Method for large language model application in wireless cellular system

Abstract

The application discloses a method for applying a large language model in a wireless cellular system, the method comprising: for an application based on a large language model in a terminal, the terminal dynamically switches the application mode of the large language model thereof by using a hierarchical triggering strategy, wherein the hierarchical triggering strategy first judges whether parameters of the terminal meet the conditions for applying the large language model, wherein the parameters at least include the remaining power of the terminal, whether the large language model can provide an answer, and the response delay of the large language model. If the above parameters meet the conditions for applying the large language model in the terminal, the application of the large language model is performed in the terminal. If the above parameters do not meet the conditions for applying the large language model in the terminal, it is judged based on network side parameters whether the application of the large language model can be performed on the network side, wherein the network side parameters at least include the data transmission cost and the data transmission delay of the terminal. If the network side parameters meet the conditions for performing the application of the large language model on the network side, the terminal transmits data to the network side for the application of the large language model. In the process of performing the application of the large language model on the network side, the terminal monitors the parameters thereof at any time, judges whether the parameters of the terminal meet the conditions for applying the large language model, and if the conditions for applying the large language model in the terminal are met, the application of the large language model in the terminal is dynamically switched back.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and more particularly to a method and device for applying a large language model in a wireless cellular system. BACKGROUND

[0002] With the progress of chip computing power and artificial intelligence algorithms, large language models (LLM) are also used in terminal devices such as mobile phones for text, speech, image processing, and wireless signal processing. The large language model generally refers to a deep learning model trained using a large amount of text data, which can generate natural language text or understand the meaning of language text. The large language model can also be referred to as a large model, which generally refers to an artificial intelligence model with a large number of parameters. Based on a mobile phone, the parameter set of the large language model can reach hundreds of millions or even tens of billions, and it can handle some simple artificial intelligence applications. However, for some more complex tasks, such as image processing or audio / video processing, due to limitations of the model, mobile phone hardware, and other conditions, the application of the large language model (such as inference, training, etc.) needs to be performed on the network side. For a use scenario based on a large language model, how to balance user experience, complexity, power consumption, and other factors, and dynamically switch the application of the large language model on the terminal side and the network side is a problem that needs to be considered.

[0003] At present, the introduction of large language model technology into terminal data processing is just beginning, the signaling mechanism is not determined, and there are many optimization problems to be solved. SUMMARY

[0004] Therefore, the present application provides the following technical solutions:

[0005] A method for applying a large language model in a wireless cellular system, applied to data processing of a terminal, characterized in that it comprises:

[0006] For an application based on a large language model in a terminal, the terminal dynamically switches the application mode of the large language model thereof by using a hierarchical triggering strategy, wherein the hierarchical triggering strategy first judges whether the parameters of the terminal satisfy the conditions for the application of the large language model, wherein the parameters at least include the remaining power of the terminal, whether the large language model can provide an answer, and the response delay of the large language model. If the above parameters satisfy the conditions for the application of the large language model in the terminal, the application of the large language model in the terminal is performed. If the above parameters do not satisfy the conditions for the application of the large language model in the terminal, it is judged based on the network side parameters whether the application of the large language model in the network side can be performed, wherein the network side parameters at least include the data transmission cost and the data transmission delay of the terminal. If the network side parameters satisfy the conditions for the application of the large language model in the network side, the terminal transmits data to the network side for the application of the large language model. In the process of the application of the large language model in the network side, the terminal monitors the parameters thereof at any time, judges whether the parameters of the terminal satisfy the conditions for the application of the large language model, and if the conditions for the application of the large language model in the terminal are satisfied, the application of the large language model in the terminal is dynamically switched back.

[0007] The method comprises, based on the application of the large language model, at least including the training, learning and reasoning of the large language model for text summarization, language translation, audio optimization, video optimization or wireless signal processing.

[0008] The method further comprises judging whether the parameters of the terminal satisfy the conditions for the application of the large language model, at least including: first, the remaining power of the terminal is greater than a first specific threshold; and second, the large language model can provide an answer within a predicted time.

[0009] The method further comprises judging based on the network side parameters whether the application of the large language model in the network side can be performed, at least including: first, the data transmission cost of the terminal is less than a second specific threshold; and second, the data transmission delay of the terminal is less than a third specific threshold.

[0010] The method further comprises that the terminal monitors the parameters thereof at any time, judges whether the parameters of the terminal satisfy the conditions for the application of the large language model, and the judgment conditions at least include: first, the remaining power of the terminal is greater than a first specific threshold; and second, the large language model can provide an answer within a predicted time.

[0011] A processing device for large language model application in a wireless cellular system, applied to terminal data processing, is characterized by comprising: for a large language model-based application in a terminal, the terminal dynamically switches its large language model application mode using a hierarchical triggering strategy. The hierarchical triggering strategy first determines whether terminal parameters meet the conditions for large language model application, where the parameters include at least the terminal's remaining battery life, whether the large language model can provide answers, and the large language model's response delay. If these parameters meet the conditions for large language model application at the terminal, the large language model is applied at the terminal. If these parameters do not meet the conditions for large language model application at the terminal, whether the large language model can be applied on the network side is determined based on network parameters, where the network parameters include at least the terminal's data transmission cost and data transmission delay. If the network parameters meet the conditions for large language model application on the network side, the terminal transmits data to the network side for large language model application. During the large language model application process on the network side, the terminal constantly monitors its own parameters to determine whether the terminal parameters meet the conditions for large language model application. If the conditions for large language model application are met, the terminal dynamically switches back to large language model application at the terminal.

[0012] As can be seen from the above technical solutions, compared to the prior art, embodiments of the present invention disclose a method for applying a large language model in a wireless cellular system. The method includes: for a large language model-based application in a terminal, the terminal dynamically switches its large language model application mode using a hierarchical triggering strategy. The hierarchical triggering strategy first determines whether the terminal's parameters meet the conditions for applying the large language model, where the parameters include at least the terminal's remaining battery life, whether the large language model can provide answers, and the large language model's response delay. If these parameters meet the conditions for applying the large language model at the terminal, the large language model is applied at the terminal. If these parameters do not meet the conditions for applying the large language model at the terminal, whether the large language model can be applied on the network side is determined based on network-side parameters, where the network-side parameters include at least the terminal's data transmission cost and data transmission delay. If the network-side parameters meet the conditions for applying the large language model on the network side, the terminal transmits data to the network side for applying the large language model. During the large language model application process on the network side, the terminal monitors its own parameters at all times to determine whether the terminal parameters meet the conditions for applying the large language model. If the conditions for applying the large language model at the terminal are met, the terminal dynamically switches back to applying the large language model at the terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description only constitute a part of the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained based on the provided drawings without creative labor.

[0014] Figure 1 A flow chart of a large language model application method in a wireless cellular system according to an embodiment of the present application;

[0015] Figure 2 A schematic diagram of a large language model application method in a wireless cellular system according to an embodiment of the present application;

[0016] Figure 3 A functional schematic diagram of a large language model application in a wireless terminal according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0018] Figure 1 A flow chart of a large language model application method in a wireless cellular system according to an embodiment of the present application. Referring to FIG. 1, Figure 1 The flow chart includes the following steps:

[0019] Step 101: The terminal receives a request of a user about a large language model application, such as image processing, language translation or optimization of a wireless received signal.

[0020] Step 102: The terminal determines whether the remaining power is greater than a first specific threshold (for example, 20%). Considering that the large language model consumes more power when applied, the terminal needs to ensure that the application of the large language model will not affect the normal use of the terminal. If the remaining power is greater than the first specific threshold in actual application, the terminal will execute step 103: determining whether the large language model can provide an answer within a predicted time. If the terminal predicts that the large language model can provide an answer within a predicted time (for example, 3 seconds), the terminal will perform data processing (step 104) based on its own large language model.

[0021] If the conditions in step 102 and step 103 are not met, the terminal will consider applying the large language model on the network side. The network side here can be a base station, a multi-access edge server (abbreviated as MEC), a core network or an artificial intelligence server at the application layer. The terminal needs to determine whether the conditions for applying the large language model on the network side are met based on some network parameters. For example, in step 105, the terminal determines whether the data transmission fee is less than the second specific threshold (for example, RMB 1 per megabyte). Applying the large language model on the network side may require the transmission of a large amount of data. The terminal needs to determine the cost of transmitting this data to avoid incurring too high a cost for the user. If step 105 can meet the conditions, the terminal further determines whether the data transmission delay is less than the third specific threshold (for example, 10 seconds) (step 107). If step 107 can also meet the conditions, the terminal transmits data to the network side and starts to apply the large language model on the network side (step 108);

[0022] If the conditions in steps 105 and 107 are not met, the terminal will reject the user's request for applying the large language model (step 106) and process the data based on the normal process.

[0023] Figure 2 This is a schematic diagram of a method for applying a large language model in a wireless cellular system disclosed in an embodiment of the present invention, including a terminal and a network. This embodiment includes the following steps:

[0024] Step 0: The terminal receives a request to apply a large language model;

[0025] Step 1: The terminal applies the large language model to determine whether the remaining battery level is greater than a specific threshold (e.g., 20%). Secondly, the terminal determines whether the large language model can provide an answer within the expected time (e.g., 3 seconds). If neither of these conditions is met, the terminal proceeds to step 2.

[0026] Step 2: The terminal determines whether its data transmission fee is less than the second specific threshold (e.g., RMB 1 / Mbyte), and then determines whether the terminal's data transmission delay is less than the third specific threshold (e.g., 10 seconds). If the above conditions are met, proceed to step 3;

[0027] Step 3: The terminal transmits relevant data to the network side;

[0028] Step 4: Apply the large language model on the network side;

[0029] Step 5: The network side returns the application result.

[0030] The embodiment balances terminal power consumption, large language model application delay, processing cost and other indicators, and can play a good optimization role for subsequent various applications of the terminal (such as text, image, audio / video, and wireless signal processing), and improve user experience.

[0031] For the foregoing method embodiments, for the sake of simple description, they are all expressed as a combination of a series of actions, but those skilled in the art should know that the present application is not limited by the action sequence described, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily necessary for the present application.

[0032] The method is described in detail in the above disclosed embodiments of the present application, and the method of the present application can be implemented in various forms of devices, so the present application also discloses a device, and the following specific embodiments are given to explain in detail.

[0033] Figure 3 A function diagram of the large language model application in the wireless terminal disclosed by the embodiment of the present application, and a large language model application control decision unit disclosed by the embodiment of the present application. The data receiving module 301 is used to receive various data, such as original data that needs to be processed by the large language model, network signaling, internal instructions of the terminal, etc.; the data processing module 302 is mainly used to process various data, such as inference based on the large language model, switching decision of the terminal side and network side large language model, and is the core module of the large language model decision; the module 303 is a data sending module, which sends internal instructions to the terminal, interaction signaling with the network layer, original data, etc.; the signaling processing module 304 is a network signaling analysis and encapsulation module.

[0034] The method and device for applying a large language model in a wireless cellular system are described in the embodiments. The method comprises: for a terminal-based application of a large language model, the terminal dynamically switches the application mode of the large language model thereof by using a hierarchical triggering strategy, wherein the hierarchical triggering strategy first determines whether the parameters of the terminal meet the conditions for applying the large language model, wherein the parameters at least include the remaining power of the terminal, whether the large language model can provide an answer, and the response delay of the large language model. If the above parameters meet the conditions for applying the large language model in the terminal, the terminal applies the large language model. If the above parameters do not meet the conditions for applying the large language model in the terminal, it is determined whether the large language model application can be performed on the network side based on the network-side parameters, wherein the network-side parameters at least include the data transmission cost and the data transmission delay of the terminal. If the network-side parameters meet the conditions for applying the large language model on the network side, the terminal transmits data to the network side for the application of the large language model. During the application of the large language model on the network side, the terminal monitors the parameters thereof at any time, determines whether the parameters of the terminal meet the conditions for applying the large language model, and if the conditions for applying the large language model in the terminal are met, the terminal is dynamically switched back to the application of the large language model in the terminal.

[0035] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0036] It should also be noted that the relational terms herein, such as first and second, are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus including the element.

[0037] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0038] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for applying a large language model in a wireless cellular system, applied to data processing of a terminal, characterized in that: include: For a large language model-based application in a terminal, the terminal uses a hierarchical triggering strategy to dynamically switch its large language model application mode. The hierarchical triggering strategy first determines whether terminal parameters meet conditions for large language model application, where the parameters include at least the remaining battery life of the terminal, whether the large language model can provide answers, and the large language model response delay. If the above parameters meet the conditions for applying the large language model in the terminal, the large language model is applied in the terminal. If the above parameters do not meet the conditions for applying the large language model in the terminal, whether the large language model can be applied on the network side is determined based on network side parameters, where the network side parameters include at least the terminal's data transmission cost and data transmission delay. If the network-side parameters meet the conditions for applying the large language model on the network side, the terminal transmits the data to the network side for application of the large language model. During the application of the large language model on the network side, the terminal monitors its own parameters at any time to determine whether the terminal parameters meet the conditions for application of the large language model. If the conditions for application of the large language model on the terminal are met, the terminal dynamically switches back to application of the large language model on the terminal. The determining whether the large language model can be applied on the network side based on the network side parameters at least includes: First, the data transmission fee of the terminal is less than the second threshold; Secondly, the data transmission delay of the terminal is less than the third threshold; The terminal monitors its own parameters at any time to determine whether the terminal parameters meet the conditions for application of the large language model, which at least includes: First, the remaining power of the terminal is greater than the first threshold. Secondly, large language models can provide answers within the expected time.

2. The method for applying a large language model in a wireless cellular system according to claim 1, characterized in that: The application based on the large language model at least includes: Training, learning, and inference of large language models for text summarization, language translation, audio optimization, video optimization, or wireless signal processing.

3. The method for applying a large language model in a wireless cellular system according to claim 1, wherein: The determination of whether the terminal parameters meet the conditions for application of the large language model includes at least: First, the remaining power of the terminal is greater than the first threshold. Secondly, large language models can provide answers within the expected time.

Images