WO2024000605A1 - 一种ai模型推理的方法及其装置 - Google Patents

一种ai模型推理的方法及其装置 Download PDF

Info

Publication number
WO2024000605A1
WO2024000605A1 PCT/CN2022/103485 CN2022103485W WO2024000605A1 WO 2024000605 A1 WO2024000605 A1 WO 2024000605A1 CN 2022103485 W CN2022103485 W CN 2022103485W WO 2024000605 A1 WO2024000605 A1 WO 2024000605A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
inference
reasoning
task
response
Prior art date
Application number
PCT/CN2022/103485
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 PCT/CN2022/103485 priority Critical patent/WO2024000605A1/zh
Priority to CN202280002424.4A priority patent/CN117651954A/zh
Publication of WO2024000605A1 publication Critical patent/WO2024000605A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models

Definitions

  • the present application relates to the field of communication technology, and in particular, to a method and device for AI model reasoning.
  • AI Artificial Intelligence
  • the main participants of AI technology are mainly base stations and terminal equipment.
  • the base station provides the AI model and the terminal performs inference. Since the terminal equipment performs inference, the terminal equipment needs to have certain hardware capabilities and software platform, which usually requires a comparison of processing capabilities. Only high-end terminal devices can perform inference. However, in actual applications, there are still a batch of terminal devices with insufficient processing power to perform inference.
  • Embodiments of the present application provide a method and device for AI model reasoning, which can be applied to wireless artificial intelligence (Artificial Intelligence, AI) systems.
  • AI Artificial Intelligence
  • the first device assists the second device.
  • the device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, thereby benefiting from wireless AI.
  • embodiments of the present application provide a method for AI model inference, which is executed by a first device.
  • the method includes:
  • the AI model inference request is when the second device responds to the need to provide inference results of the AI model or use inference results of the AI model. Sent to the first device.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • assisting the second device in performing AI model inference tasks includes any of the following:
  • the first device alone completes the AI model reasoning task
  • the first device and the second device jointly complete the AI model reasoning task
  • the first device, the second device, and the third device jointly complete the AI model inference task.
  • the method further includes:
  • the reasoning capability information of the AI model includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the method further includes:
  • the method further includes:
  • the AI model for inference In response to the AI model for inference being provided by the third device, the AI model forwarded by the second device is received.
  • the method further includes:
  • the AI model In response to the AI model for inference being provided by the first device, the AI model is sent to the second device, and the AI model is forwarded to the third device through the second device; or
  • the AI model In response to the AI model for performing inference being provided by the first device, the AI model is sent directly to the third device.
  • the method further includes:
  • the inference results are directly reported to the third device.
  • the method further includes:
  • the parameters further obtained based on the inference results are directly reported to the third device.
  • the protocol used by the first device to interact with the second device is a customized interaction protocol.
  • embodiments of the present application provide a method for artificial intelligence AI model inference, which is executed by a second device and includes:
  • an AI model inference request that needs to assist the second device in completing the AI model inference task is sent to the first device.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • the method further includes:
  • the method further includes:
  • the reasoning capability information includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the method further includes:
  • the method further includes:
  • the method further includes:
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • the protocol used by the second device to interact with the first device is a customized interaction protocol.
  • embodiments of the present application provide a method for artificial intelligence AI model inference, which method is executed by a third device.
  • the method includes:
  • the second device In response to receiving the information reported by the second device with AI model reasoning capabilities, sending an AI model reasoning task to the second device so that the first device assists the second device in completing the reasoning task, and the second device responds
  • the second device is required to provide the inference results of the AI model or use the inference results of the AI model to report information on the specific AI model's inference capabilities.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • the method further includes:
  • the method further includes:
  • the reasoning capability information of the AI model includes: AI model information, AI processing platform framework information, and AI processing capability information.
  • the method further includes:
  • the method further includes:
  • the AI model In response to the AI model for performing inference being provided by the third device, the AI model is sent to the second device, and the AI model is forwarded to the first device through the second device.
  • the method further includes:
  • the AI model for inference In response to the AI model for inference being provided by the first device, the AI model forwarded by the second device is received.
  • the first device and the second device are assisted to complete the AI model inference task.
  • the method further includes:
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • inventions of the present application provide an apparatus for AI model inference.
  • the apparatus is provided on a first device.
  • the apparatus includes:
  • a processing unit configured to assist the second device in completing the AI model inference task in response to receiving an AI model inference request sent by the second device.
  • the AI model inference request provides the inference result of the AI model or uses AI in response to the need of the second device.
  • the inference results of the model are sent to the first device.
  • assisting the second device in performing AI model inference tasks includes any of the following:
  • the first device alone completes the AI model reasoning task
  • the first device and the second device jointly complete the AI model reasoning task
  • the first device, the second device, and the third device jointly complete the AI model inference task.
  • the device further includes:
  • a sending unit configured to send the reasoning capability information of the first device to the AI model to the second device.
  • the reasoning capability information of the AI model includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • the reporting unit is configured to report the time-consuming information of processing the AI model inference task to the third device.
  • the device further includes:
  • a receiving unit configured to receive the AI model sent by the third device in response to the AI model for inference being provided by the third device; or.
  • the receiving unit is further configured to receive the AI model forwarded by the second device in response to the AI model for inference being provided by the third device.
  • the device further includes:
  • a sending unit configured to send the AI model to the second device in response to the AI model for inference being provided by the first device, and forward the AI model to the third device through the second device ;
  • the sending unit is also configured to directly send the AI model to the third device in response to the AI model for inference being provided by the first device.
  • the device further includes:
  • a sending unit configured to send the inference result to the second device, and the inference result is forwarded to the third device through the second device;
  • the inference results are directly reported to the third device.
  • the device further includes:
  • a sending unit configured to send parameters further obtained based on the inference results to the second device, and the parameters are forwarded to the third device through the second device;
  • a reporting unit is configured to directly report the parameters further obtained based on the inference results to the third device.
  • the protocol used by the first device to interact with the second device is a customized interaction protocol.
  • inventions of the present application provide a device for artificial intelligence AI model inference.
  • the device is provided on a second device and includes:
  • a sending unit configured to respond to the second device providing an inference result of the AI model or an inference result using the AI model, and sending an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • the device further includes:
  • a receiving unit configured to receive reasoning capability information sent by the first device to assist in AI model reasoning.
  • the device further includes:
  • a reporting unit is configured to report the reasoning capability information of the first device to assist in AI model reasoning to the third device.
  • the reasoning capability information includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • a receiving unit configured to respond to the AI model for inference being provided by the third device, receive the AI model sent by the third device, and forward the AI model to the first device.
  • the device further includes:
  • a receiving unit configured to respond to the AI model for inference being provided by the first device, receive the AI model sent by the first device, and forward the AI model to the third device.
  • the device further includes:
  • a receiving unit configured to receive the inference result of the AI model inference returned by the first device, and forward the inference result to the third device.
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • the protocol used by the second device to interact with the first device is a customized interaction protocol.
  • embodiments of the present application provide a device for artificial intelligence AI model inference, the device is provided in a third device, and is characterized in that the device includes:
  • the sending unit is configured to send an AI model inference task to the second device in response to receiving the information reported by the second device that has the AI model inference capability.
  • the device further includes:
  • a receiving unit configured to receive the reasoning capability information of the first device for the AI model sent by the second device.
  • the device further includes:
  • a receiving unit configured to receive the reasoning capability information of the second device for the AI model sent by the second device.
  • the reasoning capability information of the AI model includes: AI model information, AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • the receiving unit is configured to receive the time-consuming information of processing the AI model inference task reported by the first device.
  • the device further includes:
  • a sending unit configured to directly send the AI model to the first device in response to the AI model for inference being provided by the third device;
  • a sending unit configured to send the AI model to the second device in response to the AI model for inference being provided by the third device, and the AI model is forwarded to the first device through the second device.
  • the device further includes:
  • a receiving unit configured to receive the AI model sent by the first device in response to the AI model for inference being provided by the first device;
  • a receiving unit configured to respond to the AI model for inference being provided by the first device, and receive the AI model forwarded by the second device.
  • the processing unit is configured to assist the first device and the second device in completing the AI model inference task in response to receiving the AI model provided by the first device.
  • the device further includes:
  • a receiving unit configured to receive the inference result of the AI model sent by the second device.
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • inventions of the present application provide a reasoning 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 The device performs the method described in the first aspect.
  • inventions of the present application provide another reasoning 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 enable The device performs the method described in the second aspect.
  • inventions of the present application provide another reasoning 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 enable The device performs the method described in the third aspect.
  • embodiments of the present application provide another reasoning 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.
  • embodiments of the present application provide another reasoning 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 second aspect.
  • embodiments of the present application provide another reasoning 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 third aspect.
  • inventions of the present application provide another reasoning 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.
  • inventions of the present application provide another reasoning 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 second aspect.
  • inventions of the present application provide another reasoning 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 third aspect.
  • embodiments of the present application provide a reasoning system, including: a reasoning device as described in the seventh aspect, a reasoning device as described in the eighth aspect, and a reasoning device as described in the ninth aspect. ;
  • the system includes a reasoning device as described in the tenth aspect, a reasoning device as described in the eleventh aspect, and a reasoning device as described in the twelfth aspect;
  • the system includes a reasoning device as described in the thirteenth aspect, a reasoning device as described in the fourteenth aspect, and a reasoning device as described in the fifteenth aspect.
  • embodiments of the present application provide a computer-readable storage medium for storing instructions. When the instructions are executed, the method described in the first aspect is implemented.
  • embodiments of the present application provide another computer-readable storage medium for storing instructions that, when executed, enable the method described in the second aspect to be implemented.
  • embodiments of the present application provide another computer-readable storage medium for storing instructions. When the instructions are executed, the method described in the third aspect is implemented.
  • Figure 1 is a schematic architectural diagram of a reasoning system provided by an embodiment of the present application.
  • Figure 2 is a schematic flowchart of a reasoning method provided by an embodiment of the present application.
  • Figure 3 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 4 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 5 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 6 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 7 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 8 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 9 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 10 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 11 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 12 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 13 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 14 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 15 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 16 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 17 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 18 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 19 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 20 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 21 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 22 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 23 is a schematic flowchart of another reasoning method provided by an embodiment of the present application.
  • Figure 24 is a schematic structural diagram of a reasoning device provided by an embodiment of the present application.
  • Figure 25 is a schematic structural diagram of another reasoning device provided by an embodiment of the present application.
  • Figure 26 is a schematic structural diagram of another reasoning device provided by an embodiment of the present application.
  • Figure 27 is a schematic structural diagram of another reasoning device provided by an embodiment of the present application.
  • Figure 28 is a schematic structural diagram of another reasoning device provided by an embodiment of the present application.
  • Figure 1 is a schematic architectural diagram of a reasoning system provided by an embodiment of the present application.
  • the reasoning system may include but is not limited to a first device 101, a second device 102, and a third device 103.
  • 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. Practical applications may include two or more first devices 101, two or more second devices 102, and two or more third devices 103.
  • the system shown in Figure 1 includes a first device 101, a second device 102 and a third device 103.
  • the first device 101 in the embodiment of this application is a third-party AI processing platform, which is a server or processor other than a wireless cellular system.
  • the second device 102 in the embodiment of this application is an entity on the user side that is used to receive or transmit signals, such as a mobile phone.
  • the first device may also be called terminal equipment (terminal), user equipment (user equipment, UE), mobile station (mobile station, MS), mobile terminal equipment (mobile terminal, MT), etc.
  • the processing capability of the second device 102 is not sufficient to independently complete the AI model inference task.
  • the embodiments of this application do not limit the specific technology and specific device form used by the second device 102 .
  • the third device 103 in the embodiment of this application is a network device.
  • 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.
  • eNB evolved NodeB
  • TRP transmission reception point or transmit receive point
  • gNB next generation base station
  • WiFi wireless fidelity
  • 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 (central unit, CU) and a distributed unit (DU).
  • the CU may also be called a control unit (control unit), using CU-DU.
  • 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.
  • Figure 2 is a schematic flowchart of a reasoning method provided by an embodiment of the present application.
  • the method is executed by the first device, as shown in Figure 2.
  • the method may include but is not limited to the following steps:
  • Step S201 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the third device responds to receiving the information reported by the second device that it has the AI model reasoning capability, and sends the AI model reasoning task to the second device.
  • the second device does not have the conditions for independent reasoning, such as limited hardware conditions or AI processing platform When incompatible, the second device sends an AI model inference request to the first device, and the first device assists the second device in completing the AI model inference task.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • Figure 3 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S301 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the assisting the second device to perform the AI model inference task includes any of the following: the first device completes it alone The AI model reasoning task; the first device and the second device jointly complete the AI model reasoning task; the first device, the second device and the third device jointly complete the AI model reasoning task .
  • the AI model inference request is sent by the second device to the first device in response to a need to provide an inference result of the AI model or to use an inference result of the AI model.
  • the first device serves as the provider of the AI model, it can complete the model inference task alone, or the first device and the second device can jointly complete the model inference task.
  • the first device serves as the user of the AI model, it needs to jointly complete the model inference task after receiving the AI model transmitted by the third device.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • Figure 4 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S401 Send the reasoning capability information of the first device to the AI model to the second device.
  • the first device sends the reasoning capability information of the AI model to the second device.
  • the purpose is to use the second device as a relay to forward the obtained reasoning capability information to the third device to achieve information synchronization of the AI model during transmission. So that the third device determines whether to let the second device use the function of the wireless AI model or which use cases of the wireless AI model to use based on the reasoning capability information.
  • the third device responds to the AI model reasoning capability information reported by the second device and sends an AI model reasoning task to the second device.
  • the second device does not have the conditions for independent reasoning, such as limited hardware conditions or incompatible AI processing platforms
  • the first device assists the second device in completing the AI model inference task.
  • the reasoning capability information of the AI model includes: at least one of AI model information, AI processing platform framework information, and AI processing capability information.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Step S402 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds In order to receive the AI model inference request sent by the second device, assist the second device to complete the AI model inference task, so that the second device can respond to the need to provide the inference results of the AI model or use the inference results of the AI model, making the second device indirect Equipped with reasoning capabilities and benefit from wireless AI.
  • Figure 5 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S501 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Assisting the second device to perform the AI model reasoning task includes any of the following: the first device alone completes the AI model reasoning task, the first device and the second device jointly complete the AI model reasoning task, The first device, the second device and the third device jointly complete the AI model reasoning task.
  • Step S502 Report the time-consuming information of processing the AI model inference task to the third device.
  • the time-consuming information/delay information for processing each AI task is determined, and the time-consuming information/delay information is reported to the third device.
  • the third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds In order to receive the AI model inference request sent by the second device, assist the second device to complete the AI model inference task, so that the second device can respond to the need to provide the inference results of the AI model or use the inference results of the AI model, making the second device indirect Equipped with reasoning capabilities and benefit from wireless AI.
  • Figure 6 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S601 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Step S602 In response to the AI model for inference being provided by the third device, receive the AI model sent by the third device; or, in response to the AI model for inference being provided by the third device, receive the AI model sent by the third device. The AI model forwarded by the second device.
  • the first device when the first device serves as the user of the AI model and the third device serves as the provider of the AI model, the first device receives the AI model transmitted by the third device.
  • the transfer by the second device is also supported, that is, the third device serves as the provider of the AI model and transmits the AI model to the second device.
  • the second device transfers the AI model to the first device. That is, the transmission of the AI model is performed between the first device, the second device, and the third device.
  • the third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds In order to receive the AI model inference request sent by the second device, perform AI model transmission in at least two of the first device, the second device and the third device to complete the model inference task of the second device, so that the third device
  • the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • Figure 7 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S701 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Step S702 In response to the AI model for inference being provided by the first device, send the AI model to the second device, and the AI model is forwarded to the third device through the second device; or In response to the AI model for performing inference being provided by the first device, the AI model is sent directly to the third device.
  • the scenario of the embodiment of this application is that the first device serves as the provider of the AI model and needs to transmit it to the user of the AI model (the third device).
  • the third device assists the first device in performing AI model inference based on the received AI model. Task.
  • the process is similar to the process of the third device transmitting the AI model to the first device.
  • the first device is used to directly transmit the AI model from the first device to the third device to the third device.
  • the second device transmits the AI model to the third device.
  • the embodiment of the present disclosure does not specifically limit the method of transmitting the AI model.
  • the third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds In order to receive the AI model inference request sent by the second device, perform AI model transmission in at least two of the first device, the second device and the third device to complete the model inference task of the second device, so that the third device
  • the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • FIG. 8 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S801 In response to receiving the AI model inference request sent by the second device, assist the second device to complete the AI model inference task.
  • the AI model inference request is the second device's response to the need to provide inference results of the AI model or use the AI model.
  • the inference result is sent to the first device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Step S802 Send the inference result to the second device, and forward the inference result to the third device through the second device; or directly report the inference result to the third device.
  • the inference result is returned to the second device, and the second device uploads it to the third device.
  • the inference result is directly returned to the third device.
  • the network device in the embodiment of this application is an entity on the network side that is used to transmit or receive signals.
  • the network equipment can be an evolved base station (evolved NodeB, eNB), a transmission point (transmission reception point, TRP), a next generation base station (next generation NodeB, gNB) in an NR system, a base station in other future mobile communication systems, or Access nodes in wireless fidelity (WiFi) systems, etc.
  • the embodiments of this application do not limit the specific technology and specific equipment form used by the network equipment.
  • the network equipment provided by the embodiments of this application may be composed of a centralized unit (central 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 third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds Receive the AI model inference request sent by the second device, and return the inference results to the second device or the third device to assist the second device in completing the AI model inference task, so that the second device can indirectly have reasoning capabilities and benefit from wireless AI.
  • Figure 9 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the first device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S901 In response to the second device providing or using the inference result based on the AI model, the first device assists the second device in performing the AI model inference task.
  • the AI model inference task is completed by the first device alone and is performed by the first device.
  • the first device and the second device are completed together or are completed by the first device, the second device and their third device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • Step S902 Send the parameters further obtained based on the inference results to the second device, and the parameters are forwarded to the third device through the second device; or send the parameters further obtained based on the inference results.
  • the parameters are directly reported to the third device.
  • the third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds to receive the AI model inference request sent by the second device, and return the further parameters obtained from the inference result to the second device or the third device to assist the second device in completing the AI model inference task, so that the second device has the inference capability indirectly, Benefit from wireless AI.
  • the embodiment of the present disclosure provides another reasoning method, which is executed by the first device.
  • the reasoning method can be executed alone, or can be combined with any embodiment in the present disclosure or the possible implementations in the embodiment.
  • the method can be executed together, and can also be executed in combination with any technical solution in related technologies.
  • a new AI inference processing architecture including a first device, a second device, and a third device.
  • the protocol for interaction between the first device and the second device is that the first device interacts with the second device.
  • the second device customizes the interaction protocol, and the protocol between the first device and the third device is a universal interaction protocol.
  • Figure 10 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1001 In response to the second device providing an inference result of the AI model or an inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • the third device In response to receiving the information reported by the second device with AI model reasoning capabilities, the third device sends an AI model reasoning task to the second device.
  • the second device does not have the conditions for independent reasoning, the second device sends the required information to the first device.
  • the first device assists the second device in completing the AI model inference task by assisting the second device in completing the AI model inference task.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • the second device is a device that does not have the conditions for independent reasoning, such as limited hardware conditions or incompatibility of the AI processing platform.
  • the third device sends an AI model inference task to the second device.
  • the second device does not have the conditions for independent reasoning
  • the second device sends to the first device the need to assist the second device in completing the AI model inference task.
  • the first device assists the second device in completing the AI model inference task, so that the second device can respond to the need to provide the inference results of the AI model or use the inference results of the AI model, so that the second device is indirectly capable of inference. capabilities, benefiting from wireless AI.
  • Figure 11 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1101 Receive reasoning capability information sent by the first device to assist in AI model reasoning.
  • the first device sends the reasoning capability information of the AI model to the second device.
  • the purpose is for the second device to report the acquired reasoning capability information to the third device.
  • the third device infers the reasoning capability information based on the received AI model. Configure the second device to perform AI inference tasks.
  • the reasoning capability information of the AI model includes at least one of AI model information, AI processing platform framework information, and AI processing capability information.
  • AI model information For example, the type of AI model supported is convolutional neural network (Deep Convolutional Neural Network, CNN), recurrent neural network (Recurrent Neural Network, RNN), or transform, etc. Is the AI processing framework tenserflow or Pytorch, calculation speed, etc.
  • Step S1102 In response to the second device providing the inference result of the AI model or the inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • the reasoning capability information of the first device to assist in AI model reasoning is reported to the third device, and the third device configures the second device to perform AI reasoning based on the received reasoning capability information of AI model reasoning.
  • the second device when the second device does not have the conditions for independent reasoning, the second device sends an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task, and the first device assists the second device in completing the AI model inference.
  • the task enables the second device to provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has inference capabilities and benefits from wireless AI.
  • Figure 12 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1201 Report the reasoning capability information of the first device to assist in AI model reasoning to the third device.
  • the first device can be used to assist the second device in performing AI reasoning.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • the first device sends the reasoning capability information of the AI model to the second device.
  • the purpose is for the second device to report the acquired reasoning capability information to the third device.
  • the third device infers the reasoning capability information based on the received AI model. Configure the second device to perform AI inference tasks.
  • the second device acts as a relay and forwards the obtained inference capability information to the third device to achieve information synchronization of the AI model during transmission, so that the third device determines whether to allow the second device to use the function of the wireless AI model or based on the inference capability information. Use cases for which wireless AI models to use.
  • the reasoning capability information of the AI model includes at least one of AI model information, AI processing platform framework information, and AI processing capability information.
  • AI model information For example, the type of AI model supported is convolutional neural network (Deep Convolutional Neural Network, CNN), recurrent neural network (Recurrent Neural Network, RNN), or transform, etc. Is the AI processing framework tenserflow or Pytorch, calculation speed, etc.
  • Step S1202 In response to the second device providing the inference result of the AI model or the inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • the reasoning capability information of the first device to assist in AI model reasoning is reported to the third device, and the third device configures the second device to perform AI reasoning based on the received reasoning capability information of AI model reasoning.
  • the second device when the second device does not have the conditions for independent reasoning, the second device sends an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task, and the first device assists the second device in completing the AI model inference.
  • the task enables the second device to provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has inference capabilities and benefits from wireless AI.
  • Figure 13 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1301 In response to the second device providing an inference result of the AI model or an inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • Step S1302 In response to the AI model for inference being provided by the third device, receive the AI model sent by the third device, and forward the AI model to the first device.
  • the first device when the first device serves as the user of the AI model and the third device serves as the provider of the AI model, the first device receives the AI model transmitted by the third device.
  • the transfer by the second device is also supported, that is, the third device serves as the provider of the AI model and transmits the AI model to the second device.
  • the second device transfers the AI model to the first device. That is, the transmission of the AI model is performed between the first device, the second device, and the third device.
  • the AI model sent by the third device in response to the AI model for inference being provided by the third device, the AI model sent by the third device is received, and the AI model is forwarded to the first device, and in the second
  • the second device sends an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task, and the first device assists the second device in completing the AI model inference task, so that the second device
  • the device can provide inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has inference capabilities and benefits from wireless AI.
  • Figure 14 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1401 In response to the second device providing an inference result of the AI model or an inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • Step S1402 In response to the AI model for inference being provided by the first device, receive the AI model sent by the first device, and forward the AI model to the third device.
  • the scenario of the embodiment of this application is that the first device serves as the provider of the AI model and needs to transmit it to the user of the AI model (the third device).
  • the third device assists the first device in performing AI model inference based on the received AI model. Task.
  • the process is similar to the process of the third device transmitting the AI model to the first device.
  • the first device is used to directly transmit the AI model from the first device to the third device to the third device.
  • the second device transmits the AI model to the third device.
  • the embodiment of the present disclosure does not specifically limit the method of transmitting the AI model.
  • the third device sends an AI model inference task to the second device in response to receiving information reported by the second device that has AI model inference capabilities.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the third device's AI model inference capability.
  • the AI model inference request sent by the second device is executed in at least two of the first device, the second device and the third device to complete the model inference task of the second device and make the second device indirectly Sex has reasoning capabilities and benefits from wireless AI.
  • Figure 15 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by the second device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1501 In response to the second device providing an inference result of the AI model or an inference result using the AI model, send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task.
  • Step S1502 Receive the inference result of the AI model inference returned by the first device, and forward the inference result to the third device.
  • the inference result is returned to the second device, and the second device uploads it to the third device.
  • the inference result is directly returned to the third device.
  • the network device in the embodiment of this application is an entity on the network side that is used to transmit or receive signals.
  • the network equipment can be an evolved base station (evolved NodeB, eNB), a transmission point (transmission reception point, TRP), a next generation base station (next generation NodeB, gNB) in an NR system, a base station in other future mobile communication systems, or Access nodes in wireless fidelity (WiFi) systems, etc.
  • the embodiments of this application do not limit the specific technology and specific equipment form used by the network equipment.
  • the network equipment provided by the embodiments of this application may be composed of a centralized unit (central 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 third device responds to receiving information reported by the second device that has AI model reasoning capabilities, and sends an AI model reasoning task to the second device.
  • the first device responds Receive the AI model inference request sent by the second device, and return the inference results to the second device or the third device to assist the second device in completing the AI model inference task, so that the second device can indirectly have reasoning capabilities and benefit from wireless AI.
  • the embodiment of the present disclosure provides another reasoning method, which is applied to the third device side.
  • Figure 16 is a flowchart of another reasoning method provided by the embodiment of the present disclosure. The method is executed by the third device. The reasoning method can be executed alone, or in combination with any embodiment or possible implementation in the embodiment, or in combination with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1601 In response to receiving information with AI model inference capabilities reported by the second device, send an AI model inference task to the second device.
  • the third device sends an AI model inference task to the second device.
  • the second device does not have the conditions for independent inference, such as when the hardware conditions are limited or the AI processing platform is incompatible
  • the second device sends an AI model inference request to the first device.
  • the first device assists the second device in completing the AI model inference task.
  • the first device is a server or processor outside the wireless cellular system.
  • the specific equipment form of the first equipment is not limited.
  • the third device sends an AI model inference task to the second device.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the AI model inference request sent by the second device and assists the second device to complete the task.
  • the AI model inference task enables the second device to provide inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • Figure 17 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1701 Receive the reasoning capability information of the first device for the AI model sent by the second device.
  • Step S1702 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • the second device When the second device does not have the conditions for independent reasoning, so that the first device assists the second device in completing the reasoning task, the second device responds to the need for the second device to provide inference results of the AI model or use AI
  • the model's inference results report information on the specific AI model's reasoning capabilities.
  • the first device sends the reasoning capability information of the AI model to the second device.
  • the purpose is for the second device to report the acquired reasoning capability information to the third device.
  • the third device infers the reasoning capability information based on the received AI model.
  • the second device acts as a relay and forwards the obtained inference capability information to the third device to achieve information synchronization of the AI model during transmission, so that the third device determines whether to allow the second device to use the function of the wireless AI model or based on the inference capability information.
  • the reasoning capability information of the AI model includes at least one of AI model information, AI processing platform framework information, and AI processing capability information.
  • AI model information For example, the type of AI model supported is convolutional neural network (Deep Convolutional Neural Network, CNN), recurrent neural network (Recurrent Neural Network, RNN), or transform, etc. Is the AI processing framework tenserflow or Pytorch, calculation speed, etc.
  • the reasoning capability information of the first device to assist in AI model reasoning is reported to the third device, and the third device configures the second device to perform AI reasoning based on the received reasoning capability information of AI model reasoning.
  • the second device when the second device does not have the conditions for independent reasoning, the second device sends an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task, and the first device assists the second device in completing the AI model inference.
  • the task enables the second device to provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has inference capabilities and benefits from wireless AI.
  • Figure 18 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1801 Receive the inference capability information of the second device on the AI model sent by the second device.
  • the second device reports the inference capability information to the third device, and the third device configures the second device to perform the AI inference task according to the received inference capability information of AI model inference.
  • the second device's reasoning capability information for the AI model sent by the second device may include, but is not limited to, the reasoning capability information of the AI model reasoning provided by the first device, and may also include the reasoning capability information of the second device's own AI model reasoning. Capability information, the third device does not pay attention to the source of the reasoning capability information for AI model reasoning, but is based on the fact that the reasoning capability information for AI model reasoning provided by the second device can perform the AI reasoning task.
  • the reasoning capability information of the AI model includes at least one of AI model information, AI processing platform framework information, and AI processing capability information.
  • AI model information For example, the type of AI model supported is convolutional neural network (Deep Convolutional Neural Network, CNN), recurrent neural network (Recurrent Neural Network, RNN), or transform, etc. Is the AI processing framework tenserflow or Pytorch, calculation speed, etc.
  • Step S1802 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • the first device assists the second device in completing the inference task
  • the second device responds to the need for the second device to provide inference results of the AI model or use the inference results of the AI model to report information on the specific AI model inference capabilities.
  • the reasoning capability information of the first device to assist in AI model reasoning is reported to the third device, and the third device configures the second device to perform AI reasoning based on the received reasoning capability information of AI model reasoning.
  • the second device sends an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task, and the first device assists the second device in completing the AI model inference task, so that the second device can provide AI in response to the need
  • the inference results of the model or the inference results using the AI model enable the second device to indirectly have inference capabilities and benefit from wireless AI.
  • Figure 19 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S1901 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • Step S1902 Receive the time-consuming information of processing the AI model inference task reported by the first device.
  • the first device determines the time-consuming information/delay information for processing each AI task according to the category of the AI task processed in the AI model, and reports the time-consuming information/delay information to the third device.
  • the third device sends an AI model inference task to the second device in response to receiving information reported by the second device that has AI model inference capabilities.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the third device's AI model inference capability.
  • the AI model inference request sent by the second device assists the second device in completing the AI model inference task, so that the second device can respond to the need to provide the inference results of the AI model or use the inference results of the AI model, so that the second device indirectly has the ability to reason. , benefiting from wireless AI.
  • Figure 20 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S2001 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • Step S2002 In response to the AI model for inference being provided by the third device, directly send the AI model to the first device; or in response to the AI model for inference being provided by the third device, sending the AI model to the first device.
  • the AI model is sent to the second device, and the AI model is forwarded to the first device through the second device.
  • the scenario of the embodiment of this application is that the third device serves as the provider of the AI model and needs to transmit it to the user of the AI model (the first device).
  • the first device assists the first device in performing AI model inference based on the received AI model. Task.
  • the scenario of the embodiment of this application is that the third device, as the provider of the AI model, needs to transmit it to the second device, and then the second device forwards it to the user of the AI model (the first device), so as to implement the first device , the second device and the third device jointly perform the AI model inference task.
  • the third device sends an AI model inference task to the second device in response to receiving information reported by the second device that has AI model inference capabilities.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the third device's AI model inference capability.
  • the AI model inference request sent by the second device assists the second device in completing the AI model inference task, so that the second device can respond to the need to provide the inference results of the AI model or use the inference results of the AI model, so that the second device indirectly has the ability to reason. , benefiting from wireless AI.
  • Figure 21 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S2101 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • Step S2102 In response to the AI model for inference being provided by the first device, receive the AI model sent by the first device; or in response to the AI model for inference being provided by the first device, receive the The AI model forwarded by the second device.
  • Figure 22 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S2001 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • Step S2202 In response to receiving the AI model provided by the first device, assist the first device and the second device to complete the AI model inference task.
  • Figure 23 is a schematic flowchart of another reasoning method provided by the embodiment of the present disclosure.
  • the method is executed by a third device.
  • the reasoning method can be executed separately, or It can be executed in conjunction with any embodiment or possible implementation in the embodiment, and can also be executed in conjunction with any technical solution in related technologies.
  • the reasoning method may include the following steps:
  • Step S2302 In response to receiving the information reported by the second device with AI model inference capabilities, send an AI model inference task to the second device.
  • Step S2302 Receive the inference result of the AI model inference returned by the first device, and forward the inference result to the third device.
  • the inference result is: an inference result obtained by the first device alone completing the AI model inference task; or an inference result obtained by the first device and the second device jointly completing the AI model inference task. ; Or the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • the third device sends an AI model inference task to the second device in response to receiving information reported by the second device that has AI model inference capabilities.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the third device's AI model inference capability.
  • the second device sends an AI model inference request and returns the inference results to the third device, assisting the second device in completing the AI model inference task, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • the present disclosure also provides a reasoning device. Since the reasoning device provided by the embodiments of the present disclosure is the same as the reasoning method provided by the above-mentioned embodiments of FIGS. 2 to 23 Correspondingly, therefore, the implementation of the reasoning method is also applicable to the reasoning device provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.
  • Figure 24 is a schematic structural diagram of a reasoning device provided by an embodiment of the present disclosure.
  • the device is provided on the first device, and the device includes:
  • the processing unit 2401 is configured to assist the second device in completing the AI model inference task in response to receiving an AI model inference request sent by the second device.
  • the AI model inference request provides the inference result or use of the AI model in response to the need of the second device.
  • the inference results of the AI model are sent to the first device.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • assisting the second device in performing the AI model inference task includes any of the following:
  • the first device alone completes the AI model reasoning task
  • the first device and the second device jointly complete the AI model reasoning task
  • the first device, the second device, and the third device jointly complete the AI model inference task.
  • the device further includes:
  • the sending unit 2402 is configured to send the reasoning capability information of the first device to the AI model to the second device.
  • the reasoning capability information of the AI model includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • the reporting unit 2403 is configured to report the time-consuming information of processing the AI model inference task to the third device.
  • the device further includes:
  • the receiving unit 2404 is configured to receive the AI model sent by the third device in response to the AI model for inference being provided by the third device; or.
  • the receiving unit 2404 is further configured to receive the AI model forwarded by the second device in response to the AI model for inference being provided by the third device.
  • the device further includes:
  • the sending unit 2402 is configured to send the AI model to the second device in response to the AI model for inference being provided by the first device, and the AI model is forwarded to the third device through the second device. equipment; or
  • the sending unit 2402 is also configured to directly send the AI model to the third device in response to the AI model for inference being provided by the first device.
  • the device further includes:
  • Sending unit 2402 configured to send the inference result to the second device, and the inference result is forwarded to the third device through the second device;
  • the inference results are directly reported to the third device.
  • the device further includes:
  • Sending unit 2402 configured to send parameters further obtained based on the inference results to the second device, and the parameters are forwarded to the third device through the second device;
  • the reporting unit 2403 is configured to directly report the parameters further obtained based on the inference results to the third device.
  • the protocol for interaction between the first device and the second device is a customized interaction protocol.
  • the embodiment of the present application provides a device for artificial intelligence AI model inference.
  • the device is provided on the second device, as shown in Figure 25, and includes:
  • the sending unit 2501 is configured to send an AI model inference request to the first device that needs to assist the second device in completing the AI model inference task in response to the second device providing the inference result of the AI model or the inference result using the AI model.
  • the third device sends an AI model inference task to the second device.
  • the second device does not have the conditions for independent reasoning, it responds to receiving the AI model inference request sent by the second device and assists the second device to complete the task.
  • the AI model inference task enables the second device to provide inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • the device further includes:
  • the receiving unit 2502 is configured to receive reasoning capability information sent by the first device to assist in AI model reasoning.
  • the device further includes:
  • the reporting unit 2503 is configured to report the reasoning capability information of the first device to assist in AI model reasoning to the third device.
  • the reasoning capability information includes:
  • AI model information AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • the receiving unit 2502 is configured to respond to the AI model for inference being provided by the third device, receive the AI model sent by the third device, and forward the AI model to the first device.
  • the device further includes:
  • the receiving unit 2502 is configured to respond to the AI model for inference being provided by the first device, receive the AI model sent by the first device, and forward the AI model to the third device.
  • the device further includes:
  • the receiving unit 2502 is configured to receive the inference result of the AI model inference returned by the first device, and forward the inference result to the third device.
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • the protocol for the second device to interact with the first device is a customized interaction protocol.
  • the embodiment of the present application provides a device for artificial intelligence AI model inference.
  • the device is provided on a third device. As shown in Figure 26, the device includes:
  • the sending unit 2601 is configured to send an AI model reasoning task to the second device in response to receiving the information reported by the second device with AI model reasoning capabilities, so that the first device assists the second device in completing the reasoning task,
  • the second device responds to the requirement that the second device provide an inference result of the AI model or use the inference result of the AI model to report information on a specific AI model inference capability.
  • the third device sends an AI model inference task to the second device.
  • the first device responds to receiving the AI model inference request sent by the second device and assists the third device.
  • the second device completes the AI model inference task, so that the second device can provide the inference results of the AI model or use the inference results of the AI model in response to the need, so that the second device indirectly has reasoning capabilities and benefits from wireless AI.
  • the device further includes:
  • the receiving unit 2602 is configured to receive the reasoning capability information of the first device for the AI model sent by the second device.
  • the device further includes:
  • the receiving unit 2602 is configured to receive the reasoning capability information of the second device for the AI model sent by the second device.
  • the reasoning capability information of the AI model includes: AI model information, AI processing platform framework information, and AI processing capability information.
  • the device further includes:
  • the receiving unit 2602 is configured to receive the time-consuming information of processing the AI model inference task reported by the first device.
  • the device further includes:
  • the sending unit 2603 is configured to directly send the AI model to the first device in response to the AI model for inference being provided by the third device; or
  • the sending unit 2603 is configured to send the AI model to the second device in response to the AI model for inference being provided by the third device, and the second device forwards the AI model to the third device.
  • the device further includes:
  • the receiving unit 2602 is configured to receive the AI model sent by the first device in response to the AI model for inference being provided by the first device; or
  • the receiving unit 2602 is configured to respond to the AI model for inference being provided by the first device, and receive the AI model forwarded by the second device.
  • the processing unit 2601 is configured to assist the first device and the second device in completing the AI model inference task in response to receiving the AI model provided by the first device. .
  • the device further includes:
  • the receiving unit 2602 is configured to receive the inference result of the AI model sent by the second device.
  • the reasoning result is:
  • the inference result obtained by the first device and the second device jointly completing the AI model inference task;
  • the inference result obtained by the first device, the second device and its third device jointly completing the AI model inference task.
  • the present disclosure also provides another reasoning 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 shown in Figures 2 to 9, or to perform the methods shown in Figures 10 to 15, or to perform the methods shown in Figures 16 to 23.
  • the first device, the second device and the third device may include a hardware structure and a software module, in the form of a hardware structure, a software module, or a hardware structure plus a software module.
  • a hardware structure in the form of a hardware structure, a software module, or a hardware structure plus a software module.
  • One of the above functions can be executed by a hardware structure, a software module, or a hardware structure plus a software module.
  • network device 2700 includes processing component 2722, which further includes at least one processor, and memory resources represented by memory 2732 for storing instructions, such as application programs, executable by processing component 2722.
  • the application program stored in memory 2732 may include one or more modules, each corresponding to a set of instructions.
  • the processing component 2722 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 FIG. 2 to FIG. 21 .
  • Network device 2700 may also include a power supply component 2706 configured to perform power management of network device 2700, a wired or wireless network interface 2750 configured to connect network device 2700 to a network, and an input-output (I/O) interface 2758 .
  • Network device 2700 may operate based on an operating system stored in memory 2732, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or similar.
  • embodiments of the present application provide a reasoning system, including: a reasoning device as shown in Figure 24, a reasoning device as shown in Figure 25, and a reasoning device as shown in Figure 26.
  • Figure 28 is a block diagram of a reasoning device provided by an embodiment of the present disclosure.
  • user device 2800 may be a mobile phone, computer, digital broadcast user device, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, etc.
  • user equipment 2800 may include at least one of the following components: a processing component 2802, a memory 2804, a power supply component 2806, a multimedia component 2808, an audio component 2810, an input/output (I/O) interface 2812, a sensor component 2814, and Communication component 2816.
  • Processing component 2802 generally controls the overall operations of user device 2800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 2802 may include at least one processor 2820 to execute instructions to complete all or part of the steps of the above method. Additionally, processing component 2802 may include at least one module that facilitates interaction between processing component 2802 and other components. For example, processing component 2802 may include a multimedia module to facilitate interaction between multimedia component 2808 and processing component 2802.
  • Memory 2804 is configured to store various types of data to support operations at user device 2800. Examples of such data include instructions for any application or method operating on user device 2800, contact data, phonebook data, messages, pictures, videos, etc.
  • Memory 2804 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 2806 provides power to various components of user equipment 2800.
  • Power supply components 2806 may include a power management system, at least one power supply, and other components associated with generating, managing, and distributing power to user device 2800.
  • Multimedia component 2808 includes a screen that provides an output interface between the user device 2800 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 2808 includes a front-facing camera and/or a rear-facing camera.
  • the front camera and/or the rear camera can 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 2810 is configured to output and/or input audio signals.
  • audio component 2810 includes a microphone (MIC) configured to receive external audio signals when user device 2800 is in operating modes, such as call mode, recording mode, and speech recognition mode. The received audio signals may be further stored in memory 2804 or sent via communications component 2816.
  • audio component 2810 also includes a speaker for outputting audio signals.
  • the I/O interface 2812 provides an interface between the processing component 2802 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.
  • Sensor component 2814 includes at least one sensor for providing various aspects of status assessment for user device 2800 .
  • the sensor component 2814 can detect the open/closed state of the user device 2800, the relative positioning of components, such as the display and keypad of the user device 2800, the sensor component 2814 can also detect the user device 2800 or a user device 2800. Changes in position of components, presence or absence of user contact with user device 2800 , user device 2800 orientation or acceleration/deceleration and changes in temperature of user device 2800 .
  • Sensor component 2814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
  • Sensor assembly 2814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 2814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • Communication component 2815 is configured to facilitate wired or wireless communication between user device 2800 and other devices.
  • User equipment 2800 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 2815 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communications component 2815 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 user equipment 2800 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 2804 including instructions, which can be executed by the processor 2820 of the user device 2800 to complete the above-described FIGS. 2 to 21 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 may also be other names that can be understood by the inference device, and the values or expressions of the parameters can also be other values or expressions that can be understood by the inference device.
  • 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)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请实施例公开了一种推理的方法及其装置,可以应用于应用于无线人工智能(Artificial Intelligence,AI)系统中,该方法包括:在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。

Description

一种AI模型推理的方法及其装置 技术领域
本申请涉及通信技术领域,尤其涉及一种AI模型推理的方法及其装置。
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。
相关技术中,AI技术的主要参与方主要是基站和终端设备,由基站提供AI模型,由终端进行推理,由于终端设备进行推理需要终端设备具有一定的硬件能力和软件平台,通常需要处理能力比较高端的终端设备才能进行,但是,在实际的应用中,仍然存在一批处理能力不足以执行推理终端设备。
发明内容
本申请实施例提供一种AI模型推理的方法及其装置,可以应用于无线人工智能(Artificial Intelligence,AI)系统中,在第二设备不具有独立推理的条件时,通过第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,实现受益于无线AI。
第一方面,本申请实施例提供一种AI模型推理的方法,该方法被第一设备执行,该方法包括:
响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
在一种实现方式中,所述辅助第二设备执行AI模型推理任务包括以下任一种:
所述第一设备单独完成所述AI模型推理任务;
所述第一设备与所述第二设备共同完成所述AI模型推理任务;
所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
在一种实现方式中,所述方法还包括:
将所述第一设备对所述AI模型的推理能力信息发送至所述第二设备。
在一种实现方式中,所述AI模型的推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述方法还包括:
将处理AI模型推理任务的耗时信息上报给所述第三设备。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型;或者。
响应于进行推理的AI模型由所述第三设备提供,接收所述第二设备转发的所述AI模型。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第一设备提供,向所述第二设备发送的所述AI模型,所述AI模型通过所述第二设备转发至所述第三设备;或者
响应于进行推理的AI模型由所述第一设备提供,直接向所述第三设备发送的所述AI模型。
在一种实现方式中,所述方法还包括:
将所述推理结果发送至所述第二设备,所述推理结果通过所述第二设备转发至所述第三设备;或者
将所述推理结果直接上报至所述第三设备。
在一种实现方式中,所述方法还包括:
将基于所述推理结果进一步得到的参数发送至所述第二设备,所述参数通过所述第二设备转发至所述第三设备;或者
将所述基于所述推理结果进一步得到的参数直接上报至所述第三设备。
在一种实现方式中,所述第一设备与所述第二设备进行交互的协议为自定义的交互协议。
第二方面,本申请实施例提供一种人工智能AI模型推理的方法,该方法被第二设备执行,包括:
响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
在一种实现方式中,所述方法还包括:
接收所述第一设备发送的辅助进行AI模型推理的推理能力信息。
在一种实现方式中,所述方法还包括:
将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备。
在一种实现方式中,所述推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型,并将所述AI模型转发至所述第三设备。
在一种实现方式中,所述方法还包括:
接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
在一种实现方式中,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
在一种实现方式中,所述第二设备与所述第一设备进行交互的协议为自定义的交互协议。
第三方面,本申请实施例提供一种人工智能AI模型推理的方法,该方法被第三设备执行,所述方法包括:
响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务,以便第一设备辅助所述第二设备完成所述推理任务,所述第二设备响应于需要所述第二设备提供AI模型的推理结果或使用AI模型的推理结果上报具体AI模型推理能力的信息。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
在一种实现方式中,所述方法还包括:
接收所述第二设备发送的所述第一设备对所述AI模型的推理能力信息。
在一种实现方式中,所述方法还包括:
接收所述第二设备发送的所述第二设备对所述AI模型的推理能力信息。
在一种实现方式中,所述AI模型的推理能力信息包括;AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述方法还包括:
接收所述第一设备上报的处理AI模型推理任务的耗时信息。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第三设备提供,直接将所述AI模型发送至所述第一设备;或者
响应于进行推理的AI模型由所述第三设备提供,将所述所述AI模型发送至所述第二设备,所述AI模型通过第二设备转发至所述第一设备。
在一种实现方式中,所述方法还包括:
响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型;或者
响应于进行推理的AI模型由所述第一设备提供,接收所述第二设备转发的所述AI模型。
在一种实现方式中,响应于接收到第一设备提供的AI模型,辅助所述第一设备、所述第二设备完成所述AI模型推理任务。
在一种实现方式中,所述方法还包括:
接收所述第二设备发送的AI模型的推理结果。
在一种实现方式中,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
第四方面,本申请实施例提供一种AI模型推理的装置,所述装置被设置在第一设备,所述装置包括:
处理单元,用于响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在一种实现方式中,所述辅助第二设备执行AI模型推理任务包括以下任一种:
所述第一设备单独完成所述AI模型推理任务;
所述第一设备与所述第二设备共同完成所述AI模型推理任务;
所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
在一种实现方式中,所述装置还包括:
发送单元,用于将所述第一设备对所述AI模型的推理能力信息发送至所述第二设备。
在一种实现方式中,所述AI模型的推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述装置还包括:
上报单元,用于将处理AI模型推理任务的耗时信息上报给所述第三设备。
在一种实现方式中,所述装置还包括:
接收单元,用于响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型;或者。
所述接收单元,还用于响应于进行推理的AI模型由所述第三设备提供,接收所述第二设备转发的所述AI模型。
在一种实现方式中,所述装置还包括:
发送单元,用于响应于进行推理的AI模型由所述第一设备提供,向所述第二设备发送的所述AI模型,所述AI模型通过所述第二设备转发至所述第三设备;或者
发送单元,还用于响应于进行推理的AI模型由所述第一设备提供,直接向所述第三设备发送的所述AI模型。
在一种实现方式中,所述装置还包括:
发送单元,用于将所述推理结果发送至所述第二设备,所述推理结果通过所述第二设备转发至所述第三设备;或者
将所述推理结果直接上报至所述第三设备。
在一种实现方式中,所述装置还包括:
发送单元,用于将基于所述推理结果进一步得到的参数发送至所述第二设备,所述参数通过所述第二设备转发至所述第三设备;或者
上报单元,用于将所述基于所述推理结果进一步得到的参数直接上报至所述第三设备。
在一种实现方式中,所述第一设备与所述第二设备进行交互的协议为自定义的交互协议。
第五方面,本申请实施例提供一种人工智能AI模型推理的装置,所述装置被设置在第二设备,包括:
发送单元,用于响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第一设备发送的辅助进行AI模型推理的推理能力信息。
在一种实现方式中,所述装置还包括:
上报单元,用于将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备。
在一种实现方式中,所述推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述装置还包括:
接收单元,用于响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备。
在一种实现方式中,所述装置还包括:
接收单元,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型,并将所述AI模型转发至所述第三设备。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
在一种实现方式中,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
在一种实现方式中,所述第二设备与所述第一设备进行交互的协议为自定义的交互协议。
第六方面,本申请实施例提供一种人工智能AI模型推理的装置,所述装置被设置在第三设备,其特征在于,所述装置包括:
发送单元,用于响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第二设备发送的所述第一设备对所述AI模型的推理能力信息。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第二设备发送的所述第二设备对所述AI模型的推理能力信息。
在一种实现方式中,所述AI模型的推理能力信息包括;AI模型信息、AI处理平台框架信息以及AI处理能力信息。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第一设备上报的处理AI模型推理任务的耗时信息。
在一种实现方式中,所述装置还包括:
发送单元,用于响应于进行推理的AI模型由所述第三设备提供,直接将所述AI模型发送至所述第一设备;或者
发送单元,用于响应于进行推理的AI模型由所述第三设备提供,将所述所述AI模型发送至所述第二设备,所述AI模型通过第二设备转发至所述第一设备。
在一种实现方式中,所述装置还包括:
接收单元,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型;或者
接收单元,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第二设备转发的所述AI模型。
在一种实现方式中,处理单元,用于响应于接收到第一设备提供的AI模型,辅助所述第一设备、所述第二设备完成所述AI模型推理任务。
在一种实现方式中,所述装置还包括:
接收单元,用于接收所述第二设备发送的AI模型的推理结果。
在一种实现方式中,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
第七方面,本申请实施例提供一种推理的装置,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所述的方法。
第八方面,本申请实施例提供另一种推理的装置,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二方面所述的方法。
第九方面,本申请实施例提供另一种推理的装置,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三方面所述的方法。
第十方面,本申请实施例提供另一种推理的装置,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所述的方法。
第十一方面,本申请实施例提供另一种推理的装置,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二方面所述的方法。
第十二方面,本申请实施例提供另一种推理的装置,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三方面所述的方法。
第十三方面,本申请实施例提供另一种推理的装置,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所述的方法。
第十四方面,本申请实施例提供另一种推理的装置,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二方面所述的方法。
第十五方面,本申请实施例提供另一种推理的装置,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三方面所述的方法。
第十六方面,本申请实施例提供一种推理的系统,包括:如第七方面所述的推理的装置、如第八方面所述的推理的装置以及如第九方面所述的推理的装置;
或者,所述系统包括如第十方面所述的推理的装置、如第十一方面所述的推理的装置以及第十二方面所述的推理的装置;
或者,所述系统包括如第十三方面所述的推理的装置、如第十四方面所述的推理的装置以及第十五方面所述的推理的装置。
第十七方面,本申请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,使如第一方面所述的方法被实现。
第十八方面,本申请实施例提供另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,使如第二方面所述的方法被实现。
第十九方面,本申请实施例提供另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,使如第三方面所述的方法被实现。
附图说明
为了更清楚地说明本申请实施例或背景技术中的技术方案,下面将对本申请实施例或背景技术中所需要使用的附图进行说明。
图1是本申请实施例提供的一种推理的系统的架构示意图;
图2是本申请实施例提供的一种推理的方法的流程示意图;
图3是本申请实施例提供的另一种推理的方法的流程示意图;
图4是本申请实施例提供的另一种推理的方法的流程示意图;
图5是本申请实施例提供的另一种推理的方法的流程示意图;
图6是本申请实施例提供的另一种推理的方法的流程示意图;
图7是本申请实施例提供的另一种推理的方法的流程示意图;
图8是本申请实施例提供的另一种推理的方法的流程示意图;
图9是本申请实施例提供的另一种推理的方法的流程示意图;
图10是本申请实施例提供的另一种推理的方法的流程示意图;
图11是本申请实施例提供的另一种推理的方法的流程示意图;
图12是本申请实施例提供的另一种推理的方法的流程示意图;
图13是本申请实施例提供的另一种推理的方法的流程示意图;
图14是本申请实施例提供的另一种推理的方法的流程示意图;
图15是本申请实施例提供的另一种推理的方法的流程示意图;
图16是本申请实施例提供的另一种推理的方法的流程示意图;
图17是本申请实施例提供的另一种推理的方法的流程示意图;
图18是本申请实施例提供的另一种推理的方法的流程示意图;
图19是本申请实施例提供的另一种推理的方法的流程示意图;
图20是本申请实施例提供的另一种推理的方法的流程示意图;
图21是本申请实施例提供的另一种推理的方法的流程示意图;
图22是本申请实施例提供的另一种推理的方法的流程示意图;
图23是本申请实施例提供的另一种推理的方法的流程示意图;
图24是本申请实施例提供的一种推理的装置的结构示意图;
图25是本申请实施例提供的另一种推理的装置的结构示意图;
图26是本申请实施例提供的另一种推理的装置的结构示意图;
图27是本申请实施例提供的另一种推理的装置的结构示意图;
图28是本申请实施例提供的另一种推理的装置的结构示意图。
具体实施方式
请参见图1,图1为本申请实施例提供的一种推理的系统的架构示意图。该推理的系统可包括但不限于一个第一设备101、一个第二设备102和一个第三设备103,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第一设备101,两个或两个以上的第二设备102和两个或两个以上的第三设备103。图1所示的系统以包括一个第一设备101、一个第二设备102和一个第三设备103。
本申请实施例中的第一设备101,为第三方AI处理平台,为除无线蜂窝系统之外的服务器或者处理器。
本申请实施例中的第二设备102是用户侧的一种用于接收或发射信号的实体,如手机。第一设备也可以称为终端设备(terminal)、用户设备(user equipment,UE)、移动台(mobile station,MS)、移动终端设备(mobile terminal,MT)等。第二设备102的处理能力不足以独立完成AI模型推理任务,本申请的实施例对第二设备102所采用的具体技术和具体设备形态不做限定。
本申请实施例中的第三设备103是网络设备。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。
下面结合附图对本申请所提供的AI模型推理的方法及其装置进行详细地介绍。
请参见图2,图2是本申请实施例提供的一种推理的方法的流程示意图。该方法被第一设备执行,如图2所示,该方法可以包括但不限于如下步骤:
步骤S201:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,第二设备向第一设备发送AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图3为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S301:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,所述辅助第二设备执行AI模型推理任务包括以下任一种:所述第一设备单独完成所述AI模型推理任务;所述第一设备与所述第二设备共同完成所述AI模型推理任务;所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
当第一设备作为AI模型的提供方时,可以单独完成模型推理任务,或者,第一设备与所述第二设备共同完成完成模型推理任务。
当第一设备作为AI模型的使用方,需在接收第三设备传输的AI模型后,第一设备、第二设备及第三设备共同完成模型推理任务。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图4为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S401:将所述第一设备对所述AI模型的推理能力信息发送至所述第二设备。
第一设备将AI模型的推理能力信息发送至第二设备,其目的在于,将第二设备作为中转,将获得的推理能力信息转发至第三设备,以实现AI模型在传输时的信息同步,以便第三设备根据推理能力信息确定是否让第二设备使用无线AI模型的功能或者使用哪些无线AI模型的用例。
第三设备响应于第二设备上报的具备AI模型推理能力信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,第二设备向第一设备发送AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务。
示例性的,所述AI模型的推理能力信息包括:AI模型信息、AI处理平台框架信息以及AI处理能力信息中的至少一种。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
步骤S402:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图5为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S501:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
辅助第二设备执行AI模型推理任务包括以下任一种:所述第一设备单独完成所述AI模型推理任务、所述第一设备与所述第二设备共同完成所述AI模型推理任务、所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
步骤S502:将处理AI模型推理任务的耗时信息上报给所述第三设备。
根据AI模型中处理AI任务的类别,确定处理每个AI任务的耗时信息/延时信息,将耗时信息/延时信息上报至第三设备。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图6为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S601:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
步骤S602:响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型;或者,响应于进行推理的AI模型由所述第三设备提供,接收所述第二设备转发的所述AI模型。
示例性的,当第一设备作为AI模型的使用方,第三设备作为AI模型的提供方,第一设备接收第三设备传输的AI模型。
除了第一设备和第三设备之间能够直接传输AI模型外,本公开实施例中,还支持由第二设备的中转,即第三设备作为AI模型的提供方,向第二设备传输AI模型,由第二设备将AI模型转达AI模型至第一设备。即在第一设备、第二设备及第三设备之间执行AI模型的传输。
以上执行AI模型传输的过程仅为示例性说明,并非意在限定AI模型的传输顺序仅包含上述实例的实现。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,在所述第一设备、所述第二设备和第三设备至少两个设备中执行AI模型传输,以完成第二设备的模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图7为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S701:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务, AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
步骤S702:响应于进行推理的AI模型由所述第一设备提供,向所述第二设备发送的所述AI模型,所述AI模型通过所述第二设备转发至所述第三设备;或者响应于进行推理的AI模型由所述第一设备提供,直接向所述第三设备发送的所述AI模型。
本申请实施例的场景为第一设备作为AI模型的提供方,需要将其传输至AI模型的使用方(第三设备),第三设备根据接收到的AI模型辅助第一设备执行AI模型推理任务。在第一设备向第三设备传输AI模型时,与第三设备向第一设备传输AI模型的过程类似,采用第一设备直接将第一设备向第三设备传输AI模型传输至第三设备,也可采用第一设备向第二设备传输AI模型后,由第二设备向第三设备传输AI模型,本公开实施例对传输AI模型的方式不进行具体限定。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,在所述第一设备、所述第二设备和第三设备至少两个设备中执行AI模型传输,以完成第二设备的模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图8为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S801:响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
步骤S802:将所述推理结果发送至所述第二设备,所述推理结果通过所述第二设备转发至所述第三设备;或者将所述推理结果直接上报至所述第三设备。
作为一种本申请实施例的一种实现方式,在第一设备辅助第二设备执行完AI模型推理任务后,将推理结果返回至第二设备,由第二设备上传至第三设备。
作为本申请实施例的另一种实现方式,在第一设备辅助第二设备执行完AI模型推理任务后,将推理结果直接返回至第三设备。
本申请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,网络设备可以为演 进型基站(evolved NodeB,eNB)、传输点(transmission reception 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。本申请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。本申请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。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,并将推理结果辅助返回至第二设备或第三设备,辅助第二设备完成AI模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图9为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第一设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S901:响应于第二设备提供或使用基于AI模型的推理结果,第一设备辅助所述第二设备执行AI模型推理任务,所述AI模型推理任务由所述第一设备单独完成、由所述第一设备与所述第二设备共同完成或由所述第一设备、所述第二设备及其第三设备共同完成。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
步骤S902:将基于所述推理结果进一步得到的参数发送至所述第二设备,所述参数通过所述第二设备转发至所述第三设备;或者将所述基于所述推理结果进一步得到的参数直接上报至所述第三设备。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,并将推理结果进一步得到的参数返回至第二设备或第三设备,辅助第二设备完成AI模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,该方法被第一设备执行,该推理的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。
本申请实施例中,提供新的AI推理处理架构,包含第一设备、第二设备及第三设备,第一设备与所述第二设备进行交互的协议为由所述第一设备与所述第二设备自定义的交互协议,第一设备与所述第三设备之间的协议为通用的交互协议。
本公开实施例提供了另一种推理的方法,图10为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S1001:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。第二设备为不具有独立推理的条件的设备,例如硬件条件受限或者AI处理平台不兼容。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图11为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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11所示,该推理的方法可包括如下步骤:
步骤S1101:接收所述第一设备发送的辅助进行AI模型推理的推理能力信息。
第一设备将AI模型的推理能力信息发送至第二设备,其目的在于,由第二设备将获取的推理能力信息上报至第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务。
所述AI模型的推理能力信息包括AI模型信息、AI处理平台框架信息以及AI处理能力信息中的至少一种。示例性的,所支持的AI模型的类型是卷积神经网络(Deep Convolutional Neural Network,CNN)、循环神经网络(Recurrent Neural Network,RNN)还是transform等。AI处理框架是tenserflow还是Pytorch,计算速度等。
步骤S1102:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
在该方案中,将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图12为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S1201:将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备。
在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,可借助于第一设备协助第二设备进行AI推理。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
第一设备将AI模型的推理能力信息发送至第二设备,其目的在于,由第二设备将获取的推理能力信息上报至第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务。
第二设备作为中转,将获得的推理能力信息转发至第三设备,以实现AI模型在传输时的信息同步,以便第三设备根据推理能力信息确定是否让第二设备使用无线AI模型的功能或者使用哪些无线AI模型的用例。
所述AI模型的推理能力信息包括AI模型信息、AI处理平台框架信息以及AI处理能力信息中的至少一种。示例性的,所支持的AI模型的类型是卷积神经网络(Deep Convolutional Neural Network,CNN)、循环神经网络(Recurrent Neural Network,RNN)还是transform等。AI处理框架是tenserflow还是Pytorch,计算速度等。
步骤S1202:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
在该方案中,将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图13为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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,该推理的方法可包括如下步骤:
步骤S1301:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
步骤S1302:响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备。
示例性的,当第一设备作为AI模型的使用方,第三设备作为AI模型的提供方,第一设备接收第三设备传输的AI模型。
除了第一设备和第三设备之间能够直接传输AI模型外,本公开实施例中,还支持由第二设备的中转,即第三设备作为AI模型的提供方,向第二设备传输AI模型,由第二设备将AI模型转达AI模型至第一设备。即在第一设备、第二设备及第三设备之间执行AI模型的传输。
以上执行AI模型传输的过程仅为示例性说明,并非意在限定AI模型的传输顺序仅包含上述实例的实现。
在该方案中,响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图14为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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14所示,该推理的方法可包括如下步骤:
步骤S1401:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
步骤S1402:响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型,并将所述AI模型转发至所述第三设备。
本申请实施例的场景为第一设备作为AI模型的提供方,需要将其传输至AI模型的使用方(第三设备),第三设备根据接收到的AI模型辅助第一设备执行AI模型推理任务。在第一设备向第三设备传输AI模型时,与第三设备向第一设备传输AI模型的过程类似,采用第一设备直接将第一设备向第三设备传输AI模型传输至第三设备,也可采用第一设备向第二设备传输AI模型后,由第二设备向第三设备传输AI模型,本公开实施例对传输AI模型的方式不进行具体限定。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,在所述第一设备、所述第二设备和第三设备至少两个设备中执行AI模型传输,以完成第二设备的模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图15为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第二设备执行,该推理的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15所示,该推理的方法可包括如下步骤:
步骤S1501:响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
步骤S1502:接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
作为一种本申请实施例的一种实现方式,在第一设备辅助第二设备完成AI模型推理任务后,将推理结果返回至第二设备,由第二设备上传至第三设备。
作为本申请实施例的另一种实现方式,在第一设备辅助第二设备执行完AI模型推理任务后,将推理结果直接返回至第三设备。
本申请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,网络设备可以为演进型基站(evolved NodeB,eNB)、传输点(transmission reception 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。本申请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。本申请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。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,并将推理结果辅助返回至第二设备或第三设备,辅助第二设备完成AI模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,该方法应用于第三设备侧,图16为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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16所示,该推理的方法可包括如下步骤:
步骤S1601:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
第三设备向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,例如硬件条件受限或者AI处理平台不兼容时,第二设备向第一设备发送AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务。
作为本申请实施例的可行方式,所述第一设备为无线蜂窝系统之外的服务器或者处理器。具体的对第一设备的具体设备形式不进行限定。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图17为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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17所示,该推理的方法可包括如下步骤:
步骤S1701:接收所述第二设备发送的所述第一设备对所述AI模型的推理能力信息。
步骤S1702:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
在第二设备不具有独立推理的条件时,以便第一设备辅助所述第二设备完成所述推理任务,所述第 二设备响应于需要所述第二设备提供AI模型的推理结果或使用AI模型的推理结果上报具体AI模型推理能力的信息。
第一设备将AI模型的推理能力信息发送至第二设备,其目的在于,由第二设备将获取的推理能力信息上报至第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务。第二设备作为中转,将获得的推理能力信息转发至第三设备,以实现AI模型在传输时的信息同步,以便第三设备根据推理能力信息确定是否让第二设备使用无线AI模型的功能或者使用哪些无线AI模型的用例。
所述AI模型的推理能力信息包括AI模型信息、AI处理平台框架信息以及AI处理能力信息中的至少一种。示例性的,所支持的AI模型的类型是卷积神经网络(Deep Convolutional Neural Network,CNN)、循环神经网络(Recurrent Neural Network,RNN)还是transform等。AI处理框架是tenserflow还是Pytorch,计算速度等。
在该方案中,将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务,在第二设备不具有独立推理的条件时,第二设备向第一设备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图18为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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18所示,该推理的方法可包括如下步骤:
步骤S1801:接收所述第二设备发送的所述第二设备对所述AI模型的推理能力信息。
第二设备将推理能力信息上报至第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务。第二设备发送的所述第二设备对所述AI模型的推理能力信息可以包含但不限于第一设备提供的AI模型推理的推理能力信息,还可以包括第二设备自身的AI模型推理的推理能力信息,第三设备并不关注AI模型推理的推理能力信息的来源,而是以第二设备所提供的的AI模型推理的推理能力信息能够执行AI推理任务为准。
所述AI模型的推理能力信息包括AI模型信息、AI处理平台框架信息以及AI处理能力信息中的至少一种。示例性的,所支持的AI模型的类型是卷积神经网络(Deep Convolutional Neural Network,CNN)、循环神经网络(Recurrent Neural Network,RNN)还是transform等。AI处理框架是tenserflow还是Pytorch,计算速度等。
步骤S1802:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
以便第一设备辅助所述第二设备完成所述推理任务,所述第二设备响应于需要所述第二设备提供AI模型的推理结果或使用AI模型的推理结果上报具体AI模型推理能力的信息。
在该方案中,将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备,第三设备根据接收到的AI模型推理的推理能力信息配置第二设备进行执行AI推理任务,第二设备向第一设 备发送需要辅助第二设备完成AI模型推理任务的AI模型推理请求,由第一设备辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图19为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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19所示,该推理的方法可包括如下步骤:
步骤S1901:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
步骤S1902:接收所述第一设备上报的处理AI模型推理任务的耗时信息。
第一设备根据AI模型中处理AI任务的类别,确定处理每个AI任务的耗时信息/延时信息,将耗时信息/延时信息上报至第三设备。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图20为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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20所示,该推理的方法可包括如下步骤:
步骤S2001:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
步骤S2002:响应于进行推理的AI模型由所述第三设备提供,直接将所述AI模型发送至所述第一设备;或者响应于进行推理的AI模型由所述第三设备提供,将所述所述AI模型发送至所述第二设备,所述AI模型通过第二设备转发至所述第一设备。
本申请实施例的场景为第三设备作为AI模型的提供方,需要将其传输至AI模型的使用方(第一设备),第一设备根据接收到的AI模型辅助第一设备执行AI模型推理任务。
本申请实施例的场景为第三设备作为AI模型的提供方,需要将其传输至第二设备后,由第二设备转发至AI模型的使用方(第一设备),以由实现第一设备、第二设备及第三设备共同执行AI模型推理任务。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
本公开实施例提供了另一种推理的方法,图21为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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21所示,该推理的方法可包括如下步骤:
步骤S2101:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
步骤S2102:响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型;或者响应于进行推理的AI模型由所述第一设备提供,接收所述第二设备转发的所述AI模型。
有关AI模型在第一设备、第二设备及第三设备之间的传输过程,可参阅任意实施例的详细说明,本公开实施例在此不再进行赘述。
本公开实施例提供了另一种推理的方法,图22为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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22所示,该推理的方法可包括如下步骤:
步骤S2001:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
步骤S2202:响应于接收到第一设备提供的AI模型,辅助所述第一设备、所述第二设备完成所述AI模型推理任务。
本公开实施例提供了另一种推理的方法,图23为本公开实施例提供的另一种推理的方法的流程示意图,该方法被第三设备执行,该推理的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23所示,该推理的方法可包括如下步骤:
步骤S2302:响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
步骤S2302:接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
所述推理结果为:由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
在该方案中,第三设备响应于接收第二设备上报的具备AI模型推理能力的信息,向第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,并将推理结果返回至第三设备,辅助第二设备完成AI模型推理任务,使第二设备间接性具备推理能力,受益于无线AI。
与上述图2至图23实施例提供的推理的方法相对应,本公开还提供一种推理的装置,由于本公开实施例提供推理的装置与上述图2至图23实施例提供的推理的方法相对应,因此在推理的方法的实施方式也适用于本公开实施例提供的推理的装置,在本公开实施例中不再详细描述。
图24为本公开实施例所提供的一种推理的装置的结构示意图。所述装置被设置在第一设备,所述装置包括:
处理单元2401,用于响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
作为本公开实施例的的一种可能实现方式,所述辅助第二设备执行AI模型推理任务包括以下任一种:
所述第一设备单独完成所述AI模型推理任务;
所述第一设备与所述第二设备共同完成所述AI模型推理任务;
所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
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:
发送单元2402,用于将所述第一设备对所述AI模型的推理能力信息发送至所述第二设备。
作为本公开实施例的的一种可能实现方式,所述AI模型的推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
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:
上报单元2403,用于将处理AI模型推理任务的耗时信息上报给所述第三设备。
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:
接收单元2404,用于响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型;或者。
所述接收单元2404,还用于响应于进行推理的AI模型由所述第三设备提供,接收所述第二设备转发的所述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:
发送单元2402,用于响应于进行推理的AI模型由所述第一设备提供,向所述第二设备发送的所述AI模型,所述AI模型通过所述第二设备转发至所述第三设备;或者
发送单元2402,还用于响应于进行推理的AI模型由所述第一设备提供,直接向所述第三设备发送的所述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:
发送单元2402,用于将所述推理结果发送至所述第二设备,所述推理结果通过所述第二设备转发至所述第三设备;或者
将所述推理结果直接上报至所述第三设备。
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:
发送单元2402,用于将基于所述推理结果进一步得到的参数发送至所述第二设备,所述参数通过所述第二设备转发至所述第三设备;或者
上报单元2403,用于将所述基于所述推理结果进一步得到的参数直接上报至所述第三设备。
作为本公开实施例的的一种可能实现方式,所述第一设备与所述第二设备进行交互的协议为自定义的交互协议。
作为本公开实施例的的一种可能实现方式,本申请实施例提供一种人工智能AI模型推理的装置,所述装置被设置在第二设备,如图25所示,包括:
发送单元2501,用于响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
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:
接收单元2502,用于接收所述第一设备发送的辅助进行AI模型推理的推理能力信息。
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:
上报单元2503,用于将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备。
作为本公开实施例的的一种可能实现方式,所述推理能力信息包括:
AI模型信息、AI处理平台框架信息以及AI处理能力信息。
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:
接收单元2502,用于响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备。
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:
接收单元2502,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型,并将所述AI模型转发至所述第三设备。
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:
接收单元2502,用于接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
作为本公开实施例的的一种可能实现方式,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
作为本公开实施例的的一种可能实现方式,所述第二设备与所述第一设备进行交互的协议为自定义的交互协议。
作为本公开实施例的的一种可能实现方式,本申请实施例提供一种人工智能AI模型推理的装置,所述装置被设置在第三设备,如图26所示,所述装置包括:
发送单元2601,用于响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务,以便第一设备辅助所述第二设备完成所述推理任务,所述第二设备响应于需要所述第二设备提供AI模型的推理结果或使用AI模型的推理结果上报具体AI模型推理能力的信息。
在该方案中,第三设备向所述第二设备发送AI模型推理任务,在第二设备不具有独立推理的条件时,第一设备响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,使得第二设备能够响应于需要提供AI模型的推理结果或使用AI模型的推理结果,使第二设备间接性具备推理能力,受益于无线AI。
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:
接收单元2602,用于接收所述第二设备发送的所述第一设备对所述AI模型的推理能力信息。
为本公开实施例的的一种可能实现方式,所述装置还包括:
接收单元2602,用于接收所述第二设备发送的所述第二设备对所述AI模型的推理能力信息。
作为本公开实施例的的一种可能实现方式,所述AI模型的推理能力信息包括;AI模型信息、AI处理平台框架信息以及AI处理能力信息。
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:
接收单元2602,用于接收所述第一设备上报的处理AI模型推理任务的耗时信息。
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:
发送单元2603,用于响应于进行推理的AI模型由所述第三设备提供,直接将所述AI模型发送至所述第一设备;或者
发送单元2603,用于响应于进行推理的AI模型由所述第三设备提供,将所述所述AI模型发送至所述第二设备,由第二设备将所述AI模型转发至所述第一设备。
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:
接收单元2602,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型;或者
接收单元2602,用于响应于进行推理的AI模型由所述第一设备提供,接收所述第二设备转发的所述AI模型。
作为本公开实施例的的一种可能实现方式,处理单元2601,用于响应于接收到第一设备提供的AI模型,辅助所述第一设备、所述第二设备完成所述AI模型推理任务。
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:
接收单元2602,用于接收所述第二设备发送的AI模型的推理结果。
作为本公开实施例的的一种可能实现方式,所述推理结果为:
由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
为了实现上述实施例,本公开还提供另一种推理的装置,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9所示的方法,或执行图10至图15所示的方法,或执行图16至图23所示的方法。
为了实现上述本公开实施例提供的方法中的各功能,第一设备、第二设备及第三设备可以包括硬件结构、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27,图27为本公开实施例所提供的一种推理的装置的结构示意图。参照图27,网络设备2700包括处理组件2722,其进一步包括至少一个处理器,以及由存储器2732所代表的存储器资源,用于存储可由处理组件2722的执行的指令,例如应用程序。存储器2732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2722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21实施例所述的方法。
网络设备2700还可以包括一个电源组件2706被配置为执行网络设备2700的电源管理,一个有线或无线网络接口2750被配置为将网络设备2700连接到网络,和一个输入输出(I/O)接口2758。网络设备2700可以操作基于存储在存储器2732的操作系统,例如Windows Server TM,Mac OS XTM,UnixTM,LinuxTM,FreeBSDTM或类似。
为了实现上述实施例,本申请实施例提供一种推理的系统,包括:如图24所示的推理的装置、如图25所示的推理的装置以及如图26所示的推理的装置。
图28为本公开实施例所提供的一种推理的装置的框图。例如,用户设备2800可以是移动电话,计算机,数字广播用户设备,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28,用户设备2800可以包括以下至少一个组件:处理组件2802,存储器2804,电源组件2806,多媒体组件2808,音频组件2810,输入/输出(I/O)的接口2812,传感器组件2814,以及通信组件2816。
处理组件2802通常控制用户设备2800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2802可以包括至少一个处理器2820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2802可以包括至少一个模块,便于处理组件2802和其他组件之间的交互。例如,处理组件2802可以包括多媒体模块,以方便多媒体组件2808和处理组件2802之间的交互。
存储器2804被配置为存储各种类型的数据以支持在用户设备2800的操作。这些数据的示例包括用于在用户设备2800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2804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2806为用户设备2800的各种组件提供电力。电源组件2806可以包括电源管理系统,至少一个电源,及其他与为用户设备2800生成、管理和分配电力相关联的组件。
多媒体组件2808包括在所述用户设备2800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2808包括一个前置摄像头和/或后置摄像头。当用户设备2800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2810被配置为输出和/或输入音频信号。例如,音频组件2810包括一个麦克风(MIC),当用户设备2800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2804或经由通信组件2816发送。在一些实施例中,音频组件2810还包括一个扬声器,用于输出音频信号。
I/O接口2812为处理组件2802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2814包括至少一个传感器,用于为用户设备2800提供各个方面的状态评估。例如,传感器组件2814可以检测到用户设备2800的打开/关闭状态,组件的相对定位,例如所述组件为用户设备2800的显示器和小键盘,传感器组件2814还可以检测用户设备2800或用户设备2800一个组件的位置改变,用户与用户设备2800接触的存在或不存在,用户设备2800方位或加速/减速和用户设备2800的温度变化。传感器组件2814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2814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2814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2815被配置为便于用户设备2800和其他设备之间有线或无线方式的通信。用户设备2800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2815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2815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,用户设备2800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2804,上述指令可由用户设备2800的处理器2820执行以完成上述图2至图21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推理的装置可理解的其他名称,其参数的取值或表示方式也可以推理的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 (42)

  1. 一种AI模型推理的方法,该方法被第一设备执行,其特征在于,所述方法包括:
    响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
  2. 根据权利要求1所述的方法,其特征在于,所述辅助第二设备执行AI模型推理任务包括以下任一种:
    所述第一设备单独完成所述AI模型推理任务;
    所述第一设备与所述第二设备共同完成所述AI模型推理任务;
    所述第一设备与所述第二设备及第三设备共同完成所述AI模型推理任务。
  3. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    将所述第一设备对所述AI模型的推理能力信息发送至所述第二设备。
  4. 根据权利要求3所述的方法,其特征在于,所述AI模型的推理能力信息包括:
    AI模型信息、AI处理平台框架信息以及AI处理能力信息。
  5. 根据权利要求2所述的方法,其特征在于,所述方法还包括:
    将处理AI模型推理任务的耗时信息上报给所述第三设备。
  6. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型;或者,
    响应于进行推理的AI模型由所述第三设备提供,接收所述第二设备转发的所述AI模型。
  7. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述第一设备提供,向所述第二设备发送的所述AI模型,所述AI模型通过所述第二设备转发至所述第三设备;或者
    响应于进行推理的AI模型由所述第一设备提供,直接向所述第三设备发送的所述AI模型。
  8. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    将所述推理结果发送至所述第二设备,所述推理结果通过所述第二设备转发至所述第三设备;或者
    将所述推理结果直接上报至所述第三设备。
  9. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    将基于所述推理结果进一步得到的参数发送至所述第二设备,所述参数通过所述第二设备转发至所述第三设备;或者
    将所述基于所述推理结果进一步得到的参数直接上报至所述第三设备。
  10. 根据权利要求1-9中任一项所述的方法,其特征在于,
    所述第一设备与所述第二设备进行交互的协议为自定义的交互协议。
  11. 一种人工智能AI模型推理的方法,该方法被第二设备执行,其特征在于,包括:
    响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
  12. 根据权利要求11所述的方法,其特征在于,所述方法还包括:
    接收所述第一设备发送的辅助进行AI模型推理的推理能力信息。
  13. 根据权利要求12所述的方法,其特征在于,所述方法还包括:
    将所述第一设备辅助进行AI模型推理的推理能力信息上报至所述第三设备。
  14. 根据权利要求12所述的方法,其特征在于,所述推理能力信息包括:
    AI模型信息、AI处理平台框架信息以及AI处理能力信息。
  15. 根据权利要求11所述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述第三设备提供,接收所述第三设备发送的所述AI模型,并将所述AI模型转发至所述第一设备。
  16. 根据权利要求11述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述第一设备提供,接收所述第一设备发送的所述AI模型,并将所述AI模型转发至所述第三设备。
  17. 根据权利要求11-16中任一项所述的方法,其特征在于,所述方法还包括:
    接收所述第一设备返回的AI模型推理的推理结果,并将所述推理结果转发至所述第三设备。
  18. 根据权利要求17所述的方法,其特征在于,所述推理结果为:
    由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
    由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
    由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
  19. 根据权利要求18所述的方法,其特征在于,
    所述第二设备与所述第一设备进行交互的协议为自定义的交互协议。
  20. 一种人工智能AI模型推理的方法,该方法被第三设备执行,其特征在于,所述方法包括:
    响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
  21. 根据权利要求20所述的方法,其特征在于,所述方法还包括:
    接收所述第二设备发送的所述第一设备对所述AI模型的推理能力信息。
  22. 根据权利要求20所述的方法,其特征在于,所述方法还包括:
    接收所述第二设备发送的所述第二设备对所述AI模型的推理能力信息。
  23. 根据权利要求21所述的方法,其特征在于,所述AI模型的推理能力信息包括;AI模型信息、AI处理平台框架信息以及AI处理能力信息。
  24. 根据权利要求20所述的方法,其特征在于,所述方法还包括:
    接收所述第一设备上报的处理AI模型推理任务的耗时信息。
  25. 根据权利要求20所述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述第三设备提供,将所述AI模型发送至所述第一设备;或者
    响应于进行推理的AI模型由所述第三设备提供,将所述所述AI模型发送至所述第二设备,所述AI模型通过所述第二设备转发至所述第一设备。
  26. 根据权利要求20所述的方法,其特征在于,所述方法还包括:
    响应于进行推理的AI模型由所述一设备提供,接收所述第一设备发送的所述AI模型;或者
    响应于进行推理的AI模型由所述一设备提供,接收所述第二设备转发的所述AI模型。
  27. 根据权利要求20所述的方法,其特征在于,
    响应于接收到第一设备提供的AI模型,辅助所述第一设备、所述第二设备完成所述AI模型推理任务。
  28. 根据权利要求20-27中任一项所述的方法,其特征在于,所述方法还包括:
    接收所述第二设备发送的AI模型的推理结果。
  29. 根据权利要求28所述的方法,其特征在于,所述推理结果为:
    由所述第一设备单独完成所述AI模型推理任务得到的推理结果;或者
    由所述第一设备与所述第二设备共同完成所述AI模型推理任务得到的推理结果;或者
    由所述第一设备与所述第二设备及其第三设备共同完成所述AI模型推理任务得到的推理结果。
  30. 一种AI模型推理的装置,所述装置被设置在第一设备,其特征在于,所述装置包括:
    处理单元,用于响应于接收第二设备发送的AI模型推理请求,辅助第二设备完成AI模型推理任务,AI模型推理请求为所述第二设备响应于需要提供AI模型的推理结果或使用AI模型的推理结果时向所述第一设备发送的。
  31. 一种人工智能AI模型推理的装置,所述装置被设置在第二设备,其特征在于,包括:
    发送单元,用于响应于第二设备提供AI模型的推理结果或使用AI模型的推理结果,向第一设备发送需要辅助所述第二设备完成AI模型推理任务的AI模型推理请求。
  32. 一种人工智能AI模型推理的装置,所述装置被设置在第三设备,其特征在于,所述装置包括:
    发送单元,用于响应于接收第二设备上报的具备AI模型推理能力的信息,向所述第二设备发送AI模型推理任务。
  33. 一种推理的装置,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至10中任一项所述的方法。
  34. 一种推理的装置,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11至19中任一项所述的方法。
  35. 一种推理的装置,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20至29中任一项所述的方法。
  36. 一种推理的装置,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求1至10中任一项所述的方法。
  37. 一种推理的装置,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求11至19中任一项所述的方法。
  38. 一种推理的装置,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求20至29中任一项所述的方法。
  39. 一种推理的系统,其特征在于,包括:如权利要求30所述的推理的装置、如权利要求31所述的推理的装置以及如权利要求32所述的推理的装置;
    或者,所述系统包括如权利要求33所述的推理的装置、如权利要求34所述的推理的装置以及如权利要求35所述的推理的装置;
    或者,所述系统包括如权利要求36所述的推理的装置、如权利要求37所述的推理的装置以及如权利要求38所述的推理的装置。
  40. 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至10中任一项所述的方法被实现。
  41. 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11至19中任一项所述的方法被实现。
  42. 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20至29中任一项所述的方法被实现。
PCT/CN2022/103485 2022-07-01 2022-07-01 一种ai模型推理的方法及其装置 WO2024000605A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2022/103485 WO2024000605A1 (zh) 2022-07-01 2022-07-01 一种ai模型推理的方法及其装置
CN202280002424.4A CN117651954A (zh) 2022-07-01 2022-07-01 一种ai模型推理的方法及其装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/103485 WO2024000605A1 (zh) 2022-07-01 2022-07-01 一种ai模型推理的方法及其装置

Publications (1)

Publication Number Publication Date
WO2024000605A1 true WO2024000605A1 (zh) 2024-01-04

Family

ID=89383918

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/103485 WO2024000605A1 (zh) 2022-07-01 2022-07-01 一种ai模型推理的方法及其装置

Country Status (2)

Country Link
CN (1) CN117651954A (zh)
WO (1) WO2024000605A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130179391A1 (en) * 2010-09-13 2013-07-11 Siemens Aktiengesellschaft Apparatus for processing data in a computer-aided logic system, and appropriate method
CN112686374A (zh) * 2020-12-31 2021-04-20 中山大学 基于自适应负载分配的深度神经网络模型协同推理方法
CN112784989A (zh) * 2019-11-08 2021-05-11 阿里巴巴集团控股有限公司 推理系统、推理方法、电子设备及计算机存储介质
CN114254751A (zh) * 2020-09-21 2022-03-29 华为技术有限公司 协同推理方法及通信装置
CN114416863A (zh) * 2020-10-28 2022-04-29 中国电信股份有限公司 用于执行基于模型并行的分布式推理的方法、设备和介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130179391A1 (en) * 2010-09-13 2013-07-11 Siemens Aktiengesellschaft Apparatus for processing data in a computer-aided logic system, and appropriate method
CN112784989A (zh) * 2019-11-08 2021-05-11 阿里巴巴集团控股有限公司 推理系统、推理方法、电子设备及计算机存储介质
CN114254751A (zh) * 2020-09-21 2022-03-29 华为技术有限公司 协同推理方法及通信装置
CN114416863A (zh) * 2020-10-28 2022-04-29 中国电信股份有限公司 用于执行基于模型并行的分布式推理的方法、设备和介质
CN112686374A (zh) * 2020-12-31 2021-04-20 中山大学 基于自适应负载分配的深度神经网络模型协同推理方法

Also Published As

Publication number Publication date
CN117651954A (zh) 2024-03-05

Similar Documents

Publication Publication Date Title
EP3833081B1 (en) Configuration method and apparatus for transmission configuration indication
US20230217366A1 (en) Access method, access apparatus, and storage medium
WO2019183857A1 (zh) 信息传输方法和信息传输装置
US20160164808A1 (en) Method and device for instant messaging
US11792659B2 (en) Method and device for using network slice
US11985620B2 (en) Network registration method and apparatus
US20230276340A1 (en) Capability reporting method and apparatus, and capability determination method and apparatus
US20230403589A1 (en) Method for wlan sensing, apparatus for wlan sensing and non-transitory computer-readable storage medium
US11805562B2 (en) User device pairing method and apparatus
US11864190B2 (en) Method and apparatus for configuring and determining transmission block scheduling interval, and base station
US20240063980A1 (en) System information reception method and apparatus, and system information transmission method and apparatus
US10123208B2 (en) Method and device for controlling access
EP4336927A1 (en) Method and apparatus for monitoring physical downlink control channel, and storage medium
US20230388981A1 (en) Wireless sensing resource coordination method and apparatus, and storage medium
US11950162B2 (en) Unmanned aerial vehicle control method and apparatus
WO2024000605A1 (zh) 一种ai模型推理的方法及其装置
JP2018527765A (ja) サービス接続を確立する方法、装置、プログラム、及び記録媒体
WO2022036610A1 (zh) 一种通信方法、通信装置及存储介质
US20230262746A1 (en) Random access method and device
WO2020164515A1 (zh) 信号传输方法、设备及系统
CN108924085B (zh) 网络调度方法、装置及存储介质
CN114430370A (zh) 网络切片的确定方法、装置、设备及存储介质
WO2024000532A1 (zh) 一种ai模型的传输方法及其装置
WO2019183763A1 (zh) 信息记录方法和信息记录装置
WO2023206298A1 (zh) 一种协商方法/装置/设备及存储介质

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 202280002424.4

Country of ref document: CN

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

Ref document number: 22948704

Country of ref document: EP

Kind code of ref document: A1