duration and the first scaling factor corresponding to each of the at least one parameter. For instance, if the at least one parameter includes: the processing power of the terminal device, the number of parameters in the AI / ML model, and the data type of the AI / ML model, then the first duration T1 = T3 × first scaling factor A × first scaling factor D × first scaling factor E. As another example, if the at least one parameter includes: the processing power of the terminal device, the type information of the AI / ML model, and the size information of the AI / ML model, then the first duration T1 = T3 × first scaling factor A × first scaling factor B × first scaling factor F. As yet another example, if the at least one parameter includes: the processing power of the terminal device, the structural information of the AI / ML model, and the format information of the AI / ML model, then the first duration T1 = T3 × first scaling factor A × first scaling factor C × first scaling factor G.

[0349] Method 3: The first duration is determined based on the second duration corresponding to at least one parameter and / or the first scaling factor, wherein some of the parameters in the at least one parameter correspond to the second duration and the other parameters correspond to the first scaling factor.

[0350] Optionally, for at least one parameter, there are parameters corresponding to a second duration, where each parameter corresponds to a second duration.

[0351] Optionally, for any parameter in at least one of the parameters corresponding to the second duration, different values ​​or ranges of values ​​correspond to different values ​​of the second duration. For example, it is consistent with the example in Method 1.

[0352] Optionally, for at least one parameter, a portion of the parameters corresponding to the first scaling factor, wherein each parameter corresponds to a first scaling factor.

[0353] Optionally, for any parameter in at least one of the parameters corresponding to the first scaling factor, different values ​​or ranges of values ​​correspond to different values ​​of the first scaling factor. For example, this is consistent with the example in Method 2.

[0354] For example, the parameters corresponding to the second duration include: the processing power of the terminal device and / or the format information of the AI / ML model; some parameters corresponding to the first scaling factor include: the size information of the AI / ML model and / or the parameter information of the AI / ML model.

[0355] For example, the first duration is the sum of the second durations corresponding to the partial parameters of the second duration in at least one parameter, multiplied by the first scaling factor corresponding to the partial parameters of the first scaling factor in at least one parameter. For instance, if the at least one parameter includes: the processing power of the terminal device, the number of parameters of the AI / ML model, and the data type of the AI / ML model, where the processing power of the terminal device corresponds to the second duration, and the number of parameters and the data type of the AI / ML model correspond to the first scaling factor, then the first duration T1 = second duration A × first scaling factor D × first scaling factor E. As another example, if the at least one parameter includes: the processing power of the terminal device, the type information of the AI / ML model, and the size information of the AI / ML model, where the processing power of the terminal device corresponds to the second duration, and the type information and the size information of the AI / ML model correspond to the first scaling factor, then the first duration T1 = second duration A × first scaling factor B × first scaling factor F. For example, if at least one parameter includes: the processing power of the terminal device, the structural information of the AI / ML model, and the format information of the AI / ML model, wherein the processing power of the terminal device and the format information of the AI / ML model correspond to the second duration, and the structural information of the AI / ML model corresponds to the first scaling factor, then the first duration T1 = (second duration A + second duration G) × first scaling factor C.

[0356] Method 4: "The first duration is determined based on at least one parameter" includes: different values ​​or combinations of value ranges of all parameters in the at least one parameter, corresponding to different values ​​of the first duration. For example, assuming that the at least one parameter includes: the processing power of the terminal device, the number of parameters of the AI / ML model, and the data type of the AI / ML model, then the first duration is as shown in Table 14.

[0357] Optionally, the units for the first duration, the second duration, and the third duration can be one of the following: seconds, milliseconds, microseconds, time slots, symbols, half-frames, frames, etc.

[0358] This disclosure also provides an apparatus for implementing any of the above methods. For example, an apparatus is provided that includes units or modules for implementing the steps performed by the terminal device in any of the above methods. Alternatively, another apparatus is provided that includes units or modules for implementing the steps performed by the network device (e.g., access network device, core network functional node, core network device, etc.) in any of the above methods.

[0359] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC). The functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD). Taking a field-programmable gate array (FPGA) as an example, it can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0360] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), or a Deep Learning Processing Unit (DPU).

[0361] Figure 6a is a schematic diagram of the structure of the terminal device proposed in an embodiment of this disclosure. As shown in Figure 6a, the terminal device 6100 may include at least one of a transceiver module 6101 and a processing module 6102. The transceiver module 6101 is used to receive first information sent by the network device. The first information is sent by the network device at a first moment and is used to instruct the terminal device to use a model. The time interval between the first moment and a second moment is greater than or equal to a first duration, and the second moment is the moment when the network device completes sending the model to the terminal device.

[0362] In some embodiments, the first duration is determined based on second information, which includes at least one of the following: processing capability information of the terminal device; model type information; model structure information; model parameter information; model size information; and model format information.

[0363] In some embodiments, the parameter information of the model includes at least one of the following: the number of parameters of the model; the data type of the model.

[0364] In some embodiments, the first duration is determined based on the second duration corresponding to each item in the second information.

[0365] In some embodiments, the value or range of a piece of information in the second information corresponds to a value of a second duration.

[0366] In some embodiments, the first duration is equal to the sum of the second durations corresponding to each item in the second information.

[0367] In some embodiments, the first duration is determined based on a first scaling factor corresponding to each item in the third duration and the second information.

[0368] In some embodiments, the third duration is predefined by the protocol, or the third duration is sent by the terminal device to the network device.

[0369] In some embodiments, the value or range of a piece of information in the second information corresponds to the value of a first scaling factor.

[0370] In some embodiments, the first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information.

[0371] In some embodiments, the first duration is determined based on the fourth duration corresponding to the first part of the second information and the second scaling factor corresponding to the second part of the second information.

[0372] In some embodiments, the value or range of a piece of information in the first part of the information corresponds to a value of a fourth duration.

[0373] In some embodiments, the value or range of a piece of information in the second part of the information corresponds to the value of a second scaling factor.

[0374] In some embodiments, the first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

[0375] In some embodiments, the first part of the information includes at least one of the following: processing capability information of the terminal device; format information of the model; and type information of the model.

[0376] In some embodiments, the second part of the information includes at least one of the following: model parameter information; model size information; model structure information.

[0377] In some embodiments, different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

[0378] In some embodiments, the model includes at least one of the following: an artificial intelligence (AI) model; or a machine learning (ML) model.

[0379] Figure 6b is a schematic diagram of the network device proposed in an embodiment of this disclosure. As shown in Figure 6b, the network device 6200 may include at least one of a transceiver module 6201 and a processing module 6202. The transceiver module 6201 is used to send first information to a terminal device at a first moment, the first information being used to instruct the terminal device to use a model. The time interval between the first moment and the second moment is greater than or equal to a first duration, and the second moment is the moment when the network device completes sending the model to the terminal device.

[0380] In some embodiments, the first duration is determined based on second information, which includes at least one of the following: processing capability information of the terminal device; model type information; model structure information; model parameter information; model size information; and model format information.

[0381] In some embodiments, the parameter information of the model includes at least one of the following: the number of parameters of the model; the data type of the model.

[0382] In some embodiments, the first duration is determined based on the second duration corresponding to each item in the second information.

[0383] In some embodiments, the value or range of a piece of information in the second information corresponds to a second duration.

[0384] In some embodiments, the first duration is equal to the sum of the second durations corresponding to each item in the second information.

[0385] In some embodiments, the first duration is determined based on a first scaling factor corresponding to each item in the third duration and the second information.

[0386] In some embodiments, the third duration is predefined by the protocol, or the third duration is sent by the terminal device to the network device.

[0387] In some embodiments, the value or range of a piece of information in the second information corresponds to a first scaling factor.

[0388] In some embodiments, the first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information.

[0389] In some embodiments, the first duration is determined based on the fourth duration corresponding to the first part of the second information and the second scaling factor corresponding to the second part of the second information.

[0390] In some embodiments, the value or range of a piece of information in the first part of the information corresponds to a fourth duration.

[0391] In some embodiments, the value or range of a piece of information in the second part of the information corresponds to a second scaling factor.

[0392] In some embodiments, the first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

[0393] In some embodiments, the first part of the information includes at least one of the following: processing capability information of the terminal device; format information of the model; and type information of the model.

[0394] In some embodiments, the second part of the information includes at least one of the following: model parameter information; model size information; model structure information.

[0395] In some embodiments, different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

[0396] In some embodiments, the model includes at least one of the following: an artificial intelligence (AI) model; or a machine learning (ML) model.

[0397] Figure 7a is a schematic diagram of a communication device according to an embodiment of this disclosure. The communication device 7100 can be a network device, a terminal device, or a chip, chip system, or processor that supports the implementation of any of the above methods in a network device, or a chip, chip system, or processor that supports the implementation of any of the above methods in a terminal device. Optionally, the network device can be an access network device, a core network device, etc. Optionally, the terminal device can be a user equipment, etc. The communication device 7100 can be used to implement the methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0398] As shown in Figure 7a, the communication device 7100 includes one or more processors 7101. The processor 7101 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control the communication device, execute programs, and process program data. The communication device 7100 is used to execute any of the above methods. Optionally, the communication device can be a base station, a baseband chip, a terminal device, a terminal device chip, a DU (Distributed Unit), or a CU (Computer Integrated Circuit), etc.

[0399] In some embodiments, the communication device 7100 further includes one or more memories 7102 for storing instructions. Optionally, all or part of the memories 7102 may also be located outside the communication device 7100.

[0400] In some embodiments, the communication device 7100 further includes one or more transceivers 7103. When the communication device 7100 includes one or more transceivers 7103, the transceivers 7103 perform communication steps such as sending and / or receiving in the above method, such as steps S2101 and S2102, and the processor 7101 performs other steps.

[0401] In some embodiments, a transceiver may include a receiver and / or a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, etc., may be used interchangeably; the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc., may be used interchangeably; and the terms receiver, receiving unit, receiver, receiving circuit, etc., may be used interchangeably.

[0402] In some embodiments, the communication device 7100 may include one or more interface circuits 7104. Optionally, the interface circuit 7104 is connected to the memory 7102, and the interface circuit 7104 can be used to receive signals from the memory 7102 or other devices, and can be used to send signals to the memory 7102 or other devices. For example, the interface circuit 7104 can read instructions stored in the memory 7102 and send the instructions to the processor 7101.

[0403] The communication device 7100 described in the above embodiments may be a network device or a terminal device, but the scope of the communication device 7100 described in this disclosure is not limited thereto, and the structure of the communication device 7100 may not be limited by FIG. 7a. The communication device may be a standalone device or a part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data and programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0404] Figure 7b is a schematic diagram of the chip structure proposed in an embodiment of this disclosure. For cases where the communication device 7100 can be a chip or a chip system, please refer to the schematic diagram of the chip 7200 shown in Figure 7b, but it is not limited thereto.

[0405] Chip 7200 includes one or more processors 7201, which are used to perform any of the above methods.

[0406] In some embodiments, chip 7200 further includes one or more interface circuits 7202. Optionally, the interface circuit 7202 is connected to memory 7203, and the interface circuit 7202 can be used to receive signals from memory 7203 or other devices, and the interface circuit 7202 can be used to send signals to memory 7203 or other devices. For example, the interface circuit 7202 can read instructions stored in memory 7203 and send the instructions to processor 7201.

[0407] In some embodiments, the interface circuit 7202 performs communication steps such as sending and / or receiving in the above method, such as steps S2101 and S2102, while the processor 7201 performs other steps.

[0408] In some embodiments, the terms interface circuit, interface, transceiver pin, transceiver, etc., can be used interchangeably.

[0409] In some embodiments, chip 7200 further includes one or more memories 7203 for storing instructions. Optionally, all or part of the memories 7203 may be located outside of chip 7200.

[0410] This disclosure also proposes a storage medium storing instructions that, when executed on the communication device 7100, cause the communication device 7100 to perform any of the above methods. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but not limited thereto; it may also be a storage medium readable by other devices. Optionally, the storage medium may be a non-transitory storage medium, but not limited thereto; it may also be a temporary storage medium.

[0411] This disclosure also provides a program product that, when executed by the communication device 7100, causes the communication device 7100 to perform any of the above methods. Optionally, the program product is a computer program product.

[0412] This disclosure also proposes a computer program that, when run on a computer, causes the computer to perform any of the above methods.

Claims

1. A communication method, characterized in that, The method includes: The network device sends first information to the terminal device at a first moment. The first information is used to instruct the terminal device to use the model. The time interval between the first moment and the second moment is greater than or equal to a first duration. The second moment is the moment when the network device completes sending the model to the terminal device.

2. The method according to claim 1, characterized in that, The first duration is determined based on second information, which includes at least one of the following: The processing capability information of the terminal device; The type information of the model; The structural information of the model; The parameter information of the model; The size information of the model; The format information of the model.

3. The method according to claim 2, characterized in that, The parameter information of the model includes at least one of the following: The number of parameters in the model; The data type of the model.

4. The method according to any one of claims 2-3, characterized in that, The first duration is determined based on the second duration corresponding to each item in the second information.

5. The method according to claim 4, characterized in that, The value or range of values ​​of one item in the second information corresponds to a value of the second duration.

6. The method according to claim 5, characterized in that, The first duration is equal to the sum of the second durations corresponding to each item in the second information.

7. The method according to any one of claims 2-3, characterized in that, The first duration is determined based on the third duration and the first scaling factor corresponding to each item in the second information.

8. The method according to claim 7, characterized in that, The third duration is either predefined by the protocol or it is sent from the terminal device to the network device.

9. The method according to any one of claims 7-8, characterized in that, The value or range of values ​​of one item in the second information corresponds to the value of one of the first scaling factors.

10. The method according to any one of claims 7-9, characterized in that, The first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information.

11. The method according to any one of claims 2-3, characterized in that, The first duration is determined based on the fourth duration corresponding to the first part of the second information and the second scaling factor corresponding to the second part of the second information.

12. The method according to claim 11, characterized in that, The value or range of a single piece of information in the first part of the information corresponds to a value of the fourth duration.

13. The method according to any one of claims 11-12, characterized in that, The value or range of values ​​of one item in the second part of the information corresponds to the value of one of the second scaling factors.

14. The method according to any one of claims 11-13, characterized in that, The first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

15. The method according to any one of claims 11-14, characterized in that, The first part of the information includes at least one of the following: The processing capability information of the terminal device; The format information of the model; The type information of the model.

16. The method according to any one of claims 11-15, characterized in that, The second part of the information includes at least one of the following: The parameter information of the model; The size information of the model; The structural information of the model.

17. The method according to any one of claims 2-3, characterized in that, Different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

18. The method according to any one of claims 1-17, characterized in that, The model includes at least one of the following: Artificial intelligence (AI) models; Machine learning (ML) models.

19. A communication method, characterized in that, The method includes: The terminal device receives first information sent by the network device. The first information is sent by the network device at a first moment. The first information is used to instruct the terminal device to use the model. The time interval between the first moment and the second moment is greater than or equal to a first duration. The second moment is the moment when the network device completes sending the model to the terminal device.

20. The method according to claim 19, characterized in that, The first duration is determined based on second information, which includes at least one of the following: The processing capability information of the terminal device; The type information of the model; The structural information of the model; The parameter information of the model; The size information of the model; The format information of the model.

21. The method according to claim 20, characterized in that, The parameter information of the model includes at least one of the following: The number of parameters in the model; The data type of the model.

22. The method according to any one of claims 20-21, characterized in that, The first duration is determined based on the second duration corresponding to each item in the second information.

23. The method according to claim 22, characterized in that, The value or range of values ​​of one item in the second information corresponds to a value of the second duration.

24. The method according to claim 23, characterized in that, The first duration is equal to the sum of the second durations corresponding to each item in the second information.

25. The method according to any one of claims 20-21, characterized in that, The first duration is determined based on the third duration and the first scaling factor corresponding to each item in the second information.

26. The method according to claim 25, characterized in that, The third duration is either predefined by the protocol or it is sent from the terminal device to the network device.

27. The method according to any one of claims 25-26, characterized in that, The value or range of values ​​of one item in the second information corresponds to the value of one of the first scaling factors.

28. The method according to any one of claims 25-27, characterized in that, The first duration is equal to the third duration multiplied by the first scaling factor corresponding to each item in the second information.

29. The method according to any one of claims 20-21, characterized in that, The first duration is determined based on the fourth duration corresponding to the first part of the second information and the second scaling factor corresponding to the second part of the second information.

30. The method according to claim 29, characterized in that, The value or range of a single piece of information in the first part of the information corresponds to a value of the fourth duration.

31. The method according to any one of claims 29-30, characterized in that, The value or range of values ​​of one item in the second part of the information corresponds to the value of one of the second scaling factors.

32. The method according to any one of claims 29-31, characterized in that, The first duration is equal to the sum of the fourth durations corresponding to each item in the first part of information multiplied by the second scaling factor corresponding to each item in the second part of information.

33. The method according to any one of claims 29-32, characterized in that, The first part of the information includes at least one of the following: The processing capability information of the terminal device; The format information of the model; The type information of the model.

34. The method according to any one of claims 29-33, characterized in that, The second part of the information includes at least one of the following: The parameter information of the model; The size information of the model; The structural information of the model.

35. The method according to any one of claims 20-21, characterized in that, Different second information corresponds to different first durations, and at least one of the different second information has a different value or value range.

36. The method according to any one of claims 19-35, characterized in that, The model includes at least one of the following: Artificial intelligence (AI) models; Machine learning (ML) models.

37. A network device, characterized in that, include: The transceiver module is used to send first information to the terminal device at a first moment. The first information is used to instruct the terminal device to use the model. The time interval between the first moment and the second moment is greater than or equal to a first duration. The second moment is the moment when the network device completes sending the model to the terminal device.

38. A terminal device, characterized in that, include: The transceiver module is used to receive first information sent by the network device at a first moment. The first information is used to instruct the terminal device to use the model. The time interval between the first moment and the second moment is greater than or equal to a first duration. The second moment is the moment when the network device completes sending the model to the terminal device.

39. A network device, characterized in that, include: One or more processors; The processor is used to execute the communication method according to any one of claims 1-18.

40. A terminal device, characterized in that, include: One or more processors; The processor is used to execute the communication method according to any one of claims 19-36.

41. A communication system, characterized in that, include: A network device and a terminal device, wherein the network device is configured to implement the communication method of any one of claims 1-18, and the terminal device is configured to implement the communication method of any one of claims 19-36.

42. A storage medium, characterized in that, include: The storage medium stores instructions that, when executed on a communication device, cause the communication device to perform the communication method as described in any one of claims 1-18 or 19-36.

43. A program product, characterized in that, include: A computer program, when executed by a communication device, causes the communication device to perform the communication method as described in any one of claims 1-18 or 19-36.

Images