Data processing method and device

Abstract

A data processing method is applied to the field of artificial intelligence and comprises the following steps: acquiring a plurality of subsequences which are subsequences of an input sequence; according to the plurality of subsequences, through a machine learning model, obtaining a processing result of each subsequence and uncertainty of each processing result; selecting at least one sub-sequence from the plurality of sub-sequences according to the uncertainty; and obtaining a processing result of the input sequence through the machine learning model according to the at least one sub-sequence. According to the method, the length of the input sequence can be reduced, the machine learning model can perform long-sequence reasoning without training and fine tuning, the cost of model training is reduced, screening can be performed from a plurality of subsequences based on the uncertainty of the processing result of each subsequence, and the accuracy of model training is improved. And reasoning is carried out based on at least one screened sub-sequence, so that a better processing result can be obtained.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more particularly to a data processing method and apparatus thereof. Background Technology

[0002] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0003] Due to limited training resources, machine learning models (such as large language models) typically use relatively short training data during training. During inference, when the length of the inference exceeds the maximum length allowed for model training, the accuracy of the inference drops rapidly. This means that the size of the training window for large models determines the upper limit of the model's inference length, significantly limiting the application of large models in long sequence inference.

[0004] Therefore, there is an urgent need for a method that enables machine learning models to process long sequences. Summary of the Invention

[0005] In a first aspect, this application provides a data processing method, the method comprising: acquiring a plurality of subsequences, the plurality of subsequences being subsequences of an input sequence; obtaining a processing result for each of the plurality of subsequences and an uncertainty of each of the processing results through a machine learning model; selecting at least one subsequence from the plurality of subsequences based on the uncertainty; and obtaining a processing result for the input sequence through the machine learning model based on the at least one subsequence.

[0006] Uncertainty can characterize the confidence level of the processing result. For example, the processing result can be multiple candidate results and their probability values. If one candidate result has a high probability value while the others have low probabilities, then the uncertainty of the processing result is low. For example, uncertainty can be the entropy corresponding to the conditional probability distribution generated during the inference process.

[0007] Taking language models as an example of machine learning models, if the uncertainty of the processing result of a certain subsequence is lower, it indicates that the prediction of the processing result of the subsequence is more accurate, and the content of the segment is more relevant to the question (i.e., the query).

[0008] In this way, on the one hand, since the data input into the machine learning model is a subsequence of the input sequence, the length of the input sequence can be reduced. The machine learning model can perform long sequence inference without training and fine-tuning, which reduces the cost of model training. On the other hand, based on the uncertainty of the processing result of each subsequence, multiple subsequences can be selected, and inference can be performed based on at least one selected subsequence, which can yield better processing results.

[0009] In one possible implementation, selecting at least one subsequence from the plurality of subsequences based on the uncertainty includes: selecting a preset number of subsequences with the lowest uncertainty from the plurality of subsequences.

[0010] In one possible implementation, the machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

[0011] In one possible implementation, the length of the input sequence exceeds the length of the input data supported by the machine learning model, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

[0012] In one possible implementation, the input sequence is divided into M first sequence units and second sequence units, the second sequence unit being a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

[0013] In one possible implementation, obtaining the processing result of each of the multiple subsequences through a machine learning model includes: processing the multiple subsequences in parallel through a machine learning model to obtain the processing result of each of the multiple subsequences.

[0014] In one possible implementation, the multiple subsequences can be processed in parallel using a machine learning model to obtain the processing result for each subsequence. After dividing a long sequence into several short sequences, the short sequences are independent of each other, and each part will obtain its own processing result. After splitting the computation, the length of the data sent to the machine learning model for computation each time becomes smaller, which may lead to a decrease in computational efficiency. In this case, the split short sequences can be constructed into a batch and input into the machine learning model for computation all at once, thereby improving computational efficiency and reducing computational latency.

[0015] In one possible implementation, obtaining the processing result of the input sequence based on the at least one subsequence through the machine learning model includes: reading the key-value (KV) data required for processing the at least one subsequence from the host's memory into the memory of a computing device, wherein the machine learning model performs data processing on the computing device; obtaining the processing result of the input sequence and the target KV data of the unprocessed data from the machine learning model based on the KV data and the data in the at least one subsequence that has not been processed by the machine learning model; deleting the KV data required for processing the at least one subsequence from the memory, and writing the target KV data back into the host's memory.

[0016] When incremental inference reaches a point where the sentence length exceeds a certain threshold, the memory usage of the key-value cache and the memory usage of the model parameters themselves will exceed the chip's physical memory limit. The segmented solution algorithm makes memory management possible by saving historical key-value states to a larger host memory. When calculating a specific segment, the corresponding key-value states are loaded from the host memory into the device memory for computation. In this way, memory utilization can be improved, and longer sequences can be inferred using a small amount of chip resources.

[0017] Secondly, this application provides a data processing apparatus, the apparatus comprising:

[0018] The acquisition module is used to acquire multiple sub-sequences, wherein the multiple sub-sequences are sub-sequences of the input sequence;

[0019] The processing module is configured to obtain a processing result for each of the plurality of subsequences and an uncertainty of each processing result through a machine learning model; select at least one subsequence from the plurality of subsequences based on the uncertainty; and obtain a processing result for the input sequence based on the at least one subsequence through the machine learning model.

[0020] In one possible implementation, the processing module is specifically used to: select a preset number of subsequences with the lowest uncertainty from the plurality of subsequences.

[0021] In one possible implementation, the machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

[0022] In one possible implementation, the length of the input sequence exceeds the length of the input data supported by the machine learning model, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

[0023] In one possible implementation, the input sequence is divided into M first sequence units and second sequence units, the second sequence unit being a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

[0024] In one possible implementation, the processing module is specifically used for:

[0025] The multiple subsequences are processed in parallel using a machine learning model to obtain the processing result for each subsequence.

[0026] In one possible implementation, the processing module is specifically used for:

[0027] The machine learning model reads the KV data required for processing the at least one subsequence from the host's memory into the memory of the computing device, and performs data processing on the computing device.

[0028] Based on the KV data and the data in at least one subsequence that has not been processed by the machine learning model, the processing result of the input sequence and the target KV data of the data that has not been processed by the machine learning model are obtained through the machine learning model.

[0029] The KV data required for processing the at least one subsequence is deleted from the memory, and the target KV data is written to the memory of the host.

[0030] Thirdly, embodiments of this application provide a data processing apparatus, which may include a memory, a processor, and a bus system, wherein the memory is used to store a program, and the processor is used to execute the program in the memory to perform the methods described in the first aspect above and any of its optional methods.

[0031] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the methods described in the first aspect and any of its optional methods.

[0032] Fifthly, embodiments of this application provide a computer program that, when run on a computer, causes the computer to perform the first aspect and any of its optional methods described above.

[0033] Sixthly, this application provides a chip system including a processor for supporting an execution device or training device in implementing the functions involved in the foregoing aspects, such as transmitting or processing data involved in the foregoing methods; or, information. In one possible design, the chip system further includes a memory for storing program instructions and data necessary for the execution device or training device. This chip system may be composed of chips or may include chips and other discrete devices. Attached Figure Description

[0034] Figure 1A A structural diagram illustrating the main framework of artificial intelligence;

[0035] Figure 1B Hezhi Figure 1C This is a schematic diagram of the application system framework of the present invention;

[0036] Figure 1D This is a schematic diagram of an optional hardware structure for a terminal.

[0037] Figure 2 This is a schematic diagram of the structure of a server;

[0038] Figures 3 to 5 This is a schematic diagram of a system architecture according to this application;

[0039] Figure 6 A process for providing a cloud service;

[0040] Figure 7 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0041] Figure 8 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0042] Figure 9 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0043] Figure 10 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0044] Figure 11 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0045] Figure 12 This is an illustration of beneficial effects;

[0046] Figure 13 A schematic diagram of the structure of a data processing apparatus provided in an embodiment of this application;

[0047] Figure 14A schematic diagram of the structure of a terminal device provided in an embodiment of this application;

[0048] Figure 15 A schematic diagram of the structure of a server provided in an embodiment of this application;

[0049] Figure 16 This is a schematic diagram of a chip structure provided in an embodiment of this application. Detailed Implementation

[0050] The embodiments of the present invention will now be described with reference to the accompanying drawings. The terminology used in the embodiments section is for illustrative purposes only and is not intended to limit the scope of the invention.

[0051] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0052] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0053] The terms “substantially,” “about,” and similar terms used herein are used as approximations rather than as terms of degree, and are intended to take into account the inherent biases of measurements or calculations known to those skilled in the art. Furthermore, the use of “may” in describing embodiments of the invention refers to “one or more possible embodiments.” The terms “use,” “using,” and “used” used herein are to be considered synonymous with the terms “utilize,” “utilizing,” and “utilized,” respectively. Additionally, the term “exemplary” is intended to refer to an instance or illustration.

[0054] First, the overall workflow of the artificial intelligence system is described; please refer to [link / reference]. Figure 1A , Figure 1AThe diagram illustrates a structural framework for artificial intelligence (AI). The framework is further elaborated below along two dimensions: the "Intelligent Information Chain" (horizontal axis) and the "IT Value Chain" (vertical axis). The "Intelligent Information Chain" reflects a series of processes from data acquisition to processing. For example, it could be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. In this process, data undergoes a condensation process of "data—information—knowledge—wisdom." The "IT Value Chain" reflects the value that AI brings to the information technology industry, from the underlying infrastructure of human intelligence and information (provided and processed through technological means) to the industrial ecosystem of the system.

[0055] (1) Infrastructure

[0056] Infrastructure provides computing power to support artificial intelligence systems, enabling communication with the external world and providing support through a basic platform. This communication occurs through sensors; computing power is provided by intelligent chips (hardware acceleration chips such as CPUs, NPUs, GPUs, ASICs, and FPGAs); and the basic platform includes distributed computing frameworks and related platform guarantees and support, which may include cloud storage and computing, interconnected networks, etc. For example, sensors communicate with the outside world to acquire data, and this data is provided to intelligent chips in the distributed computing system provided by the basic platform for computation.

[0057] (2) Data

[0058] The data at the next layer of infrastructure is used to represent the data sources in the field of artificial intelligence. The data involves graphics, images, voice, text, and IoT data from traditional devices, including business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.

[0059] (3) Data processing

[0060] Data processing typically includes methods such as data training, machine learning, deep learning, search, reasoning, and decision-making.

[0061] Among them, machine learning and deep learning can perform intelligent information modeling, extraction, preprocessing, and training of data by symbolizing and formalizing it.

[0062] Reasoning refers to the process in which, in a computer or intelligent system, the machine thinks and solves problems by simulating human intelligent reasoning, based on reasoning control strategies and using formalized information. Typical functions include search and matching.

[0063] Decision-making refers to the process of making decisions based on intelligent information after reasoning, and it typically provides functions such as classification, sorting, and prediction.

[0064] (4) General ability

[0065] After the data processing mentioned above, the results of the data processing can be used to form some general capabilities, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.

[0066] (5) Smart Products and Industry Applications

[0067] Intelligent products and industry applications refer to products and applications of artificial intelligence systems in various fields. They are the encapsulation of overall artificial intelligence solutions, productizing intelligent information decision-making and realizing practical applications. Their application areas mainly include: intelligent terminals, intelligent transportation, intelligent healthcare, autonomous driving, smart cities, etc.

[0068] First, we will introduce the application scenarios of this application. This application can be used, but is not limited to, applications with generative artificial intelligence (AIGC) functionality (hereinafter referred to as synthetic applications) or cloud services provided by cloud-side servers, etc., which will be introduced separately below:

[0069] I. Synthesis Applications

[0070] The product form of this application embodiment can be a synthetic application. Synthetic applications can run on terminal devices or cloud-based servers.

[0071] In one possible implementation, a synthesis application can perform data generation tasks based on input data (e.g., images, text, audio, video, etc.), wherein the synthesis application can perform data generation tasks in response to the input data (e.g., images, text, audio, video, etc.) to obtain generated data.

[0072] For example, the task of generating the above data can be, but is not limited to:

[0073] Text generation task: It can generate various types of text content, including news reports, blog posts, product descriptions, social media posts, etc. It can generate logical and coherent text based on given themes and requirements.

[0074] Image generation task: This task can generate images, including illustrations, artworks, design drafts, etc. It can generate image content related to given descriptions or keywords.

[0075] Audio generation task: This task can generate speech content, including text readings and responses from voice assistants. It can simulate human speech characteristics and intonation, making the generated speech sound more natural.

[0076] Content summarization and conclusion tasks: It can read large amounts of text content and generate summaries or conclusions. It can extract key information from the text and present it to the user in a concise manner.

[0077] Language translation task: This function performs language translation, converting text from one language to another. It can handle multiple language pairs and provide accurate translation results.

[0078] Automated replies and customer service: This can be used to automatically answer user questions and provide customer service. It can understand the user's intent and provide accurate answers or suggestions.

[0079] In one possible implementation, a user can open a synthesis application installed on a terminal device and input data (such as images, text, audio, video, etc.). The synthesis application can generate data from the input data using the method provided in the embodiments of this application and present the generated data to the user (the presentation method may include, but is not limited to, displaying, saving, uploading to the cloud, etc.).

[0080] In one possible implementation, a user can open a synthesis application installed on a terminal device and input data. The synthesis application can then send the input data to a cloud-based server. The cloud-based server uses the method provided in this application to generate data from the input data and sends the generated data back to the terminal device. The terminal device can then present the generated data to the user (the presentation method may include, but is not limited to, displaying, saving, or uploading to the cloud).

[0081] The following sections will describe the synthetic application in this application from the perspectives of functional architecture and product architecture that implements the functions.

[0082] Reference Figure 1B , Figure 1B This is a schematic diagram of the functional architecture of the synthetic application in the embodiments of this application:

[0083] In one possible implementation, such as Figure 1B As shown, the synthetic application 102 can receive input parameters 101 (e.g., containing input data) and generate generated data 103. The synthetic application 102 can execute on at least one computer system (for example) and includes computer code that, when executed by one or more computers, causes the computers to execute a natural language model trained by the methods provided in the embodiments of this application.

[0084] Reference Figure 1C , Figure 1C This is a schematic diagram of the entity architecture for running a synthetic application in the embodiments of this application:

[0085] See Figure 1C , Figure 1C A schematic diagram of a system architecture is shown. The system may include a terminal 100 and a server 200. The server 200 may include one or more servers (…). Figure 1C (The example includes a server), and server 200 can provide synthesis function services for one or more terminals.

[0086] The terminal 100 may have a synthesis application installed or a webpage related to the synthesis function open. The application and webpage can provide an interface. The terminal 100 can receive relevant parameters input by the user on the synthesis function interface and send the parameters to the server 200. The server 200 can obtain the processing result based on the received parameters and return the processing result to the terminal 100.

[0087] It should be understood that in some optional implementations, the terminal 100 can also complete the action of obtaining the processing result based on the received parameters on its own, without the need for the server to cooperate. This application embodiment is not limited to this.

[0088] The following description Figure 1C The product form of the mid-terminal 100;

[0089] The terminal 100 in this application embodiment can be a mobile phone, tablet computer, wearable device, vehicle device, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc., and this application embodiment does not impose any restrictions on it.

[0090] Figure 1D A schematic diagram of an optional hardware structure for terminal 100 is shown.

[0091] refer to Figure 1D As shown, the terminal 100 may include a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160 (optional), a speaker 161 (optional), a microphone 162 (optional), a processor 170, an external interface 180, a power supply 190, and other components. Those skilled in the art will understand that... Figure 1D These are merely examples of terminals or multi-functional devices and do not constitute a limitation on terminals or multi-functional devices. They may include more or fewer components than shown in the illustration, or combine certain components, or use different components.

[0092] The input unit 130 can be used to receive input numerical or character information, and to generate key signal inputs related to user settings and function control of the portable multi-functional device. Specifically, the input unit 130 may include a touchscreen 131 (optional) and / or other input devices 132. The touchscreen 131 can collect touch operations performed by the user on or near it (such as operations performed by the user using fingers, knuckles, styluses, or any suitable object on or near the touchscreen), and drive the corresponding connection devices according to a pre-set program. The touchscreen can detect the user's touch actions, convert the touch actions into touch signals and send them to the processor 170, and can receive and execute commands sent by the processor 170; the touch signal includes at least touch point coordinate information. The touchscreen 131 can provide an input interface and an output interface between the terminal 100 and the user. In addition, various types of touchscreens, such as resistive, capacitive, infrared, and surface acoustic wave, can be used to implement the touchscreen. Besides the touchscreen 131, the input unit 130 may also include other input devices. Specifically, other input devices 132 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons 132, power buttons 133, etc.), trackball, mouse, joystick, etc.

[0093] Among them, the input device 132 can receive input data, etc.

[0094] The display unit 140 can be used to display information input by the user or information provided to the user, various menus of the terminal 100, interactive interfaces, file display, and / or playback of any multimedia file. In this embodiment, the display unit 140 can be used to display the interface of a synthesis application, generated data, etc.

[0095] The memory 120 can be used to store instructions and data. The memory 120 may primarily include an instruction storage area and a data storage area. The data storage area can store various types of data, such as multimedia files and text. The instruction storage area can store software units such as operating systems, applications, and instructions required for at least one function, or subsets or extended sets thereof. It may also include non-volatile random access memory. It provides the processor 170 with hardware, software, and data resources for managing the computing device, supporting control software and applications. It is also used for storing multimedia files, as well as storing running programs and applications.

[0096] The processor 170 is the control center of the terminal 100. It connects various parts of the terminal 100 via various interfaces and lines. By running or executing instructions stored in the memory 120 and calling data stored in the memory 120, it performs various functions and processes data of the terminal 100, thereby controlling the terminal device as a whole. Optionally, the processor 170 may include one or more processing units; preferably, the processor 170 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 170. In some embodiments, the processor and memory can be implemented on a single chip; in some embodiments, they can also be implemented separately on independent chips. The processor 170 can also be used to generate corresponding operation control signals, send them to the corresponding components of the computing processing device, read and process data in the software, especially read and process data and programs in the memory 120, so that the various functional modules therein perform corresponding functions, thereby controlling the corresponding components to act according to the instructions.

[0097] The memory 120 can be used to store software code related to the data processing method, and the processor 170 can execute the steps of the chip's data processing method, and can also schedule other units (such as the above-mentioned input unit 130 and display unit 140) to achieve the corresponding functions.

[0098] The radio frequency unit 110 (optional) can be used for receiving and transmitting signals during information transmission or calls. For example, it can receive downlink information from the base station and process it for the processor 170; additionally, it can transmit uplink data to the base station. Typically, the RF circuit includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier (LNA), a duplexer, etc. Furthermore, the radio frequency unit 110 can also communicate wirelessly with network devices and other devices. This wireless communication can use any communication standard or protocol, including but not limited to Global System for Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0099] In this embodiment of the application, the radio frequency unit 110 can send input data to the server 200 and receive generated data sent by the server 200.

[0100] It should be understood that the radio frequency unit 110 is optional and can be replaced with other communication interfaces, such as a network port.

[0101] The terminal 100 also includes a power supply 190 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 170 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0102] Terminal 100 also includes an external interface 180, which can be a standard Micro USB interface or a multi-pin connector, which can be used to connect terminal 100 to other devices for communication or to connect a charger to charge terminal 100.

[0103] Although not shown, terminal 100 may also include a flash, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with various functions, etc., which will not be described in detail here. Some or all of the methods described below can be applied to, for example... Figure 1D In the terminal 100 shown.

[0104] The following description Figure 1C The product form of the mid-range server 200;

[0105] Figure 2 A structural diagram of a server 200 is provided, as follows: Figure 2 As shown, server 200 includes bus 201, processor 202, communication interface 203, and memory 204. Processor 202, memory 204, and communication interface 203 communicate with each other via bus 201.

[0106] Bus 201 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 2 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0107] The processor 202 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0108] Memory 204 may include volatile memory, such as random access memory (RAM). Memory 204 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0109] The memory 204 can be used to store software code related to the data processing method, and the processor 202 can execute the steps of the chip's data processing method, and can also schedule other units to achieve corresponding functions.

[0110] It should be understood that the aforementioned terminal 100 and server 200 can be centralized or distributed devices. The processors (e.g., processor 170 and processor 202) in the aforementioned terminal 100 and server 200 can be hardware circuits (such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), general-purpose processors, digital signal processors (DSPs), microprocessors or microcontrollers, etc.) or combinations of these hardware circuits. For example, the processor can be a hardware system with instruction execution capabilities, such as a CPU or DSP, or a hardware system without instruction execution capabilities, such as an ASIC or FPGA, or a combination of the aforementioned hardware systems without instruction execution capabilities and hardware systems with instruction execution capabilities.

[0111] It should be understood that the steps related to the model inference process in the embodiments of this application involve AI-related operations. When performing AI operations, the instruction execution architecture of the terminal device and server is not limited to the processor-memory architecture described above. The following section will further explain... Figure 5 The system architecture provided in the embodiments of this application will be described in detail.

[0112] Figure 5 This is a schematic diagram of the system architecture provided for an embodiment of this application. Figure 5 As shown, the system architecture 500 includes an execution device 510, a training device 520, a database 530, a client device 540, a data storage system 550, and a data acquisition system 560.

[0113] The execution device 510 includes a calculation module 511, an I / O interface 512, a preprocessing module 513, and a preprocessing module 514. The calculation module 511 may include a target model / rule 501, while the preprocessing modules 513 and 514 are optional.

[0114] The execution device 510 can be a terminal device or a server that runs the aforementioned synthetic application.

[0115] The data acquisition device 560 is used to collect training samples. Training samples can be program files (including program code and program input data), etc. After collecting the training samples, the data acquisition device 560 stores these training samples in the database 530.

[0116] The training device 520 can maintain training samples in the database 530 to obtain the target model / rule 501 from the neural network to be trained.

[0117] It should be noted that in practical applications, the training samples maintained in database 530 may not all come from the data acquisition device 560; they may also be received from other devices. Furthermore, it should be noted that training device 520 may not necessarily train the target model / rule 501 entirely based on the training samples maintained in database 530; it may also obtain training samples from the cloud or other sources for model training. The above description should not be construed as limiting the embodiments of this application.

[0118] The target model / rule 501 trained using training device 520 can be applied to different systems or devices, such as... Figure 5 The execution device 510 shown can be a terminal, such as a mobile phone terminal, tablet computer, laptop computer, augmented reality (AR) / virtual reality (VR) device, vehicle terminal, etc., or it can be a server, etc.

[0119] Specifically, the training device 520 can transfer the trained model to the execution device 510.

[0120] exist Figure 5 In the execution device 510, an input / output (I / O) interface 512 is configured for data interaction with external devices. Users can input data to the I / O interface 512 through the client device 540 (e.g., input data in the embodiment of this application).

[0121] Preprocessing modules 513 and 514 are used to preprocess the input data received from the I / O interface 512. It should be understood that preprocessing modules 513 and 514 may be absent, or only one preprocessing module may be used. When preprocessing modules 513 and 514 are absent, the calculation module 511 can be used directly to process the input data.

[0122] During the preprocessing of input data by the execution device 510, or during the calculation module 511 of the execution device 510 performing calculations and other related processes, the execution device 510 can call data, code, etc. in the data storage system 550 for corresponding processing, or store the data, instructions, etc. obtained from the corresponding processing into the data storage system 550.

[0123] Finally, the I / O interface 512 provides the processing results (such as generated data) to the client device 540, thereby providing them to the user.

[0124] exist Figure 5In the illustrated scenario, the user can manually provide input data, which can be done through the interface provided by I / O interface 512. Alternatively, the client device 540 can automatically send input data to I / O interface 512. If user authorization is required for the client device 540 to automatically send input data, the user can set the corresponding permissions in the client device 540. The user can view the output results of the execution device 510 on the client device 540, which can be presented in various forms such as display, sound, or animation. The client device 540 can also act as a data acquisition terminal, collecting the input data and output results of the input I / O interface 512 as shown in the figure, and storing them as new sample data in database 530. Alternatively, data can be collected directly from the I / O interface 512 without going through the client device 540, using the input data and output results of the input I / O interface 512 as shown in the figure, and storing them as new sample data in database 530.

[0125] It is worth noting that, Figure 5 This is merely a schematic diagram of a system architecture provided in an embodiment of this application. The positional relationships between the devices, components, modules, etc., shown in the diagram do not constitute any limitation. For example, in Figure 5 In this context, the data storage system 550 is an external storage device relative to the execution device 510. However, in other cases, the data storage system 550 may also be placed within the execution device 510. It should be understood that the aforementioned execution device 510 may be deployed within the client device 540.

[0126] From the inference side of the model:

[0127] In this embodiment, the computing module 511 of the execution device 510 can obtain the code stored in the data storage system 550 to implement the steps related to the model reasoning process in this embodiment.

[0128] In this embodiment of the application, the computing module 511 of the execution device 510 may include hardware circuits (such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), general-purpose processors, digital signal processors (DSPs), microprocessors or microcontrollers, etc.) or combinations of these hardware circuits. For example, the training device 520 may be a hardware system with instruction execution capabilities, such as a CPU or DSP, or a hardware system without instruction execution capabilities, such as an ASIC or FPGA, or a combination of the aforementioned hardware systems without instruction execution capabilities and hardware systems with instruction execution capabilities.

[0129] Specifically, the computing module 511 of the execution device 510 can be a hardware system with the function of executing instructions. The steps related to the model inference process provided in this application embodiment can be software code stored in the memory. The computing module 511 of the execution device 510 can obtain the software code from the memory and execute the obtained software code to implement the steps related to the model inference process provided in this application embodiment.

[0130] It should be understood that the computing module 511 of the execution device 510 can be a combination of a hardware system without the function of executing instructions and a hardware system with the function of executing instructions. Some steps related to the model reasoning process provided in the embodiments of this application can also be implemented by the hardware system in the computing module 511 of the execution device 510 without the function of executing instructions, which is not limited here.

[0131] From the training side of the model:

[0132] In this embodiment of the application, the training device 520 can access the memory ( Figure 5 (Not shown in the diagram, but can be integrated into the training device 520 or deployed separately from the training device 520) The code stored in the diagram can be used to implement the steps related to model training in the embodiments of this application.

[0133] In this embodiment of the application, the training device 520 may include hardware circuits (such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), general-purpose processors, digital signal processors (DSPs), microprocessors or microcontrollers, etc.) or combinations of these hardware circuits. For example, the training device 520 may be a hardware system with instruction execution capabilities, such as a CPU or DSP, or a hardware system without instruction execution capabilities, such as an ASIC or FPGA, or a combination of the aforementioned hardware systems without instruction execution capabilities and hardware systems with instruction execution capabilities.

[0134] It should be understood that the training device 520 can be a combination of a hardware system without the function of executing instructions and a hardware system with the function of executing instructions. Some steps related to the neutralization model training provided in the embodiments of this application can also be implemented by the hardware system in the training device 520 without the function of executing instructions, which is not limited here.

[0135] II. Cloud services providing synthesis functionality provided by the server:

[0136] In one possible implementation, the server can provide composition services to the client side through an application programming interface (API).

[0137] In this process, the terminal device can send relevant parameters (such as input data) to the server through the API provided by the cloud. The server can obtain the processing result (such as generated data) based on the received parameters and return the processing result to the terminal.

[0138] The description of the terminal and server can be found in the above embodiments, and will not be repeated here.

[0139] like Figure 6 The process of using a synthetic functional cloud service provided by a cloud platform is illustrated.

[0140] 1. Activate and purchase content moderation services.

[0141] 2. Users can download the software development kit (SDK) corresponding to the content moderation service. Cloud platforms usually provide multiple development versions of the SDK for users to choose from according to their development environment needs, such as JAVA version SDK, Python version SDK, PHP version SDK, Android version SDK, etc.

[0142] 3. After downloading the corresponding version of the SDK to their local machine according to their needs, users can import the SDK project into their local development environment, configure and debug it in the local development environment, and develop other functions in the local development environment, thus forming an application that integrates the capabilities of composite functional classes.

[0143] 4. When a composition application is used, it can trigger an API call for the composition function when the composition function is required. When the application triggers the composition function, it sends an API request to the running instance of the composition function service in the cloud environment. The API request carries the input data, which is processed by the running instance in the cloud environment to obtain the processing result (such as generated data).

[0144] 5. The cloud environment returns the processing result to the application, thus completing a synthesis function service call.

[0145] In addition to applications and cloud services, the implementation of this application can also be in a large model inference acceleration library or a large model application SDK.

[0146] To better understand the solutions of the embodiments of this application, text generation will be used as an example below, combined with... Figures 3 to 4 A brief introduction to the possible application scenarios of the embodiments of this application is provided.

[0147] Figure 3 A natural language processing (NLP) system is illustrated, comprising user devices and data processing devices. The user devices include smart terminals such as mobile phones, personal computers, or information processing centers. The user devices are the initiators of natural language data processing, acting as the initiators of requests such as language question answering or queries; typically, users initiate requests through their user devices.

[0148] The aforementioned data processing equipment can be cloud servers, network servers, application servers, management servers, or other devices or servers with data processing capabilities. The data processing equipment receives queries / voice / text from smart terminals via an interactive interface, then performs language data processing through a storage device and a data processing processor, employing methods such as machine learning, deep learning, search, reasoning, and decision-making. The processing results are then fed back to the user device. The storage device in the data processing equipment can be a general term, including local storage and a database storing historical data. The database can be located on the data processing equipment or on other network servers.

[0149] exist Figure 3 In the natural language processing system shown, the user device can receive instructions from the user. For example, the user device can receive a piece of text input by the user and then send a request to the data processing device, so that the data processing device can perform natural language processing applications (such as natural language generation, text classification, text reasoning, named entity recognition, translation, etc.) on the piece of text obtained by the user device, thereby obtaining the processing results of the corresponding natural language processing applications on the piece of text (such as word prediction results, classification results, reasoning results, named entity recognition results, translation results, etc.).

[0150] In this embodiment of the application, the user equipment can receive instructions from the user. For example, the user equipment can receive a piece of text input by the user (e.g., input data) and then send a request to the data processing device, so that the data processing device performs a natural language processing application (e.g., text synthesis) on the piece of text obtained by the user equipment, thereby obtaining the processing result (e.g., generated data) of the corresponding natural language processing application on the piece of text.

[0151] The text is in Figure 3 In this application, the data processing device can process the above-mentioned text data using the method provided in the embodiments of this application.

[0152] Figure 4 This demonstrates another natural language processing system, in Figure 4 In this context, the user equipment (UE) directly functions as a data processing device. This UE can directly receive input from the user and process it directly through its own hardware. The specific process is similar to... Figure 3 Similar to the description above, it will not be repeated here.

[0153] Figure 4 This is a schematic diagram of the natural language processing related device 300 provided in the embodiments of this application.

[0154] The above Figure 3 and Figure 4 The user equipment in the context can specifically be Figure 4Local device 301 or local device 302 in the system. Figure 3 The data processing equipment in the middle can specifically be Figure 4 The execution device 310 in the process includes a data storage system 350 that can store the data to be processed by the execution device 310. The data storage system 350 can be integrated into the execution device 310 or set up in the cloud or on other network servers.

[0155] Figure 3 and Figure 4 The processor in the application can perform data training / machine learning / deep learning using neural network models or other models, and use the model finally trained or learned from the data (such as the natural language model in the embodiments of this application) to perform natural language processing applications (such as program synthesis, etc.) on text data (such as the input data text described in the embodiments of this application) to obtain the corresponding processing results.

[0156] Since the embodiments of this application involve a large number of neural network applications, for ease of understanding, the relevant terms and concepts such as neural networks involved in the embodiments of this application will be introduced below.

[0157] (1) Neural Network

[0158] A neural network can be composed of neural units, which can be defined as a computational unit that takes xs (i.e., input data) and an intercept of 1 as input. The output of this computational unit can be:

[0159]

[0160] Where s = 1, 2, ..., n, where n is a natural number greater than 1, Ws is the weight of xs, and b is the bias of the neural unit. f is the activation function of the neural unit, used to introduce nonlinear characteristics into the neural network to convert the input signal in the neural unit into an output signal. The output signal of this activation function can be used as the input of the next convolutional layer, and the activation function can be the sigmoid function. A neural network is a network formed by connecting multiple of the above-mentioned individual neural units together, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field, which can be a region composed of several neural units.

[0161] (2) Transformer layer

[0162] The neural network includes an embedding layer and at least one transformer layer. The at least one transformer layer can be N transformer layers (N being an integer greater than 0). Each transformer layer includes sequentially adjacent attention layers, add and normalize layers, feed-forward layers, and add and normalize layers. In the embedding layer, the current input is embedded to obtain multiple embedding vectors. In the attention layer, P input vectors are obtained from the layer above the first transformer layer. Using any first input vector among the P input vectors as the center, intermediate vectors corresponding to the first input vector are obtained based on the correlation between each input vector within a preset attention window and the first input vector. This process determines P intermediate vectors corresponding to the P input vectors. In the pooling layer, the P intermediate vectors are merged into Q output vectors, where the multiple output vectors obtained from the last transformer layer are used as feature representations of the current input.

[0163] (3) Attention mechanism

[0164] Attention mechanisms mimic the internal processes of biological observation—aligning internal experience with external senses to increase the precision of observation in specific areas. They enable the rapid sifting of high-value information from a large volume of data using limited attentional resources. Attention mechanisms can quickly extract important features from sparse data and are therefore widely used in natural language processing tasks, particularly machine translation. Self-attention mechanisms, an improvement on attention mechanisms, reduce reliance on external information and are better at capturing the internal correlations of data or features. The core idea of ​​attention mechanisms can be rewritten as follows:

[0165] In this formula, Lx = ||Source|| represents the length of the Source. The meaning is that the elements in the Source are imagined as a series of data pairs. Given a Query element in the Target, the similarity or relevance between the Query and each Key is calculated to obtain the weight coefficient of the Value corresponding to each Key. Then, the Values ​​are weighted and summed to obtain the final Attention value. Therefore, the Attention mechanism essentially performs a weighted sum of the Values ​​of the elements in the Source, while the Query and Key are used to calculate the weight coefficients of their corresponding Values. Conceptually, Attention can be understood as selectively filtering a small amount of important information from a large amount of information and focusing on this important information, ignoring most of the unimportant information. The focusing process is reflected in the calculation of the weight coefficients; the larger the weight, the more focused it is on its corresponding Value. That is, the weight represents the importance of the information, and the Value is the corresponding information. Self-attention can be understood as intra attention. The attention mechanism occurs between the elements of the Target (Query) and all elements of the Source. Self-attention refers to the attention mechanism that occurs between elements within the Source or between elements within the Target. It can also be understood as the attention calculation mechanism in the special case where Target = Source. The specific calculation process is the same, only the calculation object changes.

[0166] (4) Natural Language Processing (NLP)

[0167] Natural language is human language, and Natural Language Processing (NLP) is the processing of human language. NLP is a systematic process of analyzing, understanding, and extracting information from text data in an intelligent and efficient manner. By using NLP and its components, we can manage very large amounts of text data, perform numerous automated tasks, and solve a wide variety of problems, such as automatic summarization, machine translation (MT), named entity recognition (NER), relation extraction (RE), information extraction (IE), sentiment analysis, speech recognition, question answering systems, and topic segmentation, among others.

[0168] (5) Large Language Model: A large language model is a natural language processing model trained on large-scale data, typically with billions or tens of billions of parameters. These models learn the general features of language by studying a large amount of text data during the pre-training stage, and can then be fine-tuned on downstream tasks to adapt to the needs of specific tasks.

[0169] (6) Transformer: The transformer is a deep learning model architecture originally used for sequence-to-sequence tasks, such as machine translation. It uses a self-attention mechanism to process input sequences and has achieved great success in the field of natural language processing. Most large language models, such as BERT, GPT, and T5, are based on the Transformer architecture.

[0170] (7) Key-Value Cache: A key-value cache is a cache structure that stores key-value pairs. In large language models, key-value caches are often used to store intermediate results or other useful information that the model is processing text in order to improve efficiency. By using a key-value cache, the model can avoid redundant calculations when processing text.

[0171] (8) token: refers to the smallest unit in the text. Typically, a token can be a word, number, punctuation mark, single letter, or any single element that can be used for text analysis.

[0172] (9) Backpropagation algorithm

[0173] Convolutional neural networks can employ backpropagation (BP) to correct the parameters in the initial super-resolution model during training, thereby reducing the reconstruction error loss. Specifically, forward propagation of the input signal to the output generates an error loss; this error loss information is then propagated back to update the parameters in the initial super-resolution model, leading to convergence of the error loss. The backpropagation algorithm is an error-loss-driven backpropagation process aimed at obtaining the optimal parameters of the super-resolution model, such as the weight matrix.

[0174] (10) Loss Function

[0175] In training a deep neural network, to ensure the output closely approximates the desired predicted value, we compare the network's prediction with the target value. Based on the difference, we update the weight vector of each layer (usually pre-configuring parameters before the initial update). For example, if the prediction is too high, the weight vector is adjusted to predict a lower value. This adjustment continues until the deep neural network predicts the target value or a value very close to it. Therefore, we need to predefine "how to compare the difference between the predicted and target values," which is the loss function or objective function. These are important equations used to measure the difference between the predicted and target values. Taking the loss function as an example, a higher output value (loss) indicates a greater difference, and training the deep neural network becomes a process of minimizing this loss.

[0176] (11) Incremental Inference: Incremental inference allows the model to process only newly added parts of the input, rather than reprocessing the entire sequence each time. This is achieved by maintaining contextual information in the model's internal state, allowing the model to respond quickly when it receives new input. Incremental inference is particularly useful in interactive applications, such as chatbots or real-time translation, as it can significantly reduce latency and computational resource usage.

[0177] Due to limited training resources, machine learning models (such as large language models) typically use relatively short training data, generally less than 8k. During inference, when the inference length exceeds the maximum length allowed for model training, the inference accuracy drops rapidly. This means that the size of the training window for large models determines the upper limit of the model's inference length, significantly restricting the application of large models in long sequence inference.

[0178] This application proposes an inference framework for machine learning models that enables them to infer sequences of the maximum super-trained length without additional training or fine-tuning. This framework requires no modification to the model itself; it achieves fast and efficient inference simply by employing a novel scheduling method.

[0179] To address the aforementioned problems, embodiments of this application provide a data processing method. The model training method of this application embodiment will be described in detail below with reference to the accompanying drawings.

[0180] Reference Figure 7 , Figure 7 This is a flowchart illustrating a data processing method provided in an embodiment of this application, such as... Figure 7 As shown in the embodiment of this application, a data processing method may include steps 701 to 703, which are described in detail below.

[0181] 701. Obtain multiple subsequences, wherein the multiple subsequences are subsequences of the input sequence.

[0182] In one possible implementation, the input sequence is text data, image data, or audio data.

[0183] Machine learning models (such as large language models) typically use relatively short training data, generally less than 8k, during training. During inference, when the length of the input data exceeds the maximum length of the model's training data, the inference accuracy drops rapidly. In this embodiment, the input data for inference can be the input sequence as described in this embodiment, where the length of the input sequence exceeds the length of the input data supported by the machine learning model. The length of the input data supported by the machine learning model can be understood as the maximum length of the training data used during model training.

[0184] In this embodiment of the application, the input sequence can be divided into multiple subsequences, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

[0185] The subsequence can be obtained by concatenating the sequence units obtained by dividing the input sequence. For example, in one possible implementation, the input sequence can be divided into M first sequence units and second sequence units, where the second sequence unit is a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

[0186] For example, the input sequence can be evenly divided into m blocks, and the first m-1 blocks and the last block can be combined to form a new sequence with a length less than the pre-training length, while maintaining the same front-to-back relationship as the original sequence.

[0187] 702. Based on the plurality of subsequences, a machine learning model is used to obtain the processing result of each subsequence and the uncertainty of each processing result.

[0188] In one possible implementation, the processing result of each of the multiple subsequences can be obtained through a machine learning model.

[0189] In one possible implementation, the machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

[0190] For example, the newly formed "short" sequences can be input into the model to predict the next token, resulting in m predictions.

[0191] In one possible implementation, the multiple subsequences can be processed in parallel using a machine learning model to obtain the processing result for each subsequence. After dividing a long sequence into several short sequences, the short sequences are independent of each other, and each part will obtain its own processing result. After splitting the computation, the length of the data sent to the machine learning model for computation each time becomes smaller, which may lead to a decrease in computational efficiency. In this case, the split short sequences can be constructed into a batch and input into the machine learning model for computation all at once, thereby improving computational efficiency and reducing computational latency.

[0192] For example, you can refer to Figure 8 , Figure 8 The diagram illustrates a data processing procedure where the current inference data length is 4097. After partitioning, it is divided into three subsequences. These three subsequences can be used to construct a batching algorithm, which is then input into an LLM for computation.

[0193] In this embodiment of the application, in addition to the processing result, the uncertainty of each processing result can also be obtained.

[0194] Uncertainty can characterize the confidence level of the processing result. For example, the processing result can be multiple candidate results and their probability values. If one candidate result has a high probability value while the others have low probabilities, then the uncertainty of the processing result is low. For example, uncertainty can be the entropy corresponding to the conditional probability distribution generated during the inference process.

[0195] Taking language models as an example of machine learning models, if the uncertainty of the processing result of a certain subsequence is lower, it indicates that the prediction of the processing result of the subsequence is more accurate, and the content of the segment is more relevant to the question (i.e., the query).

[0196] 703. Select at least one subsequence from the plurality of subsequences based on the uncertainty.

[0197] In one possible implementation, a preset number of subsequences with the lowest uncertainty can be selected from the plurality of subsequences.

[0198] For example, at least one selected subsequence can be concatenated according to its temporal order in the input sequence, and the reasoning task can be completed based on the concatenated sequence.

[0199] In this way, on the one hand, since the data input into the machine learning model is a subsequence of the input sequence, the length of the input sequence can be reduced. The machine learning model can perform long sequence inference without training and fine-tuning, which reduces the cost of model training. On the other hand, based on the uncertainty of the processing result of each subsequence, multiple subsequences can be selected, and inference can be performed based on at least one selected subsequence, which can yield better processing results.

[0200] For example, taking a machine learning model as a language model, refer to... Figure 11 Given a large model and a long sequence inference task, the sequence is first segmented according to the original window size of the model (each segment is composed of a small part of the sequence and some words at the end of the original sequence) to ensure that the length of each segment does not exceed the window size; the next step is to use the large model to infer all segments (this step can be done in parallel), calculate the entropy corresponding to the conditional probability distribution generated during the inference process, and filter the segments based on the entropy, and use the filtered segments for inference.

[0201] 704. Based on the at least one subsequence, the processing result of the input sequence is obtained through the machine learning model.

[0202] In this embodiment, the method can be an incremental inference process. As incremental inference progresses, the length of the input sequence will continuously increase. In this case, it is necessary to adjust the way the subsequences are divided to ensure that the length of each subsequence does not exceed the input length that the machine learning model can support.

[0203] Here is one optional implementation method, see [link / reference] Figure 9The specific numerical values ​​can be adjusted based on the actual situation. The block division method can be as follows: the upper limit of the last block is set to 1 / 8 of the pre-training sentence length, and the lower limit is set to 1 / 16 of the pre-training sentence length. The upper limit of the size of other blocks is equal to 7 / 8 of the pre-training sentence length. When combined with incremental inference, the size of the penultimate block dynamically changes, increasing from 1 / 16 to 7 / 8 of the pre-training sentence length. After the previous blocks are calculated, the key-value vectors are stored in the cache, and then they do not need to be calculated again. For example, if the pre-training length is 2048, then the upper limit of the last block is 256, the lower limit is 128, and the upper limit of the size of other blocks is 1792.

[0204] This application proposes a scheme combining a segmented inference process and incremental inference key-value caching. The block partitioning method is pre-defined, and the key-value pairs corresponding to previously calculated blocks are cached to avoid subsequent redundant calculations.

[0205] In one possible implementation, key-value (KV) data required for processing the at least one subsequence can be read from the host's memory into the memory of a computing device, where the machine learning model performs data processing. Based on the KV data and data in the at least one subsequence that has not been processed by the machine learning model, the machine learning model obtains the processing result of the input sequence and the target KV data of the data that has not been processed by the machine learning model. The KV data required for processing the at least one subsequence is deleted from the memory, and the target KV data is written back to the host's memory.

[0206] When incremental inference reaches a point where the sentence length exceeds a certain threshold, the memory usage of the key-value cache and the memory usage of the model parameters themselves will exceed the chip's physical memory limit. The segmented solution algorithm makes memory management possible by saving historical key-value states to a larger host memory. When calculating a specific segment, the corresponding key-value states are loaded from the host memory into the device memory for computation. In this way, memory utilization can be improved, and longer sequences can be inferred using a small amount of chip resources.

[0207] This application proposes a scheme combining a segmented inference process and host-device segment exchange. When incremental inference reaches a point where the sentence length exceeds a certain threshold, the memory usage of the key-value cache and the model parameters themselves will exceed the chip's physical memory limit. The segmented solution algorithm enables better memory management by storing historical key-value states in a larger host memory. Only when calculating a specific segment are the corresponding key-value states loaded from the host memory into the device memory for computation. An exemplary flowchart can be found [link to flowchart]. Figure 10 As shown.

[0208] The beneficial effects of the embodiments of this application will be described below with reference to experiments:

[0209] The proposed framework was used to infer sequences from 0 to 16k using a 2k pre-trained model, and the results were compared with those of a 32k pre-trained model directly inferring the same sequences from 0 to 16k. Additionally, the results of segmented inference using the proposed framework with a 32k pre-trained model are also presented. Referring to Table 1, the values ​​in Table 1 represent accuracy. The proposed inference framework enables the 2k pre-trained model to infer sequences far exceeding its pre-training length, while maintaining an accuracy close to that of the 32k pre-trained model. Furthermore, using the proposed framework to infer the 32k pre-trained model further improves its accuracy.

[0210] Table 1

[0211]

[0212] Furthermore, a comparison was made with existing fine-tuning and non-fine-tuning algorithms on the task of finding a needle in a haystack, referring to... Figure 12 As shown.

[0213] The rightmost XL3M is the inference algorithm of this framework, the first two are existing fine-tuning algorithms, and 3 and 4 are existing non-fine-tuning algorithms. It can be seen that the algorithm of this scheme can achieve good results at any length, while the fine-tuning algorithm can only achieve correct results within its fine-tuning length, and the non-fine-tuning algorithm performs poorly.

[0214] Furthermore, this application compares the proposed framework with the original method, as well as various existing fine-tuning and non-fine-tuning methods, in terms of speed. All models were used to infer 128k sequences with an inference decoding length of 1k. Specific inference time consumption is shown in Table 2 below.

[0215] Table 2

[0216] Prefill time(s) Decoding time(s) Total time(s) PI 60.5 228.3 288.8 Yarn 59.4 337.4 406.8 PCW 14.8 224.1 238.9 StreamLLM 4.7 36.1 40.8 XL3M 22.1 36.2 58.3

[0217] As can be seen from the results in Table 2, the framework proposed in this scheme has a significant speed advantage compared to existing fine-tuning schemes.

[0218] Reference Figure 13 , Figure 13 This is a schematic diagram of the structure of a data processing apparatus provided in an embodiment of this application, such as... Figure 13 As shown in the figure, an embodiment of this application provides a data processing apparatus 1300, comprising:

[0219] The acquisition module 1301 is used to acquire multiple sub-sequences, wherein the multiple sub-sequences are sub-sequences of the input sequence;

[0220] For a detailed description of the acquisition module 1301, please refer to the above embodiments. Figure 7 The similarities between the corresponding embodiments will not be repeated here.

[0221] The processing module 1302 is configured to obtain a processing result for each of the plurality of subsequences and an uncertainty of each processing result through a machine learning model; select at least one subsequence from the plurality of subsequences based on the uncertainty; and obtain a processing result for the input sequence based on the at least one subsequence through the machine learning model.

[0222] For a detailed description of the processing module 1302, please refer to the above embodiments. Figure 7 The similarities between the corresponding embodiments will not be repeated here.

[0223] In one possible implementation, the processing module 1302 is specifically used to: select a preset number of subsequences with the lowest uncertainty from the plurality of subsequences.

[0224] In one possible implementation, the machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

[0225] In one possible implementation, the length of the input sequence exceeds the length of the input data supported by the machine learning model, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

[0226] In one possible implementation, the input sequence is divided into M first sequence units and second sequence units, the second sequence unit being a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

[0227] In one possible implementation, the processing module 1302 is specifically used for:

[0228] The multiple subsequences are processed in parallel using a machine learning model to obtain the processing result for each subsequence.

[0229] In one possible implementation, the processing module 1302 is specifically used for:

[0230] The machine learning model reads the KV data required for processing the at least one subsequence from the host's memory into the memory of the computing device, and performs data processing on the computing device.

[0231] Based on the KV data and the data in at least one subsequence that has not been processed by the machine learning model, the processing result of the input sequence and the target KV data of the data that has not been processed by the machine learning model are obtained through the machine learning model.

[0232] The KV data required for processing the at least one subsequence is deleted from the memory, and the target KV data is written to the memory of the host.

[0233] In one possible implementation, the input sequence is text data, image data, or audio data.

[0234] The following describes an execution device provided in an embodiment of this application. Please refer to [link / reference]. Figure 14 , Figure 14 This is a schematic diagram of an execution device provided in an embodiment of this application. The execution device 1400 can specifically be a virtual reality (VR) device, a mobile phone, a tablet, a laptop, a smart wearable device, a monitoring data processing device, or a server, etc., and is not limited thereto. Specifically, the execution device 1400 includes: a receiver 1401, a transmitter 1402, a processor 1403, and a memory 1404 (wherein the execution device 1400 may have one or more processors 1403). Figure 14 (Taking a processor as an example), processor 1403 may include application processor 14031 and communication processor 14032. In some embodiments of this application, receiver 1401, transmitter 1402, processor 1403 and memory 1404 may be connected via a bus or other means.

[0235] Memory 1404 may include read-only memory and random access memory, and provides instructions and data to processor 1403. A portion of memory 1404 may also include non-volatile random access memory (NVRAM). Memory 1404 stores processor and operation instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operation instructions may include various operation instructions for implementing various operations.

[0236] Processor 1403 controls the operation of the execution device. In specific applications, the various components of the execution device are coupled together through a bus system, which may include not only the data bus, but also power buses, control buses, and status signal buses. However, for clarity, all buses are referred to as the bus system in the diagram.

[0237] The methods disclosed in the embodiments of this application can be applied to or implemented by the processor 1403. The processor 1403 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by the integrated logic circuits in the hardware of the processor 1403 or by instructions in software form. The processor 1403 can be a general-purpose processor, a digital signal processor (DSP), a microprocessor, or a microcontroller, and may further include an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The processor 1403 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 1404. Processor 1403 reads the information from memory 1404 and, in conjunction with its hardware, completes the steps involved in the model inference process described above.

[0238] Receiver 1401 can be used to receive input digital or character information, and to generate signal inputs related to the settings and function control of the execution device. Transmitter 1402 can be used to output digital or character information through the first interface; transmitter 1402 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; transmitter 1402 may also include a display device such as a display screen.

[0239] This application also provides a server device; please refer to [link / reference]. Figure 15 , Figure 15This is a schematic diagram of a server structure provided in an embodiment of this application. Specifically, server 1500 is implemented by one or more servers. Server 1500 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 1515 (e.g., one or more processors) and memory 1532, and one or more storage media 1530 (e.g., one or more mass storage devices) for storing application programs 1542 or data 1544. The memory 1532 and storage media 1530 can be temporary or persistent storage. The program stored in storage media 1530 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the server. Furthermore, the CPU 1515 may be configured to communicate with storage media 1530 and execute the series of instruction operations in storage media 1530 on server 1500.

[0240] Server 1500 may also include one or more power supplies 1526, one or more wired or wireless network interfaces 1550, one or more input / output interfaces 1558; or, one or more operating systems 1541, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0241] In this embodiment, the central processing unit 1515 is used to execute the data processing method described in the above embodiment.

[0242] This application also provides a computer program product that, when run on a computer, causes the computer to perform steps as performed by the aforementioned execution device, or causes the computer to perform steps as performed by the aforementioned training device.

[0243] This application also provides a computer-readable storage medium storing a program for signal processing, which, when run on a computer, causes the computer to perform steps as performed by the aforementioned execution device, or causes the computer to perform steps as performed by the aforementioned training device.

[0244] The execution device, training device, or terminal device provided in this application embodiment can specifically be a chip. The chip includes a processing unit and a communication unit. The processing unit can be, for example, a processor, and the communication unit can be, for example, an input / output interface, pins, or circuits. The processing unit can execute computer execution instructions stored in the storage unit to cause the chip within the execution device to execute the data processing method described in the above embodiments, or to cause the chip within the training device to execute the data processing method described in the above embodiments. Optionally, the storage unit can be a storage unit within the chip, such as a register or cache. Alternatively, the storage unit can be a storage unit located outside the chip within the wireless access device, such as a read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, such as random access memory (RAM).

[0245] For details, please refer to Figure 16 , Figure 16 This is a schematic diagram of a chip provided in an embodiment of this application. The chip can be represented as a neural network processor (NPU) 1600. The NPU 1600 is mounted as a coprocessor on the host CPU, and tasks are assigned by the host CPU. The core part of the NPU is the arithmetic circuit 1603, which is controlled by the controller 1604 to extract matrix data from the memory and perform multiplication operations.

[0246] In some implementations, the arithmetic circuit 1603 internally includes multiple processing engines (PEs). In some implementations, the arithmetic circuit 1603 is a two-dimensional pulsating array. The arithmetic circuit 1603 can also be a one-dimensional pulsating array or other electronic circuits capable of performing mathematical operations such as multiplication and addition. In some implementations, the arithmetic circuit 1603 is a general-purpose matrix processor.

[0247] For example, suppose we have an input matrix A, a weight matrix B, and an output matrix C. The arithmetic circuit retrieves the corresponding data of matrix B from the weight memory 1602 and caches it in each PE of the arithmetic circuit. The arithmetic circuit retrieves the data of matrix A from the input memory 1601 and performs matrix operations with matrix B. The partial result or the final result of the obtained matrix is ​​stored in the accumulator 1608.

[0248] Unified memory 1606 is used to store input and output data. Weight data is directly transferred to weight memory 1602 via Direct Memory Access Controller (DMAC) 1605. Input data is also transferred to unified memory 1606 via DMAC.

[0249] BIU stands for Bus Interface Unit, which is used for interaction between the AXI bus and the DMAC and the Instruction Fetch Buffer (IFB) 1609.

[0250] The Bus Interface Unit (BIU) 1610 is used by the instruction fetch memory 1609 to fetch instructions from external memory, and also by the memory access controller 1605 to fetch the original data of the input matrix A or the weight matrix B from external memory.

[0251] The DMAC is mainly used to move input data from external memory DDR to unified memory 1606, or to weight data to weight memory 1602, or to input data to input memory 1601.

[0252] The vector computation unit 1607 includes multiple arithmetic processing units that, when needed, further process the output of the computation circuit 1603, such as vector multiplication, vector addition, exponential operations, logarithmic operations, size comparisons, etc. It is mainly used for computation in non-convolutional / fully connected layers of neural networks, such as batch normalization, pixel-level summation, and upsampling of feature planes.

[0253] In some implementations, the vector computation unit 1607 can store the processed output vector in the unified memory 1606. For example, the vector computation unit 1607 can apply a linear function, or a nonlinear function, to the output of the computation circuit 1603, such as performing linear interpolation on the feature planes extracted by the convolutional layer, or, for example, accumulating a vector of values ​​to generate activation values. In some implementations, the vector computation unit 1607 generates normalized values, pixel-level summed values, or both. In some implementations, the processed output vector can be used as an activation input to the computation circuit 1603, for example, for use in subsequent layers of the neural network.

[0254] The instruction fetch buffer 1609 connected to the controller 1604 is used to store the instructions used by the controller 1604.

[0255] The unified memory 1606, input memory 1601, weight memory 1602, and instruction fetch memory 1609 are all on-chip memories. External memory is proprietary to this NPU hardware architecture.

[0256] The processor mentioned above can be a general-purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits used to control the execution of the above program.

[0257] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0258] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0259] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0260] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A data processing method, characterized in that, The method includes: Obtain multiple subsequences, wherein the multiple subsequences are subsequences of the input sequence; Based on the multiple subsequences, a machine learning model is used to obtain the processing result of each subsequence and the uncertainty of each processing result; At least one subsequence is selected from the plurality of subsequences based on the uncertainty; Based on the at least one subsequence, the processing result of the input sequence is obtained through the machine learning model.

2. The method according to claim 1, characterized in that, The step of selecting at least one subsequence from the plurality of subsequences based on the uncertainty includes: Select the preset number of subsequences with the lowest uncertainty from the plurality of subsequences.

3. The method according to claim 1 or 2, characterized in that, The machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

4. The method according to any one of claims 1 to 3, characterized in that, The length of the input sequence exceeds the length of the input data supported by the machine learning model, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

5. The method according to any one of claims 1 to 4, characterized in that, The input sequence is divided into M first sequence units and second sequence units, the second sequence unit being a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

6. The method according to any one of claims 1 to 5, characterized in that, The step of obtaining the processing result of each of the multiple sub-sequences through a machine learning model includes: The multiple subsequences are processed in parallel using a machine learning model to obtain the processing result for each subsequence.

7. The method according to any one of claims 1 to 6, characterized in that, The step of obtaining the processing result of the input sequence based on the at least one sub-sequence through the machine learning model includes: The machine learning model reads the key-value (KV) data required for processing the at least one subsequence from the host's memory into the memory of the computing device, and performs data processing on the computing device. Based on the KV data and the data in at least one subsequence that has not been processed by the machine learning model, the processing result of the input sequence and the target KV data of the data that has not been processed by the machine learning model are obtained through the machine learning model. The KV data required for processing the at least one subsequence is deleted from the memory, and the target KV data is written to the memory of the host.

8. A data processing apparatus, characterized in that, The device includes: The acquisition module is used to acquire multiple sub-sequences, wherein the multiple sub-sequences are sub-sequences of the input sequence; The processing module is configured to obtain a processing result for each of the plurality of subsequences and an uncertainty of each processing result through a machine learning model; select at least one subsequence from the plurality of subsequences based on the uncertainty; and obtain a processing result for the input sequence based on the at least one subsequence through the machine learning model.

9. The apparatus according to claim 8, characterized in that, The processing module is specifically used to: select a preset number of subsequences with the lowest uncertainty from the plurality of subsequences.

10. The apparatus according to claim 8 or 9, characterized in that, The machine learning model is a language model, the input sequence is text, and the processing result is the prediction result of the next adjacent text unit.

11. The apparatus according to any one of claims 8 to 10, characterized in that, The length of the input sequence exceeds the length of the input data supported by the machine learning model, and the length of each subsequence does not exceed the length of the input data supported by the machine learning model.

12. The apparatus according to claim 11, characterized in that, The input sequence is divided into M first sequence units and second sequence units, the second sequence unit being a query, and each subsequence includes at least one of the M sequence units and the second sequence unit.

13. The apparatus according to claim 12, characterized in that, The processing module is specifically used for: The multiple subsequences are processed in parallel using a machine learning model to obtain the processing result for each subsequence.

14. The apparatus according to any one of claims 8 to 13, characterized in that, The processing module is specifically used for: The machine learning model reads the KV data required for processing the at least one subsequence from the host's memory into the memory of the computing device, and performs data processing on the computing device. Based on the KV data and the data in at least one subsequence that has not been processed by the machine learning model, the processing result of the input sequence and the target KV data of the data that has not been processed by the machine learning model are obtained through the machine learning model. Delete the KV data required for processing the at least one subsequence from the memory, and write the target KV data back into the host's memory.

15. A computer storage medium, characterized in that, The computer storage medium stores one or more instructions that, when executed by one or more computers, cause the one or more computers to perform the operation of the method according to any one of claims 1 to 7.

16. A computer program product, characterized in that, Includes computer-readable instructions that, when executed on a computer device, cause the computer device to perform the method as described in any one of claims 1 to 7.

17. A system comprising at least one processor and at least one memory; the processor and the memory are connected via a communication bus and communicate with each other. The at least one memory is used to store code; The at least one processor is used to execute the code to perform the method as described in any one of claims 1 to 7.

18. A chip, characterized in that, It includes at least one processing unit and an interface circuit, the interface circuit being used to provide program instructions or data to the at least one processing unit, the at least one processing unit being used to execute the program instructions to implement the method of any one of claims 1 to 7.

Images