WO2024000532A1 - 一种ai模型的传输方法及其装置 - Google Patents

一种ai模型的传输方法及其装置 Download PDF

Info

Publication number
WO2024000532A1
WO2024000532A1 PCT/CN2022/103168 CN2022103168W WO2024000532A1 WO 2024000532 A1 WO2024000532 A1 WO 2024000532A1 CN 2022103168 W CN2022103168 W CN 2022103168W WO 2024000532 A1 WO2024000532 A1 WO 2024000532A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
data type
receiving node
node
bit stream
Prior art date
Application number
PCT/CN2022/103168
Other languages
English (en)
French (fr)
Inventor
牟勤
Original Assignee
北京小米移动软件有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京小米移动软件有限公司 filed Critical 北京小米移动软件有限公司
Priority to CN202280002427.8A priority Critical patent/CN118202645A/zh
Priority to PCT/CN2022/103168 priority patent/WO2024000532A1/zh
Publication of WO2024000532A1 publication Critical patent/WO2024000532A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals

Definitions

  • the present application relates to the field of communication technology, and in particular, to an AI model transmission method and its device.
  • AI Artificial Intelligence
  • AI Artificial Intelligence
  • the first stage is the training stage of the model, that is, the stage of obtaining the model; the second stage is the deployment stage of the model, that is, the inference application stage of the model.
  • the model will be transferred from the training node to the inference node.
  • the AI model needs to be digitized at this time, that is, the model Structure, model parameters, etc. are expressed numerically.
  • terminal device types support different data types. For example, some terminal devices only support 8-bit integers, and some processors support 16-bit floating point types. Different data types lead to different reasoning complexities.
  • Embodiments of the present disclosure provide an AI model transmission method and its device, which can be applied in fields such as artificial intelligence (Artificial Intelligence, AI), and provide a flexible AI model representation method that digitizes the AI model according to the data type. Representation can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • AI Artificial Intelligence
  • embodiments of the present disclosure provide a method for transmitting an AI model.
  • the method is executed by a provisioning node.
  • the method includes:
  • the node determines the data type for digital representation of the AI model, it converts the AI model into the corresponding AI model bit stream according to the data type, so that the AI model interactors can synchronize the AI model transmission time.
  • the data type used can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • determining the data type for digital representation of the AI model includes:
  • the data type for digitally representing the AI model is determined based on the data type support capability of the AI model receiving node.
  • it also includes:
  • determining the data type for digital representation of the AI model according to the data type support capability of the AI model receiving node includes:
  • the data type corresponding to the maximum support capability data type of the AI model receiving node is determined as the data type for digital representation of the AI model.
  • determining the data type for digital representation of the AI model according to the data type support capability of the AI model receiving node includes:
  • the AI model receiving node select a data type supported by the AI model providing node as the data type for digital representation of the AI model.
  • determining the data type of the AI model receiving node includes:
  • the data type for digital representation of the AI model is determined.
  • determining the data type of the AI model receiving node includes:
  • determining the data type of the AI model receiving node includes:
  • the data type represented by the AI model is determined.
  • it also includes:
  • Instruction information of a data type for digital representation of the AI model is sent to the AI model receiving node, where the instruction information includes a data type for digital representation of the AI model.
  • an AI model transmission method is also provided.
  • the method is executed by the receiving node.
  • the method includes:
  • the AI model bit stream In response to receiving the AI model bit stream sent by the AI model providing node, the AI model bit stream is reversely converted according to predetermined conversion rules to obtain the corresponding AI model, and the AI model is used at the AI receiving node.
  • the receiving node when the receiving node responds to receiving the AI model bit stream sent by the AI model providing node, it reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses it at the AI receiving node.
  • the AI model realizes the data type used when the AI model interactor synchronizes the AI model transmission, which can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • it also includes:
  • the instruction information of the data type that digitally represents the AI model is sent by the receiving AI model provision node, and the instruction information includes the data type that is digitally represented by the AI model.
  • the step of reversely converting the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model includes:
  • the AI model bit stream is reverse-converted according to the predetermined conversion rules to obtain the corresponding AI model.
  • the method further includes:
  • the method further includes:
  • Node reports are provided to the AI model based on power consumption and/or storage capabilities.
  • an AI model transmission device is also provided.
  • the device is provided at a providing node, and the device includes:
  • Determination unit used to determine the data type for digital representation of the AI model
  • a conversion unit configured to convert the AI model into a corresponding AI model bit stream according to the data type.
  • the determining unit is further configured to determine the data type for digitally representing the AI model based on the data type support capability of the AI model receiving node.
  • it also includes:
  • a receiving unit configured to receive the data type support capability of the AI model receiving node sent by the AI model receiving node.
  • the determining unit is further configured to determine the data type corresponding to the maximum support capability data type of the AI model receiving node according to the supporting capability of the data type of the AI model receiving node as the basis for the AI model.
  • a numerically represented data type is further configured to determine the data type corresponding to the maximum support capability data type of the AI model receiving node according to the supporting capability of the data type of the AI model receiving node as the basis for the AI model.
  • the determining unit is further configured to select a data type supported by the AI model providing node as a data type for digital representation of the AI model based on the data type support capability of the AI model receiving node.
  • the determining unit is further configured to determine the data type for digitally representing the AI model based on the power consumption and/or storage capacity reported by the AI model receiving node.
  • the determining unit is also used to determine the data type for digital representation of the AI model according to business requirements.
  • the determining unit is further configured to determine the data type represented by the AI model according to the resource overhead of the AI model.
  • it also includes:
  • a sending unit configured to send indication information of a data type that digitally represents the AI model to the AI model receiving node, where the indication information includes a data type that digitally represents the AI model.
  • an AI model transmission device is also provided.
  • the device is provided at a receiving node, and the device includes:
  • a conversion unit configured to, in response to receiving the AI model bit stream sent by the AI model providing node, reversely convert the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model;
  • the use unit is used for using the AI model at the AI receiving node.
  • it also includes:
  • the receiving unit is configured to receive indication information of a data type that digitally represents the AI model sent by the AI model providing node, where the indication information includes a data type that digitally represents the AI model.
  • the conversion unit is also configured to respond to the instruction information and reverse the AI model bit stream according to the predetermined conversion rules according to the data type included in the instruction information for digital representation of the AI model. to obtain the corresponding AI model.
  • the device further includes:
  • the first reporting unit is configured to provide the AI model with a node's support capability for reporting the data type.
  • the device further includes:
  • the second reporting unit is used to provide node reporting based on power consumption and/or storage capabilities to the AI model.
  • inventions of the present disclosure provide an AI model transmission device.
  • the device includes a processor and a memory.
  • a computer program is stored in the memory.
  • the processor executes the computer program stored in the memory to The device is caused to perform the method described in the first aspect or the second aspect.
  • an AI model transmission device including: a processor and an interface circuit
  • the interface circuit is used to receive code instructions and transmit them to the processor
  • the processor is configured to run the code instructions to perform the method described in the first aspect or the second aspect.
  • embodiments of the present disclosure provide a computer-readable storage medium for storing instructions that, when executed, enable the method described in the first or second aspect to be implemented.
  • Figure 1 is a schematic flowchart of an AI model transmission method provided by an embodiment of the present disclosure
  • FIG. 2 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • FIG. 3 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 4 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • FIG. 5 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 6 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 7 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 8 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 9 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 10 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 11 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 12 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 13 is a schematic flowchart of another AI model transmission method provided by an embodiment of the present disclosure.
  • Figure 14 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • Figure 15 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • Figure 16 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • Figure 17 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • first, second, third, etc. may be used to describe various information in the embodiments of the present disclosure, the information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other.
  • first information may also be called second information, and similarly, the second information may also be called first information.
  • the words "if” and “if” as used herein may be interpreted as “when” or “when” or “in response to determining.”
  • first, second, third, etc. may be used to describe various information in the embodiments of the present disclosure, the information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other.
  • first information may also be called second information, and similarly, the second information may also be called first information.
  • the words "if” and “if” as used herein may be interpreted as “when” or “when” or “in response to determining.”
  • Figure 1 is a schematic architectural diagram of a communication system provided by an embodiment of the present application.
  • the communication system may include but is not limited to a network device and a terminal device, where the terminal device communicates with the network device.
  • the number and form of devices shown in Figure 1 are only for examples and do not constitute a limitation on the embodiments of the present application.
  • the application may include two or more network devices and two or more terminal devices.
  • the communication system shown in Figure 1 includes a network device 101 and a terminal device 102.
  • a terminal device can serve as both an AI model providing node and an AI model receiving node.
  • a network device can serve as both an AI model providing node and an AI model receiving node.
  • LTE long term evolution
  • 5th generation fifth generation
  • 5G new radio (NR) system 5th generation new radio
  • the network device in the embodiment of the present disclosure is an entity on the network side that is used to transmit or receive signals.
  • the network device 101 can be an evolved base station (evolved NodeB, eNB), a transmission reception point (transmission reception point or transmit receive point, TRP), a next generation base station (next generation NodeB, gNB) in an NR system, or other future mobile Base stations in communication systems or access nodes in wireless fidelity (WiFi) systems, etc.
  • the embodiments of the present disclosure do not limit the specific technologies and specific equipment forms used by network equipment.
  • the network equipment provided by the embodiments of the present disclosure may be composed of a centralized unit (CU) and a distributed unit (DU).
  • the CU may also be called a control unit (control unit).
  • the structure can separate the protocol layers of network equipment, such as base stations, and place some protocol layer functions under centralized control on the CU. The remaining part or all protocol layer functions are distributed in the DU, and the CU centrally controls the DU.
  • the terminal device in the embodiment of the present disclosure is an entity on the user side that is used to receive or transmit signals, such as a mobile phone.
  • Terminal equipment can also be called terminal equipment (terminal), mobile station (mobile station, MS), mobile terminal equipment (mobile terminal, MT), etc.
  • the terminal device can be a car with communication functions, a smart car, a mobile phone, a wearable device, a tablet computer (Pad), a computer with wireless transceiver functions, a virtual reality (VR) terminal device, an augmented reality (augmented reality (AR) terminal equipment, wireless terminal equipment in industrial control, wireless terminal equipment in self-driving, wireless terminal equipment in remote medical surgery, smart grid ( Wireless terminal equipment in smart grid, wireless terminal equipment in transportation safety, wireless terminal equipment in smart city, wireless terminal equipment in smart home, etc.
  • the embodiments of the present disclosure do not limit the specific technology and specific equipment form used by the terminal equipment.
  • terminal device types support different data types. For example, some terminal devices only support 8-bit integers, and some processors support 16-bit floating point types. Different data types lead to different reasoning complexities.
  • embodiments of the present disclosure propose an AI model transmission method and a device thereof.
  • Figure 2 is a schematic flowchart of a data type transmission method provided by an embodiment of the present disclosure. As shown in Figure 2, this method is applied to provide nodes. This method may include but is not limited to the following steps:
  • Step S201 Determine the data type for digital representation of the AI model.
  • the artificial intelligence (Artificial Intelligence, AI) model involves two stages: stage one: the training stage of the AI model, that is, the stage of obtaining the AI model; stage two: the deployment stage of the AI model, that is, the inference application stage of the AI model.
  • stage one the training stage of the AI model
  • stage two the deployment stage of the AI model, that is, the inference application stage of the AI model.
  • the AI model will be transmitted from the training node (providing node) to the inference node (receiving node).
  • the AI model is transmitted from one node to another node, at this time
  • the AI model needs to be digitized, that is, the structure of the AI model, the parameters of the model, etc. are expressed numerically.
  • the AI model is transmitted from one node to another. It is confirmed that the two parties supporting the AI model interaction support different data types, and the AI model is digitally represented according to the data type, that is, the AI model is converted into a corresponding AI model bit stream.
  • the different data types supported by the AI model interaction parties include but are not limited to the user terminal supporting 8-bit integers, the AI model interaction parties supporting 16-bit floating point types, etc.
  • Different data types will result in different resource overhead for transmitting AI models. For example, the overhead of using a 16-bit floating point type float is higher than that of an 8-bit integer type Int.
  • both sides of the AI model interaction support 8-bit integers, and when transmitting the structure of the AI model, model parameters, etc., using the data type 16-bit floating point for digital representation, it may increase the complexity of reasoning. Problems such as speed, increased latency or unnecessary functions, or even caused by the inability of both parties to interact with the AI model to process the 16-bit floating point data type, that is, the processing capabilities of the AI model receiving node cannot be satisfied. Unable to complete transfer of AI model.
  • Step S202 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • the converted AI model bit stream is transmitted to the AI model receiving node, and the receiving node reversely converts the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node.
  • the provider node after determining the data type for digital representation of the AI model, converts the AI model into the corresponding AI model bit stream according to the data type. , realize the data type used when the AI model interactor synchronizes the AI model transmission, which can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • the embodiment of the present disclosure provides another AI model transmission method.
  • Figure 2 is a schematic flowchart of another AI model transmission method provided by the embodiment of the disclosure. It can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S301 Determine the data type for digitally representing the AI model according to the data type support capability of the AI model receiving node.
  • the AI model involves two stages: stage one: the training stage of the AI model, that is, the stage of obtaining the AI model; stage two: the deployment stage of the AI model, that is, the inference application stage of the AI model.
  • stage one the training stage of the AI model, that is, the stage of obtaining the AI model
  • stage two the deployment stage of the AI model, that is, the inference application stage of the AI model.
  • the inference application stage corresponds to the AI model receiving node. Used to obtain inference results through inference calculations.
  • the data type for digital representation of the AI model is determined based on the data type support capabilities of the AI model receiving node.
  • the support capability of the AI model receiving node is integer, then it is determined that the data type represented by the AI model is also integer.
  • Step S302 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • the digital processing of the AI model is implemented according to the data type determined in step S301, that is, the structure of the AI model, parameters of the AI model, etc. are expressed digitally and converted into a corresponding AI model bit stream.
  • the converted AI model bit stream is transmitted to the AI model receiving node, and the receiving node reversely converts the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node.
  • the data type for digital representation of the AI model is determined based on the ability of the AI model receiving node to support the data type.
  • FIG. 4 is a schematic flowchart of another method of transmitting an AI model provided by an embodiment of the present disclosure. It can be applied to the process of transmitting the AI model between both parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S401 Receive the data type support capability of the AI model receiving node sent by the AI model receiving node.
  • the AI model interactor determines the data type represented by the AI model based on the data type support capability provided by the AI model receiving node to the AI model providing node. .
  • Step S402 Determine the data type for digital representation of the AI model according to the data type support capability of the AI model receiving node.
  • the data type corresponding to the maximum support capability data type of the AI model receiving node is determined to digitally represent the AI model.
  • data type For example, the maximum precision data types supported by the AI model receiving node are floating-point float and integer Int.
  • the AI model providing node determines that the data type for digital representation of the AI model is floating-point float, and converts the data type according to the floating-point float. The AI model is converted into the corresponding AI model bit stream.
  • a data type supported by the AI model providing node is selected as the data type for digital representation of the AI model.
  • the AI model receiving node reports that the data types it supports are floating-point float and integer Int.
  • the AI model providing node selects any data type based on the floating point type float and integer type Int supported by the AI model receiving node to determine the data type for digital representation of the AI model.
  • the AI model receiving node can support floating point type float and integer type Int.
  • the AI model receiving node selects float as the data type for digital representation of the AI model.
  • Step S403 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • the converted AI model bit stream is transmitted to the AI model receiving node, and the receiving node reversely converts the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node.
  • the AI model receiving node receives the data type support sent by the AI model receiving node. Capability, according to the data type support capability of the AI model receiving node, determine the data type for digital representation of the AI model, and convert the AI model into the corresponding AI model bit stream according to the data type, which can not only meet the requirements of AI
  • the model receives the processing power of the node and can flexibly represent the AI model according to the data type.
  • FIG. 5 is a schematic flowchart of another AI model transmission method provided by the embodiment of the disclosure, which can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S501 According to the data type support capability of the AI model receiving node, determine the data type corresponding to the maximum support capability data type of the AI model receiving node as the data type for digital representation of the AI model.
  • the data type corresponding to the maximum support capability data type of the AI model receiving node is determined as the data type for digital representation of the AI model.
  • the maximum precision data types supported by the AI model receiving node are floating-point float and integer type Int.
  • the AI model providing node determines that the data type for digital representation of the AI model is floating-point float, and converts the data type according to the floating-point float.
  • the AI model is converted into the corresponding AI model bit stream.
  • Step S502 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • Digital processing of the AI model is implemented according to the data type determined in step S501, that is, the structure of the AI model, parameters of the AI model, etc. are converted into corresponding AI model bit streams.
  • the converted AI model bit stream is transmitted to the AI model receiving node.
  • the receiving node reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node to complete the process. Transmission of AI models.
  • the maximum support capability of the AI model receiving node is The data type corresponding to the data type is determined as the data type that digitally represents the AI model.
  • the AI model is converted into the corresponding AI model bit stream according to the data type, which can satisfy the processing capabilities of the AI model receiving node. , and can flexibly represent the AI model according to the data type.
  • the embodiment of the present disclosure provides another method of transmitting an AI model.
  • Figure 6 is a schematic flowchart of another method of transmitting an AI model provided by the embodiment of the present disclosure. It can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S601 According to the data type support capability of the AI model receiving node, select a data type supported by the AI model providing node as the data type for digital representation of the AI model.
  • the AI model receiving node reports that the data types it supports are floating-point type float and integer type Int.
  • the AI model providing node selects any data type based on the floating point type float and integer type Int supported by the AI model receiving node to determine the data type for digital representation of the AI model.
  • the AI model receiving node can support floating point type float and integer type Int.
  • the AI model receiving node selects float as the data type for digital representation of the AI model, or the AI model receiving node selects integer Int as the data type for digital representation of the AI model.
  • Step S602 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • Digital processing of the AI model is implemented according to the data type determined in step S601, that is, the structure of the AI model, parameters of the AI model, etc. are converted into corresponding AI model bit streams.
  • the converted AI model bit stream is transmitted to the AI model receiving node.
  • the receiving node reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node to complete the process. Transmission of AI models.
  • an AI model is selected to provide node support based on the ability of the AI model receiving node to support data types.
  • the data type is used as the data type to digitally represent the AI model.
  • the AI model is converted into the corresponding AI model bit stream, which can not only meet the processing capabilities of the AI model receiving node, but also can process the AI model according to the data type.
  • AI models are represented flexibly.
  • the embodiment of the present disclosure provides another AI model transmission method.
  • Figure 7 is a schematic flowchart of another AI model transmission method provided by the embodiment of the disclosure. It can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S701 Determine the data type for digital representation of the AI model based on the power consumption and/or storage capacity reported by the AI model receiving node.
  • the determination is based on the power consumption and/or storage capabilities of the AI model receiving node, and cannot exceed the data corresponding to the power consumption and/or storage capabilities of the AI model receiving node.
  • type For example, when the storage capacity of the terminal is weak, a data type with lower precision can be selected for representation; or, when the receiving node does not expect to use larger power consumption to process AI tasks, data with lower precision can also be selected at this time. Type is represented.
  • Step S702 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • Digital processing of the AI model is implemented according to the data type determined in step S701, that is, the structure of the AI model, parameters of the AI model, etc. are converted into corresponding AI model bit streams.
  • the converted AI model bit stream is transmitted to the AI model receiving node.
  • the receiving node reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node to complete the process. Transmission of AI models.
  • the AI model is determined based on the power consumption and/or storage capacity reported by the AI model receiving node.
  • the data type for digital representation of the model, and the AI model is converted into the corresponding AI model bit stream according to the data type, which can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • FIG. 8 is a schematic flowchart of another AI model transmission method provided by the embodiment of the disclosure, which can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S801 Determine the data type for digital representation of the AI model according to business requirements.
  • the data type for digitally representing the AI model is determined according to the business requirements for delay. For example, if you have high latency requirements, you can choose Int, and if you have low latency requirements, you can choose float.
  • Step S802 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • Digital processing of the AI model is implemented according to the data type determined in step S701, that is, the structure of the AI model, parameters of the AI model, etc. are converted into corresponding AI model bit streams.
  • the converted AI model bit stream is transmitted to the AI model receiving node.
  • the receiving node reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses the AI model at the AI receiving node to complete the process. Transmission of AI models.
  • the data type for digital representation of the AI model is determined. According to the data type Converting the AI model into the corresponding AI model bit stream can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • the embodiment of the present disclosure provides another method of transmitting an AI model.
  • Figure 9 is a schematic flowchart of another method of transmitting an AI model provided by the embodiment of the present disclosure. It can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S901 Determine the data type represented by the AI model according to the resource overhead of the AI model.
  • the data type of the AI model receiving node is determined according to the size of the resources occupied by the AI model. For example, if the AI model is expected to occupy less or have smaller resource overhead, the number of digital representations of the AI model can be determined.
  • the data type is Int, and there is no limit on resource overhead. It can be determined that the data type used to digitally represent the AI model is floating point float.
  • Step S902 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • the data type represented by the AI model is determined according to the resource overhead of the AI model.
  • the data type converts the AI model into a corresponding AI model bit stream, which can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • the embodiment of the present disclosure provides another method of transmitting an AI model.
  • Figure 10 is a schematic flowchart of another method of transmitting an AI model provided by the embodiment of the present disclosure. It can be applied to the process of transmitting the AI model between the two parties in the AI model interaction.
  • the AI model transmission method can be executed alone, or can be executed in combination with any embodiment or possible implementation method in the embodiment, or can be executed in combination with any technical solution in related technologies. .
  • the transmission method of the AI model may include the following steps:
  • Step S1001 Convert the AI model into a corresponding AI model bit stream according to the data type.
  • Step S1002 Send indication information of the data type that digitally represents the AI model to the AI model receiving node, where the indication information includes the data type that digitally represents the AI model.
  • the AI model In response to the transmission of the AI model, transmit indication information of the data type that digitally represents the AI model, and the indication information includes the data type that digitally represents the AI model, so that the receiving node follows the data type in the indication information.
  • the AI model bit stream is reversely transformed to obtain the corresponding AI model, and the AI model is used at the AI receiving node.
  • the AI model is converted into the corresponding AI model bit stream according to the data type, and then The AI model receiving node sends indication information of the data type for digital representation of the AI model.
  • the indication information includes the data type for digital representation of the AI model, which can not only satisfy the processing capabilities of the AI model receiving node, but also be based on Data types provide flexible representation of AI models.
  • FIG. 11 is a schematic flowchart of another AI model transmission method provided by this embodiment of the disclosure. This method is applied to the receiving node side. As shown in Figure 11, the transmission method of the AI model may include the following steps:
  • Step S1101 In response to receiving the AI model bit stream sent by the AI model providing node, reverse-convert the AI model bit stream according to predetermined conversion rules to obtain the corresponding AI model, and use the AI model at the AI receiving node.
  • the AI model providing node determines the data type for digital representation of the AI model, converts the AI model into a corresponding AI model bit stream according to the data type, and sends the AI model bit stream to the AI model receiving node.
  • indication information of the data type for digitally representing the AI model is sent to the AI model receiving node, and the indication information includes digitizing the AI model.
  • the receiving node reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model.
  • the AI model bit stream is reversely converted according to the predetermined conversion rules according to the maximum support capacity data type of the AI model receiving node to obtain the corresponding AI model. .
  • Embodiments of the present disclosure do not limit the predetermined conversion rules used to reversely convert the AI model bit stream to obtain the corresponding AI model.
  • the providing node determines the data type for digital representation of the AI model, and based on the data type, The AI model is converted into the corresponding AI model bit stream.
  • the receiving node responds to the AI model bit stream sent by the AI model providing node, it reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and Using the AI model in the AI receiving node can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • FIG. 12 is a schematic flowchart of another AI model transmission method provided by this embodiment of the disclosure. It can be applied to the receiving node to receive the AI model bit stream sent by the providing node.
  • the transmission method of the AI model can be executed alone, or it can be executed in combination with any embodiment in the present disclosure or the possible implementation methods in the embodiment, or in combination with any of the related technologies.
  • Technical solutions are implemented together.
  • the transmission method of the AI model may include the following steps:
  • Step S1201 Receive indication information of the data type for digital representation of the AI model sent by the AI model providing node, where the indication information includes the data type for digital representation of the AI model.
  • the AI model providing node determines the data type for digital representation of the AI model, converts the AI model into a corresponding AI model bit stream according to the data type, and sends the AI model bit stream to the AI model receiving node.
  • indication information of the data type for digitally representing the AI model is sent to the AI model receiving node, and the indication information includes digitizing the AI model.
  • the data type represented is a code that represents the AI model.
  • Step S1202 In response to the instruction information, according to the data type that includes the digital representation of the AI model in the instruction information, reverse-convert the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model.
  • the providing node determines the data type for digital representation of the AI model, and based on the data type, The AI model is converted into a corresponding AI model bit stream.
  • the receiving node receives the AI model and provides the indication information of the data type for digital representation of the AI model sent by the node.
  • the indication information includes the data type for digital representation of the AI model.
  • the AI model bit stream is reversely converted according to the predetermined conversion rules to obtain the corresponding AI model, which can satisfy the requirements of the AI model receiving node. processing capabilities, and can flexibly represent AI models according to data types.
  • FIG. 13 is a schematic flowchart of another AI model transmission method provided by this embodiment of the disclosure. It can be applied to the receiving node to receive the AI model bit stream sent by the providing node.
  • the transmission method of the AI model can be executed alone, or it can be executed in combination with any embodiment in the present disclosure or the possible implementation methods in the embodiment, or in combination with any of the related technologies.
  • Technical solutions are implemented together.
  • the transmission method of the AI model may include the following steps:
  • Step S1301 Provide the AI model with the node's ability to report the data type.
  • Step S1302 Provide the node reporting to the AI model based on power consumption and/or storage capabilities.
  • the present disclosure also provides a device for transmitting the AI model, because the device for transmitting the AI model provided by the embodiment of the present disclosure is the same as the above-mentioned embodiment of FIGS. 2 to 13 .
  • the AI model transmission method provided corresponds to the AI model transmission method. Therefore, the implementation of the AI model transmission method is also applicable to the AI model transmission device provided by the embodiment of the disclosure, and will not be described in detail in the embodiment of the disclosure.
  • FIG. 14 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • the device is installed at the provision node, and the device includes:
  • Determining unit 1401 used to determine the data type for digital representation of the AI model
  • the conversion unit 1402 is used to convert the AI model into a corresponding AI model bit stream according to the data type.
  • the node determines the data type for digital representation of the AI model, it converts the AI model into the corresponding AI model bit stream according to the data type, so that the AI model interactors can synchronize the AI model transmission time.
  • the data type used can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • the determining unit 1401 is also configured to determine the data type for digitally representing the AI model according to the data type support capability of the AI model receiving node.
  • the receiving unit 1403 is configured to receive the data type support capability of the AI model receiving node sent by the AI model receiving node.
  • the determining unit 1401 is also configured to determine the data type corresponding to the maximum support capability data type of the AI model receiving node according to the support capability of the data type of the AI model receiving node. It is the data type that digitally represents the AI model.
  • the determining unit 1401 is also configured to select a data type supported by the AI model providing node according to the data type support capability of the AI model receiving node as the AI model.
  • a numerically represented data type is also configured to select a data type supported by the AI model providing node according to the data type support capability of the AI model receiving node as the AI model.
  • the determining unit 1401 is also configured to determine the data that digitally represents the AI model based on the power consumption and/or storage capacity reported by the AI model receiving node. type.
  • the determining unit 1401 is also configured to determine the data type for digitally representing the AI model according to business requirements.
  • the determining unit 1401 is also configured to determine the data type represented by the AI model according to the resource overhead of the AI model.
  • the sending unit 1404 is configured to send indication information of the data type that digitally represents the AI model to the AI model receiving node, where the indication information includes the data type that digitally represents the AI model.
  • the present disclosure also provides an AI model transmission device. Since the transmission device of the AI model provided by the embodiment of the present disclosure is consistent with the above-mentioned embodiment of FIGS. 11 to 13
  • the AI model transmission method provided corresponds to the AI model transmission method. Therefore, the implementation of the AI model transmission method is also applicable to the AI model transmission device provided by the embodiment of the disclosure, and will not be described in detail in the embodiment of the disclosure.
  • FIG. 15 is a schematic structural diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • the device is provided at the receiving node, and the device includes:
  • the conversion unit 1501 is configured to, in response to receiving the AI model bit stream sent by the AI model providing node, reversely convert the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model;
  • the use unit 1502 is used to use the AI model at the AI receiving node.
  • the receiving node when the receiving node responds to receiving the AI model bit stream sent by the AI model providing node, it reversely converts the AI model bit stream according to the predetermined conversion rules to obtain the corresponding AI model, and uses it at the AI receiving node.
  • the AI model realizes the data type used when the AI model interactor synchronizes the AI model transmission, which can not only meet the processing capabilities of the AI model receiving node, but also flexibly represent the AI model according to the data type.
  • the receiving unit 1503 is configured to receive indication information of a data type that digitally represents the AI model sent by the AI model providing node, where the indication information includes a data type that digitally represents the AI model.
  • the conversion unit 1501 is also configured to respond to the instruction information, according to the data type included in the instruction information to digitally represent the AI model, according to predetermined conversion rules. Reverse the AI model bit stream to obtain the corresponding AI model.
  • the device further includes:
  • the first reporting unit 1504 is configured to provide the AI model with the ability for nodes to report the data type.
  • the device further includes:
  • the second reporting unit 1505 is configured to provide node reporting based on power consumption and/or storage capabilities to the AI model.
  • the present disclosure also proposes an AI model transmission device.
  • the device includes a processor and a memory.
  • a computer program is stored in the memory.
  • the processor executes the computer program stored in the memory. So that the device performs the methods described in Figures 2 to 13.
  • the present disclosure also proposes another AI model transmission device, including: a processor and an interface circuit;
  • the interface circuit is used to receive code instructions and transmit them to the processor
  • the processor is configured to run the code instructions to perform the methods described in Figures 2 to 13.
  • the present disclosure proposes a computer-readable storage medium for storing instructions.
  • the instructions When the instructions are executed, the methods described in Figures 2 to 13 are implemented.
  • the methods provided by the embodiments of the present disclosure are introduced from the perspective of the AI model interactor.
  • the AI model interactor may include a hardware structure and a software module to implement the above functions in the form of a hardware structure, a software module, or a hardware structure plus a software module.
  • a certain function among the above functions can be executed by a hardware structure, a software module, or a hardware structure plus a software module.
  • network device 1600 includes a processing component 1622, which further includes at least one processor, and memory resources represented by memory 1632 for storing instructions, such as application programs, executable by processing component 1622.
  • the application program stored in memory 1632 may include one or more modules, each corresponding to a set of instructions.
  • the processing component 1622 is configured to execute instructions to perform any of the foregoing methods applied to the network device, for example, the methods described in the embodiments of FIGS. 2 to 12 .
  • Network device 1600 may also include a power supply component 1626 configured to perform power management of network device 1600, a wired or wireless network interface 1650 configured to connect network device 1600 to a network, and an input-output (I/O) interface 1658 .
  • Network device 1600 may operate based on an operating system stored in memory 1632, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, or the like.
  • Figure 17 is a block diagram of an AI model transmission device provided by an embodiment of the present disclosure.
  • the terminal device 1700 may be a mobile phone, a computer, a digital broadcast terminal device, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.
  • the terminal device 1700 may include at least one of the following components: a processing component 1702, a memory 1704, a power supply component 1706, a multimedia component 1708, an audio component 1710, an input/output (I/O) interface 1712, a sensor component 1714, and Communication component 1716.
  • the processing component 1702 generally controls the overall operations of the terminal device 1700, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 1702 may include at least one processor 1720 to execute instructions to complete all or part of the steps of the above method. Additionally, processing component 1702 may include at least one module that facilitates interaction between processing component 1702 and other components. For example, processing component 1702 may include a multimedia module to facilitate interaction between multimedia component 1708 and processing component 1702.
  • the memory 1704 is configured to store various types of data to support operations at the terminal device 1700 . Examples of such data include instructions for any application or method operating on the terminal device 1700, contact data, phonebook data, messages, pictures, videos, etc.
  • Memory 1704 may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EEPROM), Programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EEPROM erasable programmable read-only memory
  • EPROM Programmable read-only memory
  • PROM programmable read-only memory
  • ROM read-only memory
  • magnetic memory flash memory, magnetic or optical disk.
  • Power supply component 1706 provides power to various components of terminal device 1700.
  • Power supply component 1706 may include a power management system, at least one power supply, and other components associated with generating, managing, and distributing power to end device 1700 .
  • Multimedia component 1708 includes a screen providing an output interface between the terminal device 1700 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes at least one touch sensor to sense touches, slides, and gestures on the touch panel. The touch sensor may not only sense the boundary of the touch or sliding operation, but also detect the wake-up time and pressure related to the touch or sliding operation.
  • multimedia component 1708 includes a front-facing camera and/or a rear-facing camera.
  • the front camera and/or the rear camera may receive external multimedia data.
  • Each front-facing camera and rear-facing camera can be a fixed optical lens system or have a focal length and optical zoom capabilities.
  • Audio component 1710 is configured to output and/or input audio signals.
  • the audio component 1710 includes a microphone (MIC) configured to receive external audio signals when the terminal device 1700 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode.
  • the received audio signals may be further stored in memory 1704 or sent via communication component 1716 .
  • audio component 1710 also includes a speaker for outputting audio signals.
  • the I/O interface 1712 provides an interface between the processing component 1702 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, a button, etc. These buttons may include, but are not limited to: Home button, Volume buttons, Start button, and Lock button.
  • the sensor component 1714 includes at least one sensor for providing various aspects of status assessment for the terminal device 1700 .
  • the sensor component 1714 can detect the open/closed state of the terminal device 1700 and the relative positioning of components, such as the display and keypad of the terminal device 1700.
  • the sensor component 1714 can also detect the terminal device 1700 or one of the terminal devices 1700. Changes in the position of components, presence or absence of user contact with the terminal device 1700 , orientation or acceleration/deceleration of the terminal device 1700 and temperature changes of the terminal device 1700 .
  • Sensor component 1714 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
  • Sensor assembly 1714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 1714 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • the communication component 1716 is configured to facilitate wired or wireless communication between the terminal device 1700 and other devices.
  • the terminal device 1700 can access a wireless network based on communication standards, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 1716 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communications component 1716 also includes a near field communications (NFC) module to facilitate short-range communications.
  • NFC near field communications
  • the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • the terminal device 1700 may be configured by at least one application specific integrated circuit (ASIC), digital signal processor (DSP), digital signal processing device (DSPD), programmable logic device (PLD), field programmable gate Array (FPGA), controller, microcontroller, microprocessor or other electronic components are implemented for executing the methods shown in Figures 1 to 11 above.
  • ASIC application specific integrated circuit
  • DSP digital signal processor
  • DSPD digital signal processing device
  • PLD programmable logic device
  • FPGA field programmable gate Array
  • controller microcontroller, microprocessor or other electronic components are implemented for executing the methods shown in Figures 1 to 11 above.
  • a non-transitory computer-readable storage medium including instructions such as a memory 1704 including instructions, which can be executed by the processor 1720 of the terminal device 1700 to complete the above-described FIGS. 2 to 13 is also provided. method shown.
  • the non-transitory computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.
  • the above embodiments it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof.
  • software it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer programs.
  • the computer program When the computer program is loaded and executed on a computer, the processes or functions described in accordance with the embodiments of the present disclosure are generated in whole or in part.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.
  • the computer program may be stored in or transferred from one computer-readable storage medium to another, for example, the computer program may be transferred from a website, computer, server, or data center Transmission to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more available media integrated.
  • the usable media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVD)), or semiconductor media (e.g., solid state disks, SSD)) etc.
  • magnetic media e.g., floppy disks, hard disks, magnetic tapes
  • optical media e.g., high-density digital video discs (DVD)
  • DVD digital video discs
  • semiconductor media e.g., solid state disks, SSD
  • At least one in this application can also be described as one or more, and the plurality can be two, three, four or more, which is not limited by this application.
  • the technical feature is distinguished by “first”, “second”, “third”, “A”, “B”, “C” and “D” etc.
  • the technical features described in “first”, “second”, “third”, “A”, “B”, “C” and “D” are in no particular order or order.
  • the corresponding relationships shown in each table in this application can be configured or predefined.
  • the values of the information in each table are only examples and can be configured as other values, which are not limited by this application.
  • the corresponding relationships shown in some rows may not be configured.
  • appropriate deformation adjustments can be made based on the above table, such as splitting, merging, etc.
  • the names of the parameters shown in the titles of the above tables can also be other names that can be understood by the transmission device of the AI model, and the values or expressions of the parameters can also be other values or expressions that can be understood by the transmission device of the AI model.
  • other data structures can also be used, such as arrays, queues, containers, stacks, linear lists, pointers, linked lists, trees, graphs, structures, classes, heaps, hash tables or hash tables. wait.
  • Predefinition in this application can be understood as definition, pre-definition, storage, pre-storage, pre-negotiation, pre-configuration, solidification, or pre-burning.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Signal Processing (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Multimedia (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Communication Control (AREA)

Abstract

一种AI模型的传输方法及其装置，可以应用于AI模型交互方传输AI模型中，方法包括：提供节点确定对AI模型进行数字化表示的数据类型(201)，根据数据类型将AI模型转化为对应的AI模型bit流(202)，接收节点响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在AI接收节点使用AI模型(1101)，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。

Description

一种AI模型的传输方法及其装置 技术领域
本申请涉及通信技术领域，尤其涉及一种AI模型的传输方法及其装置。
背景技术
近年来，人工智能(Artificial Intelligence，AI)技术在多个领域取得不断突破。智能语音、计算机视觉等领域的持续发展不仅为智能终端带来丰富多彩的各种应用，在教育、交通、家居、医疗、零售、安防等多个领域也有广泛应用，给人们生活带来便利同时，也在促进各个行业进行产业升级。AI技术也正在加速与其他学科领域交叉渗透，其发展融合不同学科知识同时，也为不同学科的发展提供了新的方向和方法。
在人工智能(Artificial Intelligence，AI)技术中涉及两个重要的阶段，第一个阶段是模型的训练阶段，即获得模型的阶段；第二个阶段是模型的部署阶段，即模型的推理应用阶段。当模型的训练与推理不在同一个节点时，模型将会从训练的节点传输到推理的节点，当模型从一个节点传输到另一个节点时，此时需要对AI模型进行数字化处理，即将模型的结构，模型的参数等用数字表示出来。
AI模型节点传输过程中，由于不同的终端设备类型支持不同的数据类型，例如有的终端设备只支持8位整型,有的处理器支持16位浮点型。不同的数据类型会导致不同的推理复杂度。
发明内容
本公开实施例提供一种AI模型的传输方法及其装置，可以应用人工智能(Artificial Intelligence，AI)等领域，提供了一种灵活的AI模型表示方法，根据所述数据类型对AI模型进行数字化表示，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
第一方面，本公开实施例提供一种AI模型的传输方法，所述方法被提供节点执行，所述方法包括：
确定对AI模型进行数字化表示的数据类型；
根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
在该技术方案中，提供节点在确定对AI模型进行数字化表示的数据类型后，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
在一种实现方式中，所述确定对AI模型进行数字化表示的数据类型包括：
根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
在一种实现方式中，还包括：
接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力。
在一种实现方式中，所述根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型包括：
根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据 类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。
在一种实现方式中，所述根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型包括：
根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定AI模型接收节点的数据类型包括：
根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定AI模型接收节点的数据类型包括：
根据业务需求，确定对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定AI模型接收节点的数据类型包括：
根据所述AI模型的资源开销，确定所述AI模型表示的数据类型。
在一种实现方式中，还包括：
向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
第二方面，还提供一种AI模型的传输方法，所述方法被接收节点执行，所述方法包括：
响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
在该技术方案中，接收节点响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
在一种实现方式中，还包括：
接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型包括：
响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型。
在一种实现方式中，所述方法还包括：
向所述AI模型提供节点上报所述数据类型的支持能力。
在一种实现方式中，所述方法还包括：
向所述AI模型提供节点上报基于功耗和/或存储能力。
第三方面，还提供一种AI模型的传输装置，所述装置设置于提供节点，所述装置包括：
确定单元，用于确定对AI模型进行数字化表示的数据类型；
转化单元，用于根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
在一种实现方式中，所述所述确定单元，还用于根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
在一种实现方式中，还包括：
接收单元，用于接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力。
在一种实现方式中，所述确定单元，还用于根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定单元，还用于根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述所述确定单元，还用于根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定单元，还用于根据业务需求，确定对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述确定单元，还用于根据所述AI模型的资源开销，确定所述AI模型表示的数据类型。
在一种实现方式中，还包括：
发送单元，用于向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
第四方面，还提供一种AI模型的传输装置，所述装置设置于接收节点，所述装置包括：
转化单元，用于响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型；
使用单元，用于在所述AI接收节点使用所述AI模型。
在一种实现方式中，还包括：
接收单元，用于接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
在一种实现方式中，所述转化单元，还用于响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型。
在一种实现方式中，所述装置还包括：
第一上报单元，用于向所述AI模型提供节点上报所述数据类型的支持能力。
在一种实现方式中，所述装置还包括：
第二上报单元，用于向所述AI模型提供节点上报基于功耗和/或存储能力。
第五方面，本公开实施例提供一种AI模型的传输装置，所述装置包括处理器和存储器，所述存储器中存储有计算机程序，所述处理器执行所述存储器中存储的计算机程序，以使所述装置执行如第一方面或第二方面所述的方法。
第六方面，本公开实施例提供一种AI模型的传输装置，包括：处理器和接口电路；
所述接口电路，用于接收代码指令并传输至所述处理器；
所述处理器，用于运行所述代码指令以执行如第一方面或第二方面所述的方法。
第七方面，本公开实施例提供一种计算机可读存储介质，用于存储有指令，当所述指 令被执行时，使如第一方面或第二方面所述的方法被实现。
附图说明
为了更清楚地说明本公开实施例或背景技术中的技术方案，下面将对本公开实施例或背景技术中所需要使用的附图进行说明。
图1是本公开实施例提供的一种AI模型的传输方法流程示意图；
图2是本公开实施例提供的另一种AI模型的传输输方法的流程示意图；
图3是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图4是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图5是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图6是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图7是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图8是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图9是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图10是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图11是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图12是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图13是本公开实施例提供的另一种AI模型的传输方法的流程示意图；
图14是本公开实施例提供的一种AI模型的传输装置的结构示意图；
图15是本公开实施例提供的一种AI模型的传输装置的结构示意图；
图16是本公开实施例提供的一种AI模型的传输装置的结构示意图；
图17是本公开实施例提供的一种AI模型的传输装置的结构示意图。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开实施例相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本公开实施例的一些方面相一致的装置和方法的例子。
在本公开实施例使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本公开实施例。在本公开实施例和所附权利要求书中所使用的单数形式的“一种”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解，尽管在本公开实施例可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如，在不脱离本公开实施例范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，如在此所使用的词语“如果”及“若”可以被解释成为“在……时”或“当……时”或“响应于确定”。
下面详细描述本公开的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描 述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开实施例相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本公开实施例的一些方面相一致的装置和方法的例子。
在本公开实施例使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本公开实施例。在本公开实施例和所附权利要求书中所使用的单数形式的“一种”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解，尽管在本公开实施例可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如，在不脱离本公开实施例范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，如在此所使用的词语“如果”及“若”可以被解释成为“在……时”或“当……时”或“响应于确定”。
下面详细描述本公开的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的要素。下面通过参考附图描述的实施例是示例性的，旨在用于解释本公开，而不能理解为对本公开的限制。
为了更好的理解本公开实施例公开的一种直连sidelink非连续接收DRX的控制方法，下面首先对本公开实施例适用的通信系统进行描述。示相同或类似的要素。下面通过参考附图描述的实施例是示例性的，旨在用于解释本公开，而不能理解为对本公开的限制。
为了更好的理解本公开实施例公开的一种AI模型的传输方法，下面首先对本公开实施例适用的通信系统进行描述。
请参见图1，图1为本申请实施例提供的一种通信系统的架构示意图。该通信系统可包括但不限于一个网络设备、一个终端设备，其中，终端设备与网络设备进行通信，图1所示的设备数量和形态仅用于举例并不构成对本申请实施例的限定，实际应用中可以包括两个或两个以上的网络设备，两个或两个以上的终端设备。图1所示的通信系统以包括一个网络设备101、终端设备102。
作为一种示例，一个终端设备既可以作为AI模型的提供节点，也可以作为AI模型接收节点。作为另一种示例，一个网络设备既可以作为AI模型的提供节点，也可以作为AI模型接收节点。
需要说明的是，本公开实施例的技术方案可以应用于各种通信系统。例如：长期演进(long term evolution，LTE)系统、第五代(5th generation，5G)移动通信系统、5G新空口(new radio，NR)系统，或者其他未来的新型移动通信系统等。
本公开实施例中的网络设备是网络侧的一种用于发射或接收信号的实体。例如，网络设备101可以为演进型基站(evolved NodeB，eNB)、传输接收点(transmission reception point或transmit receive point，TRP)、NR系统中的下一代基站(next generation NodeB，gNB)、其他未来移动通信系统中的基站或无线保真(wireless fidelity，WiFi)系统中的接入节点等。本公开的实施例对网络设备所采用的具体技术和具体设备形态不做限定。本公开实施例提供的网络设备可以是由集中单元(central unit，CU)与分布式单元(distributed unit，DU) 组成的，其中，CU也可以称为控制单元(control unit)，采用CU-DU的结构可以将网络设备，例如基站的协议层拆分开，部分协议层的功能放在CU集中控制，剩下部分或全部协议层的功能分布在DU中，由CU集中控制DU。
本公开实施例中的终端设备是用户侧的一种用于接收或发射信号的实体，如手机。终端设备也可以称为终端设备(terminal)、移动台(mobile station，MS)、移动终端设备(mobile terminal，MT)等。终端设备可以是具备通信功能的汽车、智能汽车、手机(mobile phone)、穿戴式设备、平板电脑(Pad)、带无线收发功能的电脑、虚拟现实(virtual reality，VR)终端设备、增强现实(augmented reality，AR)终端设备、工业控制(industrial control)中的无线终端设备、无人驾驶(self-driving)中的无线终端设备、远程手术(remote medical surgery)中的无线终端设备、智能电网(smart grid)中的无线终端设备、运输安全(transportation safety)中的无线终端设备、智慧城市(smart city)中的无线终端设备、智慧家庭(smart home)中的无线终端设备等等。本公开的实施例对终端设备所采用的具体技术和具体设备形态不做限定。
AI模型节点传输过程中，由于不同的终端设备类型支持不同的数据类型，例如有的终端设备只支持8位整型,有的处理器支持16位浮点型。不同的数据类型会导致不同的推理复杂度。
针对上述问题，本公开实施例提出一种AI模型的传输方法及其装置。
请参见图2，图2是本公开实施例提供的一种数据类型传输方法的流程示意图。如图2所示，该方法应用于提供节点，该方法可以包括但不限于如下步骤：
步骤S201：确定对AI模型进行数字化表示的数据类型。
其中，人工智能(Artificial Intelligence，AI)模型，涉及两个阶段：阶段一：AI模型的训练阶段，即获得AI模型的阶段，阶段二：AI模型的部署阶段，即AI模型的推理应用阶段，当AI模型的训练与推理不在同一个节点时，AI模型将会从训练的节点(提供节点)传输到推理的节点(接收节点)，当AI模型从一个节点传输到另一个节点时，此时需要对AI模型进行数字化处理，即将AI模型的结构，模型的参数等用数字表示出来。
本公开实施例中，应用于AI模型从一个节点传输到另一个节点的场景中，确认AI模型交互双方所支持不同的数据类型，根据数据类型对AI模型进行数字化表示，即将AI模型转化为对应的AI模型bit流。
作为一种可能的实现方式，AI模型交互双方所支持不同的数据类型包括但不限于用户终端支持8位整型，AI模型交互双方支持16位浮点型等等。不同的数据类型会导致不同的用以传输AI模型的资源开销。例如使用16位浮点型float的开销比8位整型Int的开销高。
除此之外，若AI模型交互双方支持8位整型，而在传输AI模型的结构，模型的参数等时，用数据类型16位浮点型进行数字化表示时，可能会出现增加了推理复杂度、增长时延或者增大不必要的功能等问题，甚至由于AI模型交互双方无法处理16位浮点型的数据类型，即不能满足AI模型接收节点的处理能力，而导致的AI模型交互双方无法完成AI模型的传输。
因此，在执行AI模型的传输时，需参考AI模型交互双方所支持的数据类型，即确定 对AI模型进行数字化表示的数据类型。
步骤S202：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
综上，由于不同的数据类型也会导致不同的推理复杂度，提供节点在确定对AI模型进行数字化表示的数据类型后，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图2为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图3所示，该AI模型的传输方法可包括如下步骤：
步骤S301：根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
AI模型涉及两个阶段：阶段一：AI模型的训练阶段，即获得AI模型的阶段，阶段二：AI模型的部署阶段，即AI模型的推理应用阶段，在推理应用阶段对应AI模型接收节点，用于通过推理计算得到推理结果。在AI模型交互过程中，根据AI模型接收节点对数据类型的支持能力，确定对AI模型进行数字化表示的数据类型。
示例性的，AI模型接收节点的支持能力为整型，那么确定所述AI模型表示的数据类型也为整型。
步骤S302：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
根据步骤S301确定的数据类型实现AI模型的数字化处理，即将AI模型的结构、AI模型的参数等用数字表示出来，转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
综上，不同的数据类型也会导致不同的推理复杂度，在该技术方案中，根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型，根据数据类型将AI模型转化为对应的AI模型bit流，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图4为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图4所示，该AI模型的传输方法可包括如下步骤：
步骤S401：接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力。
在本公开实施例中，AI模型交互方(AI模型提供节点及AI模型接收)根据AI模型接收节点向AI模型提供节点，提供的对数据类型的支持能力，确定所述AI模型表示的数据类型。
步骤S402：根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
作为本公开实施例的一种可实现方式，根据AI模型接收节点支持的最大精度的数据类型，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。例如，AI模型接收节点支持的最大精度的数据类型为浮点型float，整型Int，AI模型提供节点确定对AI模型进行数字化表示的数据类型为浮点型float，并根据浮点型float将AI模型转化为对应的AI模型bit流。
作为本公开实施例的另一种可实现方式，根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型，示例性的，AI模型接收节点上报其支持的数据类型为浮点型float,整型Int。AI模型提供节点根据AI模型接收节点所支持的浮点型float,整型Int，选择任意数据类型确定为对AI模型进行数字化表示的数据类型。例如AI模型接收节点可以支持浮点型float,整型Int。此时AI模型接收节点选择浮点型float作为对AI模型进行数字化表示的数据类型。
步骤S403：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力，根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图5为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图5所示，该AI模型的传输方法可包括如下步骤：
步骤S501：根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。
为了便于理解，根据AI模型接收节点支持的最大精度的数据类型，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。例如，AI模型接收节点支持的最大精度的数据类型为浮点型float，整型Int，AI模型提供节点确定对AI模型进行数字化表示的数据类型为浮点型float，并根据浮点型float将AI模型转化为对应的AI模型bit流。
步骤S502：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
根据步骤S501确定的数据类型实现AI模型的数字化处理，即将AI模型的结构、AI 模型的参数等转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，完成AI模型的传输。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图6为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图6所示，该AI模型的传输方法可包括如下步骤：
步骤S601：根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型。
示例性的，AI模型接收节点上报其支持的数据类型为浮点型float,整型Int。AI模型提供节点根据AI模型接收节点所支持的浮点型float,整型Int，选择任意数据类型确定为对AI模型进行数字化表示的数据类型。例如AI模型接收节点可以支持浮点型float,整型Int。此时AI模型接收节点选择浮点型float作为对AI模型进行数字化表示的数据类型，或者，AI模型接收节点选择整型Int作为对AI模型进行数字化表示的数据类型。
步骤S602：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
根据步骤S601确定的数据类型实现AI模型的数字化处理，即将AI模型的结构、AI模型的参数等转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，完成AI模型的传输。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图7为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图7所示，该AI模型的传输方法可包括如下步骤：
步骤S701：根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行 数字化表示的数据类型。
在确定AI模型进行数字化表示的数据类型时，根据该AI模型接收节点的基于功耗和/或存储能力进行确定，而不能超出AI模型接收节点的基于功耗和/或存储能力所对应的数据类型。例如当终端的存储能力较弱时，此时可以选择精度较低的数据类型进行表示；或者，当接收节点不期待使用较大功耗处理AI任务时，此时也可以选择精度较低的数据类型进行表示。
步骤S702：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
根据步骤S701确定的数据类型实现AI模型的数字化处理，即将AI模型的结构、AI模型的参数等转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，完成AI模型的传输。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图8为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图8所示，该AI模型的传输方法可包括如下步骤：
步骤S801：根据业务需求，确定对AI模型进行数字化表示的数据类型。
在本公开实施例中，根据业务对时延的要求确定对AI模型进行数字化表示的数据类型。示例性，对时延要求高的可以选择Int，对时延要求低的可以选择float。
步骤S802：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
根据步骤S701确定的数据类型实现AI模型的数字化处理，即将AI模型的结构、AI模型的参数等转化为对应的AI模型bit流。
将转化后的AI模型bit流传输至AI模型接收节点，由接收节点按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，完成AI模型的传输。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据业务需求，确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图9为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图9所示，该AI模型的传输方法可包括如下步骤：
步骤S901：根据所述AI模型的资源开销，确定所述AI模型表示的数据类型。
在本公开实施例中，根据AI模型占用资源的大小确定AI模型接收节点的数据类型，示例性的，若期望AI模型占用较少或较小的资源开销，可以确定对AI模型进行数字化表示的数据类型为整型Int，对资源开销没有限制的，可以确定对AI模型进行数字化表示的数据类型为浮点型float。
步骤S902：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据所述AI模型的资源开销，确定所述AI模型表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图10为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于AI模型交互双方传输AI模型过程中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图10所示，该AI模型的传输方法可包括如下步骤：
步骤S1001：根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
步骤S1002：向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
响应于所述AI模型的传输，传输对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型，以使得接收节点按照指示信息中的数据类型将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图11为本公开实施例提供的另一种AI模型的传输方法的流程示意图，该方法应用于接收节点侧。如图11所示，该AI模型的传输方法可包括如下步骤：
步骤S1101：响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
在AI模型提供节点确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，并将AI模型bit流发送至AI模型接收节点。
作为本公开实施例的示例，在提供节点传输AI模型bit流时，向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进 行数字化表示的数据类型。接收节点根据预定转换规将AI模型bit流进行逆转化得到对应的AI模型。
作为本公开实施例的另一示例，在接收节点接收到AI模型bit流后，根据AI模型接收节点最大支持能力数据类型，按照预定转化规则执行将AI模型bit流进行逆转化得到对应的AI模型。
本公开的实施例，对将AI模型bit流进行逆转化得到对应的AI模型所使用的预定转化规则不进行限定。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，由提供节点确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，接收节点响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图12为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于接收节点接收提供节点发送的AI模型bit流的场景中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图12所示，该AI模型的传输方法可包括如下步骤：
步骤S1201：接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
在AI模型提供节点确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，并将AI模型bit流发送至AI模型接收节点。
作为本公开实施例的示例，在提供节点传输AI模型bit流时，向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
步骤S1202：响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型。
综上，由于不同的数据类型也会导致不同的推理复杂度，时延和功耗等，在该技术方案中，由提供节点确定对AI模型进行数字化表示的数据类型，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，接收节点接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型，响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
本公开实施例提供了另一种AI模型的传输方法，图13为本公开实施例提供的另一种AI模型的传输方法的流程示意图，可应用于接收节点接收提供节点发送的AI模型bit流的场景中，该AI模型的传输方法可以单独被执行，也可以结合本公开中的任一个实施例或是 实施例中的可能的实现方式一起被执行，还可以结合相关技术中的任一种技术方案一起被执行。
如图13所示，该AI模型的传输方法可包括如下步骤：
步骤S1301：向所述AI模型提供节点上报所述数据类型的支持能力。
步骤S1302：向所述AI模型提供节点上报基于功耗和/或存储能力。
有关AI模型接收节点向提供节点发送数据类型的支持能力、基于功耗和/或存储能力，请参阅上述相关内容的详细说明，本公开实施例在此不再进行一一赘述。
与上述图2至图10实施例提供的AI模型的传输相对应，本公开还提供一种AI模型的传输装置，由于本公开实施例提供AI模型的传输装置与上述图2至图13实施例提供的AI模型的传输方法相对应，因此在AI模型的传输方法的实施方式也适用于本公开实施例提供的AI模型的传输装置，在本公开实施例中不再详细描述。
图14为本公开实施例所提供的一种AI模型的传输装置的结构示意图。所述装置设置于提供节点，所述装置包括：
确定单元1401，用于确定对AI模型进行数字化表示的数据类型；
转化单元1402，用于根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
在该技术方案中，提供节点在确定对AI模型进行数字化表示的数据类型后，根据所述数据类型将所述AI模型转化为对应的AI模型bit流，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
作为本公开实施例的的一种可能实现方式，所述所述确定单元1401，还用于根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，还包括：
接收单元1403，用于接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力。
作为本公开实施例的的一种可能实现方式，所述确定单元1401，还用于根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，所述确定单元1401，还用于根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，所述所述确定单元1401，还用于根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，所述确定单元1401，还用于根据业务需求，确定对AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，所述确定单元1401，还用于根据所述AI模型的资源开销，确定所述AI模型表示的数据类型。
作为本公开实施例的的一种可能实现方式，还包括：
发送单元1404，用于向所述AI模型接收节点发送对AI模型进行数字化表示的数据类 型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
与上述图11至图13实施例提供的AI模型的传输相对应，本公开还提供一种AI模型的传输装置，由于本公开实施例提供AI模型的传输装置与上述图11至图13实施例提供的AI模型的传输方法相对应，因此在AI模型的传输方法的实施方式也适用于本公开实施例提供的AI模型的传输装置，在本公开实施例中不再详细描述。
图15为本公开实施例所提供的一种AI模型的传输装置的结构示意图。所述装置设置于接收节点，所述装置包括：
转化单元1501，用于响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型；
使用单元1502，用于在所述AI接收节点使用所述AI模型。
在该技术方案中，接收节点响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型，实现AI模型交互方同步AI模型传输时使用的数据类型，既能满足AI模型接收节点的处理能力，又能根据数据类型对AI模型进行灵活表示。
作为本公开实施例的的一种可能实现方式，还包括：
接收单元1503，用于接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
作为本公开实施例的的一种可能实现方式，所述转化单元1501，还用于响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型。
作为本公开实施例的的一种可能实现方式，所述装置还包括：
第一上报单元1504，用于向所述AI模型提供节点上报所述数据类型的支持能力。
作为本公开实施例的的一种可能实现方式，所述装置还包括：
第二上报单元1505，用于向所述AI模型提供节点上报基于功耗和/或存储能力。
为了实现上述实施例，本公开还提出一种AI模型的传输装置，所述装置包括处理器和存储器，所述存储器中存储有计算机程序，所述处理器执行所述存储器中存储的计算机程序，以使所述装置执行如图2至图13所述的方法。
为了实现上述实施例，本公开还提出另一种AI模型的传输装置，包括：处理器和接口电路；
所述接口电路，用于接收代码指令并传输至所述处理器；
所述处理器，用于运行所述代码指令以执行如图2至图13所述的方法。
为了实现上述实施例，本公开提出一种计算机可读存储介质，用于存储有指令，当所述指令被执行时，使如图2至图13所述的方法被实现。
上述本申请提供的实施例中，从AI模型交互方的角度对本公开实施例提供的方法进行了介绍。为了实现上述本公开实施例提供的方法中的各功能，AI模型交互方可以包括硬件结构、软件模块，以硬件结构、软件模块、或硬件结构加软件模块的形式来实现上述各功能。上述各功能中的某个功能可以以硬件结构、软件模块、或者硬件结构加软件模块的方式来执行。
请参见图16，图16为本公开实施例所提供的一种AI模型的传输装置的结构示意图。参照图16，网络设备1600包括处理组件1622，其进一步包括至少一个处理器，以及由存储器1632所代表的存储器资源，用于存储可由处理组件1622的执行的指令，例如应用程序。存储器1632中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1622被配置为执行指令，以执行上述方法前述应用在所述网络设备的任意方法，例如，如图2至图12实施例所述的方法。
网络设备1600还可以包括一个电源组件1626被配置为执行网络设备1600的电源管理，一个有线或无线网络接口1650被配置为将网络设备1600连接到网络，和一个输入输出(I/O)接口1658。网络设备1600可以操作基于存储在存储器1632的操作系统，例如Windows Server TM，Mac OS XTM，UnixTM，LinuxTM，FreeBSDTM或类似。
图17为本公开实施例所提供的一种AI模型的传输装置的框图。例如，终端设备1700可以是移动电话，计算机，数字广播终端设备，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等。
参照图17，终端设备1700可以包括以下至少一个组件：处理组件1702，存储器1704，电源组件1706，多媒体组件1708，音频组件1710，输入/输出(I/O)的接口1712，传感器组件1714，以及通信组件1716。
处理组件1702通常控制终端设备1700的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件1702可以包括至少一个处理器1720来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件1702可以包括至少一个模块，便于处理组件1702和其他组件之间的交互。例如，处理组件1702可以包括多媒体模块，以方便多媒体组件1708和处理组件1702之间的交互。
存储器1704被配置为存储各种类型的数据以支持在终端设备1700的操作。这些数据的示例包括用于在终端设备1700上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器1704可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件1706为终端设备1700的各种组件提供电力。电源组件1706可以包括电源管理系统，至少一个电源，及其他与为终端设备1700生成、管理和分配电力相关联的组件。
多媒体组件1708包括在所述终端设备1700和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括至少一个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的唤醒时间和压力。在一些实施例中，多媒体组件1708包括一个前置摄像头和/或后置摄像头。当终端设备1700处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件1710被配置为输出和/或输入音频信号。例如，音频组件1710包括一个麦克 风(MIC)，当终端设备1700处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器1704或经由通信组件1716发送。在一些实施例中，音频组件1710还包括一个扬声器，用于输出音频信号。
I/O接口1712为处理组件1702和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件1714包括至少一个传感器，用于为终端设备1700提供各个方面的状态评估。例如，传感器组件1714可以检测到终端设备1700的打开/关闭状态，组件的相对定位，例如所述组件为终端设备1700的显示器和小键盘，传感器组件1714还可以检测终端设备1700或终端设备1700一个组件的位置改变，用户与终端设备1700接触的存在或不存在，终端设备1700方位或加速/减速和终端设备1700的温度变化。传感器组件1714可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件1714还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件1714还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件1716被配置为便于终端设备1700和其他设备之间有线或无线方式的通信。终端设备1700可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件1716经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件1716还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，终端设备1700可以被至少一个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述图1至11所示的方法。
在示例性实施例中，还提供了一种包括指令的非临时性计算机可读存储介质，例如包括指令的存储器1704，上述指令可由终端设备1700的处理器1720执行以完成上述图2至图13所示的方法。例如，所述非临时性计算机可读存储介质可以是ROM、随机存取存储器(RAM)、CD-ROM、磁带、软盘和光数据存储设备等。
本领域技术人员还可以了解到本公开实施例列出的各种说明性逻辑块(illustrative logical block)和步骤(step)可以通过电子硬件、电脑软件，或两者的结合进行实现。这样的功能是通过硬件还是软件来实现取决于特定的应用和整个系统的设计要求。本领域技术人员可以对于每种特定的应用，可以使用各种方法实现所述的功能，但这种实现不应被理解为超出本公开实施例保护的范围。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机程序。在计算机上加载和执行所述计算机程序时，全部或部分地 产生按照本公开实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机程序可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，所述计算机程序可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如，高密度数字视频光盘(digital video disc，DVD))、或者半导体介质(例如，固态硬盘(solid state disk，SSD))等。
本领域普通技术人员可以理解：本申请中涉及的第一、第二等各种数字编号仅为描述方便进行的区分，并不用来限制本公开实施例的范围，也表示先后顺序。
本申请中的至少一个还可以描述为一个或多个，多个可以是两个、三个、四个或者更多个，本申请不做限制。在本公开实施例中，对于一种技术特征，通过“第一”、“第二”、“第三”、“A”、“B”、“C”和“D”等区分该种技术特征中的技术特征，该“第一”、“第二”、“第三”、“A”、“B”、“C”和“D”描述的技术特征间无先后顺序或者大小顺序。
本申请中各表所示的对应关系可以被配置，也可以是预定义的。各表中的信息的取值仅仅是举例，可以配置为其他值，本申请并不限定。在配置信息与各参数的对应关系时，并不一定要求必须配置各表中示意出的所有对应关系。例如，本申请中的表格中，某些行示出的对应关系也可以不配置。又例如，可以基于上述表格做适当的变形调整，例如，拆分，合并等等。上述各表中标题示出参数的名称也可以采用AI模型的传输装置可理解的其他名称，其参数的取值或表示方式也可以AI模型的传输装置可理解的其他取值或表示方式。上述各表在实现时，也可以采用其他的数据结构，例如可以采用数组、队列、容器、栈、线性表、指针、链表、树、图、结构体、类、堆、散列表或哈希表等。
本申请中的预定义可以理解为定义、预先定义、存储、预存储、预协商、预配置、固化、或预烧制。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (19)

  1. 一种AI模型的传输方法，所述方法被提供节点执行，其特征在于，所述方法包括：
    确定对AI模型进行数字化表示的数据类型；
    根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
  2. 根据权利要求1所述的传输方法，其特征在于，所述确定对AI模型进行数字化表示的数据类型包括：
    根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型。
  3. 根据权利要求2所述的传输方法，其特征在于，还包括：
    接收所述AI模型接收节点发送的所述AI模型接收节点对数据类型的支持能力。
  4. 根据权利要求2所述的传输方法，其特征在于，所述根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型包括：
    根据AI模型接收节点的数据类型的支持能力，将AI模型接收节点最大支持能力数据类型对应的数据类型，确定为对所述AI模型进行数字化表示的数据类型。
  5. 根据权利要求2所述的传输方法，其特征在于，所述根据AI模型接收节点对数据类型的支持能力，确定所述对AI模型进行数字化表示的数据类型包括：
    根据AI模型接收节点对数据类型的支持能力，选择一种AI模型提供节点支持的数据类型作为对AI模型进行数字化表示的数据类型。
  6. 根据权利要求1所述的传输方法，其特征在于，所述确定AI模型接收节点的数据类型包括：
    根据所述AI模型接收节点上报的功耗和/或存储能力，确定对AI模型进行数字化表示的数据类型。
  7. 根据权利要求1所述的传输方法，其特征在于，所述确定AI模型接收节点的数据类型包括：
    根据业务需求，确定对AI模型进行数字化表示的数据类型。
  8. 根据权利要求1所述的传输方法，其特征在于，所述确定AI模型接收节点的数据类型包括：
    根据所述AI模型的资源开销，确定所述AI模型表示的数据类型。
  9. 根据权利要求1-7中任一项所述的传输方法，其特征在于，还包括：
    向所述AI模型接收节点发送对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
  10. 一种AI模型的传输方法，所述方法被接收节点执行，其特征在于，所述方法包括：
    响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型，并在所述AI接收节点使用所述AI模型。
  11. 根据权利要求10所述的传输方法，其特征在于，还包括：
    接收AI模型提供节点发送的对AI模型进行数字化表示的数据类型的指示信息，所述指示信息中包括对AI模型进行数字化表示的数据类型。
  12. 根据权利要求11所述的传输方法，其特征在于，所述按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型包括：
    响应于所述指示信息，根据所述指示信息中包括对AI模型进行数字化表示的数据类型，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型。
  13. 根据权利要求10所述的传输方法，其特征在于，所述方法还包括：
    向所述AI模型提供节点上报所述数据类型的支持能力。
  14. 根据权利要求10所述的传输方法，其特征在于，所述方法还包括：
    向所述AI模型提供节点上报基于功耗和/或存储能力。
  15. 一种AI模型的传输装置，所述装置设置于提供节点，其特征在于，所述装置包括：
    确定单元，用于确定对AI模型进行数字化表示的数据类型；
    转化单元，用于根据所述数据类型将所述AI模型转化为对应的AI模型bit流。
  16. 一种AI模型的传输装置，所述装置设置于接收节点，其特征在于，所述装置包括：
    转化单元，用于响应于接收到AI模型提供节点发送的AI模型bit流时，按照预定转化规则将AI模型bit流进行逆转化得到对应的AI模型；
    使用单元，用于在所述AI接收节点使用所述AI模型。
  17. 一种AI模型的传输装置，其特征在于，所述装置包括处理器和存储器，所述存储器中存储有计算机程序，所述处理器执行所述存储器中存储的计算机程序，以使所述装置执行如权利要求1～9中任一项所述的方法，或如权利要求10～14中任一项所述的方法。
  18. 一种AI模型的传输装置，其特征在于，包括：处理器和接口电路；
    所述接口电路，用于接收代码指令并传输至所述处理器；
    所述处理器，用于运行所述代码指令以执行如权利要1～9中任一项所述的方法，或如权利要求10～14中任一项所述的方法。
  19. 一种计算机可读存储介质，用于存储有指令，当所述指令被执行时，使如权利要求1～9中任一项所述的方法，或如权利要求10～14中任一项所述的方法被实现。
PCT/CN2022/103168 2022-06-30 2022-06-30 一种ai模型的传输方法及其装置 WO2024000532A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202280002427.8A CN118202645A (zh) 2022-06-30 2022-06-30 一种ai模型的传输方法及其装置
PCT/CN2022/103168 WO2024000532A1 (zh) 2022-06-30 2022-06-30 一种ai模型的传输方法及其装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/103168 WO2024000532A1 (zh) 2022-06-30 2022-06-30 一种ai模型的传输方法及其装置

Publications (1)

Publication Number Publication Date
WO2024000532A1 true WO2024000532A1 (zh) 2024-01-04

Family

ID=89383867

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/103168 WO2024000532A1 (zh) 2022-06-30 2022-06-30 一种ai模型的传输方法及其装置

Country Status (2)

Country Link
CN (1) CN118202645A (zh)
WO (1) WO2024000532A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101188757A (zh) * 2003-12-01 2008-05-28 三星电子株式会社 用于可伸缩视频编码和解码的方法和设备
US20080235672A1 (en) * 2007-03-20 2008-09-25 Lozano Luis A Data layout using data type information
CN104041036A (zh) * 2012-01-03 2014-09-10 杜比实验室特许公司 指定视觉动态范围编码操作及参数
CN114175652A (zh) * 2019-08-19 2022-03-11 三星电子株式会社 解码设备及其操作方法以及人工智能(ai)放大设备及其操作方法
CN114301573A (zh) * 2021-11-24 2022-04-08 超讯通信股份有限公司 联邦学习模型参数传输方法及系统
CN114616832A (zh) * 2019-09-11 2022-06-10 南洋理工大学 基于网络的视觉分析

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101188757A (zh) * 2003-12-01 2008-05-28 三星电子株式会社 用于可伸缩视频编码和解码的方法和设备
US20080235672A1 (en) * 2007-03-20 2008-09-25 Lozano Luis A Data layout using data type information
CN104041036A (zh) * 2012-01-03 2014-09-10 杜比实验室特许公司 指定视觉动态范围编码操作及参数
CN114175652A (zh) * 2019-08-19 2022-03-11 三星电子株式会社 解码设备及其操作方法以及人工智能(ai)放大设备及其操作方法
CN114616832A (zh) * 2019-09-11 2022-06-10 南洋理工大学 基于网络的视觉分析
CN114301573A (zh) * 2021-11-24 2022-04-08 超讯通信股份有限公司 联邦学习模型参数传输方法及系统

Also Published As

Publication number Publication date
CN118202645A (zh) 2024-06-14

Similar Documents

Publication Publication Date Title
US20230276430A1 (en) Resource scheduling method and apparatus, communication device and storage medium
JP7194201B2 (ja) Ｄｃｉの伝送方法、端末及び基地局
US11910377B2 (en) Method and apparatus for requesting scheduling sidelink resource, and storage medium
EP3644671A1 (en) Method and device for indicating radio bearer
EP3125163A1 (en) Method and device for flight notification, and method and device for flight setting
KR102580710B1 (ko) 전자 장치들의 연동 방법 및 장치
US20230276340A1 (en) Capability reporting method and apparatus, and capability determination method and apparatus
EP4030817A1 (en) Data processing method and apparatus, and electronic device and computer readable storage medium
WO2019237930A1 (zh) 小区管理方法、终端及网络侧设备
US11864190B2 (en) Method and apparatus for configuring and determining transmission block scheduling interval, and base station
WO2020143503A1 (zh) 测量上报方法及装置、终端设备信息获取方法及装置
JP2021519015A (ja) セル設定方法、端末及びネットワークノード
WO2023207046A1 (zh) 调度信息确定、下行控制信息发送方法和装置
US10123208B2 (en) Method and device for controlling access
WO2019227318A1 (zh) 物理下行控制信道监测配置、监测方法及装置和基站
WO2024000532A1 (zh) 一种ai模型的传输方法及其装置
US11751235B2 (en) Information transmission method and device
CN113162738B (zh) 一种上行传输方法、装置、设备及存储介质
CN111144845B (zh) 一种移动终端会议提醒方法和装置
WO2020164515A1 (zh) 信号传输方法、设备及系统
WO2024000605A1 (zh) 一种ai模型推理的方法及其装置
US20210152280A1 (en) Methods and apparatuses for determining number of times of blind decoding schedule signaling, user equipment and base station
CN110311961A (zh) 信息分享的方法及系统、客户端和服务端
WO2023206032A2 (zh) 直连sidelink非连续接收DRX的控制方法及装置
WO2024020808A1 (zh) 非激活态组播业务接收、配置指示信息发送方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22948634

Country of ref document: EP

Kind code of ref document: A1