Data processing method and device, equipment and medium

Abstract

The invention provides a data processing method and device, equipment and a medium, and relates to the field of artificial intelligence, in particular to natural language processing, image processing, audio processing, multi-modal understanding, deep learning and large model technologies. According to the implementation scheme, the method comprises the following steps: acquiring a plurality of original lexical element vectors, wherein the plurality of original lexical element vectors respectively represent one of a text, an image and an audio; dividing the plurality of original lexical element vectors into a plurality of groups, and calculating by using the plurality of original lexical element vectors included in each group in the plurality of groups to obtain a first comprehensive vector corresponding to the group; inputting a plurality of first comprehensive vectors corresponding to the plurality of groups into the large model to obtain a second comprehensive vector output by the large model; and reconstructing by using the second comprehensive vector to obtain a plurality of target lexical vectors, wherein the plurality of target lexical vectors respectively represent one of the text, the image and the audio.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and more specifically to natural language processing, image processing, audio processing, multimodal understanding, deep learning and large model technology, and more specifically to a data processing method, a data processing device, an electronic device, a computer-readable storage medium and a computer program product. Background Art

[0002] Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0003] In recent years, large multimodal models have achieved great success in a variety of understanding and generation tasks. These models can process inputs of different modalities including text, images, and audio, and generate corresponding multimodal responses based on user questions.

[0004] The methods described in this section are not necessarily methods that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any method described in this section is considered to be prior art simply because it is included in this section. Similarly, unless otherwise indicated, the issues mentioned in this section should not be considered to have been recognized in any prior art. Summary of the invention

[0005] The present disclosure provides a data processing method, a data processing device, an electronic device, a computer-readable storage medium, and a computer program product.

[0006] According to one aspect of the present disclosure, a data processing method is provided, including: obtaining a plurality of original word-unit vectors, wherein each of the plurality of original word-unit vectors represents one of text, image and audio; dividing the plurality of original word-unit vectors into a plurality of groups, and using the plurality of original word-unit vectors included in each of the plurality of groups to calculate a first comprehensive vector corresponding to the group; inputting the plurality of first comprehensive vectors corresponding to the plurality of groups into a large model to obtain a second comprehensive vector output by the large model; and reconstructing a plurality of target word-unit vectors using the second comprehensive vectors, wherein each of the plurality of target word-unit vectors represents one of text, image and audio.

[0007] According to another aspect of the present disclosure, a data processing device is provided, including: an acquisition unit, configured to acquire a plurality of original word-unit vectors, wherein each of the plurality of original word-unit vectors represents one of text, image and audio; a grouping unit, configured to divide the plurality of original word-unit vectors into a plurality of groups, and calculate a first comprehensive vector corresponding to the group using the plurality of original word-unit vectors included in each of the plurality of groups; a large model, configured to receive a plurality of first comprehensive vectors corresponding to the plurality of groups to output a second comprehensive vector; and a reconstruction unit, configured to reconstruct a plurality of target word-unit vectors using the second comprehensive vector, wherein each of the plurality of target word-unit vectors represents one of text, image and audio.

[0008] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided by the present disclosure.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method provided by the present disclosure.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements the method provided by the present disclosure when executed by a processor.

[0011] According to one or more embodiments of the present disclosure, after a word is converted into a corresponding vector representation, multiple word units are grouped together and a total vector representation corresponding to each group is calculated, thereby achieving sequence shortening at the model input level while retaining the original multimodal word unit information. At the output level, the second comprehensive vector output by the large model is reconstructed to obtain a complete output word unit sequence, thereby achieving a dual improvement in effect and efficiency.

[0012] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0014] Figure 1 A schematic diagram showing an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure;

[0015] Figure 2 An exemplary process of a data processing method according to an embodiment of the present disclosure is shown;

[0016] Figure 3 An exemplary process of obtaining multiple original word element vectors according to an embodiment of the present disclosure is shown;

[0017] Figure 4 An exemplary process of obtaining multiple original word element vectors according to an embodiment of the present disclosure is shown;

[0018] Figure 5 A schematic diagram showing folding, processing and restoration of word element vectors according to an embodiment of the present disclosure is shown;

[0019] Figure 6 An exemplary block diagram of a data processing device according to an embodiment of the present disclosure is shown;

[0020] Figure 7 A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0021] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0022] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0023] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0024] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0025] Figure 1 FIG. 1 is a schematic diagram of an exemplary system 100 in which various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure. Figure 1 , the system 100 includes one or more client devices 101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105, and 106 may be configured to execute one or more applications.

[0026] In an embodiment of the present disclosure, the server 120 may run one or more services or software applications that enable execution of the method according to an embodiment of the present disclosure.

[0027] In some embodiments, server 120 may also provide other services or software applications, which may include non-virtualized environments and virtualized environments. In some embodiments, these services may be provided as web-based services or cloud services, such as provided to users of client devices 101, 102, 103, 104, 105, and / or 106 under a software as a service (SaaS) model.

[0028] exist Figure 1 In the configuration shown, the server 120 may include one or more components that implement the functions performed by the server 120. These components may include software components, hardware components, or a combination thereof that can be executed by one or more processors. Users operating client devices 101, 102, 103, 104, 105, and / or 106 may in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that a variety of different system configurations are possible, which may differ from the system 100. Therefore, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0029] The user may use client devices 101, 102, 103, 104, 105 and / or 106 to obtain data processed by the method according to the embodiments of the present disclosure. The client device may provide an interface that enables the user of the client device to interact with the client device. The client device may also output information to the user via the interface. Figure 1 Only six client devices are depicted, but one skilled in the art will appreciate that the present disclosure may support any number of client devices.

[0030] Client devices 101, 102, 103, 104, 105 and / or 106 may include various types of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, game systems, thin clients, various messaging devices, sensors or other sensing devices, etc. These computer devices may run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices may include cellular phones, smart phones, tablet computers, personal digital assistants (PDAs), etc. Wearable devices may include head-mounted displays (such as smart glasses) and other devices. Game systems may include various handheld game devices, Internet-enabled game devices, etc. Client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and may use various communication protocols.

[0031] The network 110 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including but not limited to TCP / IP, SNA, IPX, etc. By way of example only, the one or more networks 110 may be a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0032] Server 120 may include one or more general purpose computers, dedicated server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that may be virtualized to maintain a server's virtual storage device). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0033] The computing units in the server 120 may run one or more operating systems including any of the above operating systems and any commercially available server operating systems. The server 120 may also run any of a variety of additional server applications and / or middle-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0034] In some implementations, server 120 may include one or more applications to analyze and consolidate data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and / or 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and / or 106.

[0035] In some embodiments, the server 120 may be a server of a distributed system, or a server combined with a blockchain. The server 120 may also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in a cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services.

[0036] The system 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. The databases 130 may reside in various locations. For example, the database used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The databases 130 may be of different types. In some embodiments, the databases used by the server 120 may be, for example, relational databases. One or more of these databases may store, update, and retrieve data to and from the databases in response to commands.

[0037] In some embodiments, one or more of the databases 130 may also be used by applications to store application data. The databases used by the applications may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.

[0038] Figure 1The system 100 may be configured and operated in various ways to enable the application of various methods and apparatuses described in the present disclosure.

[0039] Figure 2 An exemplary process 200 of a data processing method according to an embodiment of the present disclosure is shown.

[0040] like Figure 2 As shown, in step S202, a plurality of original word-element vectors may be obtained, wherein each of the plurality of original word-element vectors represents one of text, image, and audio.

[0041] In step S204, the plurality of original word-unit vectors may be divided into a plurality of groups, and a first comprehensive vector corresponding to each group may be calculated using the plurality of original word-unit vectors included in each group of the plurality of groups.

[0042] In step S206, a plurality of first integrated vectors corresponding to a plurality of groups may be input into the large model to obtain a second integrated vector output by the large model. In the present disclosure, the large model is also referred to as a backbone large model.

[0043] In step S208, a plurality of target word element vectors may be reconstructed using the second comprehensive vector, wherein each of the plurality of target word element vectors represents one of text, image, and audio.

[0044] By using the embodiments of the present disclosure, after converting word units into corresponding vector representations, multiple word units are grouped together and a total vector representation corresponding to each group is calculated, thereby achieving sequence shortening at the model input level while retaining the original multimodal word unit information. At the output level, the second comprehensive vector output by the large model is reconstructed to obtain a complete output word unit sequence, thereby achieving a dual improvement in effect and efficiency.

[0045] The principles of the present disclosure will be described in detail below.

[0046] In the encoding and decoding process of multimodal data such as images and audio, residual quantization strategies represented by residual quantization variational autoencoder (RQ-VAE) and residual vector quantization (RVQ) are widely used for their excellent reconstruction effect. Such methods first divide the original multimodal data into multiple segments, then calculate a token sequence for each segment, and finally output a multimodal token matrix. However, the multimodal token matrix obtained by this method usually faces the problem of a large number of tokens, which brings great computational overhead to the training and reasoning of large models.

[0047] During the training and inference phases, large models (especially Transformer-based multimodal large models) usually receive a word sequence and learn to predict the next word in the form of autoregressive decoding. For raw data, current large models usually process it in two ways: one is to directly expand the word matrix into a one-dimensional word sequence to preserve the original information of the data as much as possible; the other is to shorten the sequence length by selectively using some words, thereby improving the data processing efficiency of the model.

[0048] In this context, whether directly expanding the word-unit matrix into a one-dimensional word-unit sequence or selecting some words to shorten the sequence length, both face their own limitations:

[0049] 1. Directly expanding the word unit into a one-dimensional word unit sequence can retain the original information of the data, but it will face the problem of a large number of word units after conversion. This problem is more serious in long sequence tasks.

[0050] 2. Although selective use of lemmas can shorten the sequence length, it may lose more details of the original data, making it difficult to answer some questions that are strongly related to data details. In particular, for multimodal generation tasks, the lack of details has a significant impact on the generation of multimodal data. Therefore, selective use of some multimodal lemmas is not a unified solution for multimodal understanding and generation tasks.

[0051] The data processing method provided by the present invention not only retains the original input information, but also shortens the length of data processed by the main large model, which can achieve a dual improvement in effect and efficiency.

[0052] In step S202, a plurality of original word-element vectors may be obtained by processing original data of text, image, audio and / or other modalities. In some examples, the original data may be encoded to obtain a word-element sequence. Each word-element in the word-element sequence may be a single numerical value. Furthermore, each word-element may be converted into a corresponding vector representation to obtain an original word-element vector.

[0053] For different modalities, different methods can be used to obtain the corresponding original word element vectors.

[0054] Figure 3 An exemplary process 300 of obtaining multiple original word-element vectors according to an embodiment of the present disclosure is shown. The process 300 can be used to at least partially implement step S202 in the process 200.

[0055] In step S302, the target image may be divided into a plurality of image blocks. The target image may be any image data to be processed. The present disclosure does not limit the specific manner of dividing the target image.

[0056] In step S304, multiple image blocks may be encoded to obtain a first preset number of image word units corresponding to each image block. The present disclosure does not limit the image encoding method and the specific value of the first preset number, and those skilled in the art may select and set it according to needs.

[0057] In step S306, each image word can be mapped to an original image word vector, and the multiple original word vectors include a first preset number of original image word vectors corresponding to each of the multiple image blocks. In some examples, a pre-trained image coding model (such as RQ-VAE or Vision Transformer) can be used to generate image word embeddings. This mapping process maps the high-dimensional image block feature vector to the corresponding low-dimensional original word vector by querying the word embedding table or codebook. In addition, the mapping can also be completed through linear transformation, neural network or other methods, which are not limited here.

[0058] In an exemplary embodiment, the RQ-VAE model can be used to complete the division and encoding of the target image. Taking a target image with a resolution of 1024×1024 as an example, the RQ-VAE model can compress it into a 32×32×4 word-meta matrix. That is, the target image is divided into 32×32 image blocks, and each image block is encoded into 4 image words. Furthermore, each image word-meta can be mapped to the corresponding original image word-meta vector.

[0059] In some examples, in step S204, among the multiple original word element vectors, the first preset number of original image word element vectors corresponding to the same image block can be divided into the same group. In this way, the grouping of these word elements is consistent with the semantic information of the target image, thereby improving the rationality of the grouping. This method ensures that the first comprehensive vector representing the image modality used to input the large model can be effectively associated with an actual area in the target image, enhances the semantic expression ability, and thus improves the overall effect and accuracy of the large model when processing multimodal data.

[0060] Figure 4 An exemplary process 400 of obtaining multiple original word-unit vectors according to an embodiment of the present disclosure is shown. The process 400 can be used to at least partially implement step S202 in the process 200.

[0061] In step S402, the target audio may be divided into a plurality of audio segments. The target audio may be any audio data to be processed. The present disclosure does not limit the specific manner of dividing the target audio.

[0062] In step S404, multiple audio segments may be encoded to obtain a second preset number of audio words corresponding to each audio segment. The present disclosure does not limit the audio encoding method and the specific value of the second preset number, and those skilled in the art may select and set them according to needs.

[0063] In step S406, each audio word can be mapped to an original audio word vector, and the multiple original word vectors include a second preset number of original audio word vectors corresponding to each of the multiple audio clips. In some examples, the audio word obtained after encoding the audio clip can be a feature representation obtained after time-frequency conversion (such as short-time Fourier transform (STFT) or Mel spectrum). In addition, pre-trained audio coding models such as Wav2Vec and Hubert can also be used. In order to map the word features of the audio clip to the original audio word vector, a neural network structure such as linear transformation or multi-layer perceptron (MLP) can be used to implement the above mapping process.

[0064] In an exemplary embodiment, each second of audio may be divided into 75 audio segments, and each audio segment may be encoded into 8 audio tokens.

[0065] In some examples, in step S204, among the multiple original word element vectors, a second preset number of original audio word element vectors corresponding to the same audio segment can be divided into the same group. In this way, the grouping of these words is consistent with the semantic information of the target audio, thereby improving the rationality of the grouping. This method ensures that the first comprehensive vector representing the audio modality used to input the large model can be effectively associated with an actual segment in the target audio, enhancing the semantic expression ability, thereby improving the overall effect and accuracy of the large model when processing multimodal data.

[0066] In some examples, multiple original word unit vectors may be divided into multiple groups in other ways, which are not limited here.

[0067] In addition to the methods provided in the above embodiments, the original content in the form of images, audio, text, etc. may also be processed in other ways to obtain the corresponding original word element vectors, which are not limited here.

[0068] In step S204, using the multiple original word-unit vectors included in each of the multiple groups to calculate the first comprehensive vector corresponding to the group may include calculating a total vector representation, i.e., the first comprehensive vector, for each group of original word-unit vectors by adding, concatenating, or using a neural network (e.g., an encoding network based on a Transformer structure). This process may also be referred to as word-unit folding.

[0069] By folding multi-word information, sequence shortening is achieved at the input level of the large model, which significantly reduces the length of sequences that the backbone large model needs to process, while retaining the word information of the original data (especially the original information of multimodal data).

[0070] In step S206, the large model may receive multiple first integrated vectors corresponding to multiple groups, and encode these first integrated vectors to output a second integrated vector. The second integrated vector may represent the next vector representation predicted by the large model in the multiple first integrated vectors. The vector representation may contain multiple word-unit information, and thus it may be used to reconstruct in step S208 to restore to multiple target word-unit vectors.

[0071] In some examples, the large model may be based on a Transformer structure. Step S206, inputting a plurality of first integrated vectors corresponding to a plurality of groups into the large model to obtain a second integrated vector output by the large model may include: using the large model to perform autoregressive decoding based on the plurality of first integrated vectors to obtain one or more second integrated vectors.

[0072] The Transformer model is a neural network model based on the attention mechanism, which is mainly used for natural language processing (NLP) tasks such as language translation and text classification. The Transformer structure is widely used in the field of natural language processing. The Transformer model mainly consists of two parts: the encoder and the decoder. The encoder converts the input sequence into a vector representation of a fixed length, while the decoder converts the vector into an output sequence. Between the encoder and the decoder, there are one or more attention mechanism layers to capture the correlation between the input sequence and the output sequence.

[0073] The input sequence and output sequence of the Transformer model are both composed of tokens or corresponding token vectors. Each token can correspond to a single word, character, word or other text unit in the text mode, a pixel, image block or other image unit in the image mode, an audio frame, audio clip or other audio unit in the audio mode, or a segment with specific semantics in other modes. In addition, tokens can also correspond to special symbols, such as classification tags [CLS], separators [SEP], mask tags [MASK], etc.

[0074] The Transformer model is mainly composed of multiple layers of TransformerLayer, each of which contains two main parts: Self-Attention Layer and Feed-Forward Network (FFN). The self-attention mechanism is a method of calculating the correlation between each word and other words and combining them. This method can capture the relationship between each word (or word element vector) in the input sequence, regardless of their distance in the sequence. In the Transformer model, the self-attention mechanism is used to calculate the influence of each word (or word element vector) on other words (or word element vector) in the input sequence or output sequence. Feedforward neural network is a neural network based on fully connected layers, which is used to perform nonlinear transformations on input sequences. In the Transformer model, FFN can be composed of two fully connected layers.

[0075] By using autoregressive decoding, the large model can continuously iterate and predict the subsequent multimodal information output until the large model determines that the output is complete. By making the large model output one or more second comprehensive vectors, and then reconstructing the second comprehensive vectors to obtain the target word element vector, the length of data processed by the large model is shortened, which can achieve a dual improvement in effect and efficiency.

[0076] In some examples, step S208, reconstructing multiple target word-unit vectors using the second comprehensive vector may include: using a word-unit reconstruction unit based on a Transformer structure to perform autoregressive decoding based on the second comprehensive vector to obtain multiple target word-unit vectors.

[0077] The word unit reconstruction unit based on the Transformer structure has a strong semantic learning ability. Through the autoregressive decoding method, the model can gradually strengthen the semantic expression during the generation process, so that the generated target word unit vector has a high semantic consistency and coherence. This has a significant effect on improving the performance and understanding ability of the multimodal model.

[0078] In one example, the model size of the word-gram reconstruction unit can be smaller than the large model.

[0079] Figure 5 A schematic diagram of word element vector folding, processing and restoration according to an embodiment of the present disclosure is shown.

[0080] During the word-meta folding 502, operations such as division, encoding and / or mapping can be performed on the original data of a certain modality (e.g., image, audio, text, etc.) to obtain four original word-meta vectors A, B, C, and D corresponding to the same semantic segment (e.g., image block, audio segment, text segment, etc.). These four word-meta vectors can be folded into a first comprehensive vector 508 by addition, concatenation, or neural network. The first comprehensive vector 508 can be processed by the backbone network 504 to obtain a second comprehensive vector 510. The second comprehensive vector 510 can be subjected to word-meta restoration 506 to obtain a target word-meta vector E.

[0081] The second integrated vector 510 can be regarded as obtained by performing word folding 502 on the target word element vector E. Therefore, if the backbone network 504 is required to perform autoregressive decoding, the first integrated vector 508 (obtained by performing word folding on the original word element vectors A, B, C, and D) and the second integrated vector 510 (corresponding to the target word element vector E) can be re-input into the backbone network 504 to obtain the next second integrated vector 512. The second integrated vector 512 can be subjected to word restoration 506 to obtain the target word element vectors F, G, and H in sequence.

[0082] In one example, the first integrated vector 508, the second integrated vector 510, and the second integrated vector 512 can be re-input into the backbone network 504. After completing two rounds of autoregressive decoding, the second integrated vector 514 and the corresponding target word element vectors I and J, as well as the second integrated vector 516 and the corresponding target word element vectors K, L, M, and N are obtained.

[0083] In some examples, the multiple original word-element vectors may have a preset vector length. Step S208, reconstructing multiple target word-element vectors using the second comprehensive vector includes: segmenting the second comprehensive vector based on the preset vector length to obtain multiple target word-element vectors.

[0084] By presetting the vector length for segmentation, the complexity of processing the comprehensive vector is reduced, and the high computational overhead of step-by-step generation or reconstruction is avoided. Segmentation can be used as a direct operation to quickly divide the second comprehensive vector into multiple target word element vectors, thereby improving the overall processing speed.

[0085] Figure 6 FIG. 6 shows an exemplary block diagram of a data processing device according to an embodiment of the present disclosure. As shown in FIG. 6 , the data processing device 600 may include an acquisition unit 610, a grouping unit 620, a large model 630, and a reconstruction unit 640. Figure 6 The data processing device described in the embodiment performs the combination Figure 2 Describe the data processing methods.

[0086] The acquisition unit 610 may be configured to acquire a plurality of original word-element vectors, wherein each of the plurality of original word-element vectors represents one of text, image, and audio.

[0087] The grouping unit 620 may be configured to divide the plurality of original word-unit vectors into a plurality of groups, and calculate a first comprehensive vector corresponding to each group using the plurality of original word-unit vectors included in each of the plurality of groups.

[0088] The large model 630 may be configured to receive a plurality of first synthesis vectors corresponding to a plurality of groups to output a second synthesis vector.

[0089] The reconstruction unit 640 may be configured to reconstruct a plurality of target word element vectors using the second comprehensive vector, wherein each of the plurality of target word element vectors represents one of text, image, and audio.

[0090] In some embodiments, the acquisition unit may include: a blocking subunit, configured to divide the target image into multiple image blocks; a first encoding subunit, configured to encode the multiple image blocks to obtain a first preset number of image word units corresponding to each image block; and a first mapping subunit, in a step of mapping each image word unit to an original image word unit vector, the multiple original word unit vectors including the first preset number of original image word unit vectors corresponding to each of the multiple image blocks.

[0091] In some embodiments, among the plurality of original word-unit vectors, a first preset number of original image word-unit vectors corresponding to the same image block may be divided into the same group.

[0092] In some embodiments, the acquisition unit may include: a segmentation subunit, configured to divide the target audio into multiple audio segments; a second encoding subunit, configured to encode the multiple audio segments to obtain a second preset number of audio words corresponding to each audio segment; and a second mapping subunit, configured to map each audio word into an original audio word vector, the multiple original word vectors including a second preset number of original audio word vectors corresponding to each of the multiple audio segments.

[0093] In some embodiments, among the plurality of original word unit vectors, a second preset number of original audio word unit vectors corresponding to the same audio segment may be divided into the same group.

[0094] In some embodiments, the reconstruction unit may be configured to perform autoregressive decoding based on the second comprehensive vector using a word unit reconstruction unit based on a Transformer structure to obtain a plurality of target word unit vectors.

[0095] In some embodiments, the multiple original word-unit vectors may have a preset vector length, and the reconstruction unit may be configured to segment the second comprehensive vector based on the preset vector length to obtain multiple target word-unit vectors.

[0096] In some embodiments, the large model may be based on a Transformer structure, and may be configured to perform autoregressive decoding based on a plurality of first synthesis vectors to obtain one or more second synthesis vectors.

[0097] It should be understood that Figure 6 The various modules or units of the apparatus 600 shown in FIG. 6 can be used in conjunction with the reference Figure 2 The steps in the method 200 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 200 are also applicable to the device 600 and the modules and units included therein. For the sake of brevity, some operations, features and advantages are not repeated here.

[0098] Although specific functionality is discussed above with reference to specific modules, it should be noted that the functionality of the various units discussed herein may be separated into multiple units, and / or at least some functionality of multiple units may be combined into a single unit.

[0099] It should also be understood that various techniques may be described herein in the general context of software hardware elements or program modules. Figure 6 The various units described can be implemented in hardware or in hardware in combination with software and / or firmware. For example, these units can be implemented as computer program codes / instructions, which are configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, these units can be implemented as hardware logic / circuits. For example, in some embodiments, one or more of units 610 to 640 can be implemented together in a system on chip (System on Chip, SoC). SoC may include an integrated circuit chip (which includes a processor (e.g., a central processing unit (CPU), a microcontroller, a microprocessor, a digital signal processor (DSP), etc.), a memory, one or more communication interfaces, and / or one or more components in other circuits), and may optionally execute the received program code and / or include embedded firmware to perform functions.

[0100] According to another aspect of the present disclosure, an electronic device is also provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a method according to an embodiment of the present disclosure.

[0101] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is further provided, where the computer instructions are used to cause the computer to execute the method according to the embodiment of the present disclosure.

[0102] According to another aspect of the present disclosure, a computer program product is further provided, including a computer program, wherein the computer program implements the method according to the embodiment of the present disclosure when being executed by a processor.

[0103] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0104] According to an embodiment of the present disclosure, an electronic device, a readable storage medium and a computer program product are also provided.

[0105] refer to Figure 7 , a block diagram of an electronic device 700 that can be used as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0106] like Figure 7 As shown, the electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the electronic device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0107] Multiple components in the electronic device 700 are connected to the I / O interface 705, including: an input unit 706, an output unit 707, a storage unit 708, and a communication unit 709. The input unit 706 can be any type of device that can input information to the electronic device 700. The input unit 706 can receive input digital or character information, and generate key signal input related to user settings and / or function control of the electronic device, and can include but is not limited to a mouse, a keyboard, a touch screen, a trackpad, a trackball, a joystick, a microphone and / or a remote control. The output unit 707 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator and / or a printer. The storage unit 708 can include but is not limited to a disk, an optical disk. The communication unit 709 allows the electronic device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver and / or a chipset, such as Bluetooth TM devices, 802.11 devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0108] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as processes 200, 300, and 400. For example, in some embodiments, processes 200, 300, and 400 may be implemented as computer software programs that are tangibly included in machine-readable media, such as storage units 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of processes 200, 300, and 400 described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the processes 200 , 300 , 400 in any other appropriate manner (eg, by means of firmware).

[0109] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0110] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0111] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0112] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0113] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0114] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0115] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0116] Although the embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above-mentioned methods, systems and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but only by the claims after authorization and their equivalent scope. Various elements in the embodiments or examples can be omitted or replaced by their equivalent elements. In addition, each step can be performed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples can be combined in various ways. It is important that with the evolution of technology, many elements described herein can be replaced by equivalent elements that appear after the present disclosure.

Claims

1. A data processing method based on a large model, comprising: Obtain a plurality of original word element vectors, wherein each of the plurality of original word element vectors represents one of text, image, and audio; Dividing the plurality of original word element vectors into a plurality of groups, and calculating a first comprehensive vector corresponding to each group using the plurality of original word element vectors included in each group of the plurality of groups; Inputting a plurality of first integrated vectors corresponding to the plurality of groups into a large model to obtain a second integrated vector output by the large model; as well as The second comprehensive vector is used to reconstruct a plurality of target word element vectors, wherein each of the plurality of target word element vectors represents one of text, image and audio.

2. The method according to claim 1, wherein: Obtaining multiple original word meta-vectors includes: Divide the target image into multiple image blocks; Encoding the plurality of image blocks to obtain a first preset number of image word units corresponding to each image block; and Each image word-element is mapped into an original image word-element vector, wherein the plurality of original word-element vectors include the first preset number of original image word-element vectors corresponding to each of the plurality of image blocks.

3. The method according to claim 2, wherein: Among the plurality of original word-unit vectors, the first preset number of original image word-unit vectors corresponding to the same image block are divided into the same group.

4. The method according to claim 1, wherein: Obtaining multiple original word meta-vectors includes: Divide the target audio into multiple audio segments; Encoding the multiple audio segments to obtain a second preset number of audio words corresponding to each audio segment; and Each audio word is mapped into an original audio word vector, wherein the multiple original word vectors include the second preset number of original audio word vectors corresponding to each of the multiple audio segments.

5. The method according to claim 4, wherein: Among the multiple original word-unit vectors, the second preset number of original audio word-unit vectors corresponding to the same audio segment are divided into the same group.

6. The method according to any one of claims 1 to 5, wherein: Reconstructing a plurality of target word element vectors using the second comprehensive vector includes: A word unit reconstruction unit based on a Transformer structure is used to perform autoregressive decoding based on the second comprehensive vector to obtain the multiple target word unit vectors.

7. The method according to any one of claims 1 to 5, wherein: The multiple original word element vectors have a preset vector length, and the multiple target word element vectors are reconstructed by using the second comprehensive vector, including: The second comprehensive vector is segmented based on the preset vector length to obtain the multiple target word element vectors.

8. The method according to any one of claims 1 to 5, wherein: The large model is based on a Transformer structure, and the plurality of first comprehensive vectors corresponding to the plurality of groups are input into the large model to obtain a second comprehensive vector output by the large model, including: The large model is used to perform autoregressive decoding based on the plurality of first integrated vectors to obtain one or more second integrated vectors.

9. A data processing device based on a large model, comprising: An acquisition unit is configured to acquire a plurality of original word element vectors, wherein each of the plurality of original word element vectors represents one of text, image, and audio; A grouping unit, configured to divide the plurality of original word element vectors into a plurality of groups, and calculate a first comprehensive vector corresponding to each group using the plurality of original word element vectors included in each of the plurality of groups; A large model configured to receive a plurality of first integrated vectors corresponding to the plurality of groups to output a second integrated vector; as well as The reconstruction unit is configured to reconstruct a plurality of target word element vectors using the second comprehensive vector, wherein each of the plurality of target word element vectors represents one of text, image and audio.

10. The device according to claim 9, wherein: The acquisition unit comprises: A block division subunit is configured to divide the target image into a plurality of image blocks; A first encoding subunit is configured to encode the plurality of image blocks to obtain a first preset number of image word units corresponding to each image block; and The first mapping subunit maps each image word unit into an original image word unit vector in step, wherein the plurality of original word unit vectors include the first preset number of original image word unit vectors corresponding to each of the plurality of image blocks.

11. The device according to claim 10, wherein: Among the plurality of original word-unit vectors, the first preset number of original image word-unit vectors corresponding to the same image block are divided into the same group.

12. The device according to claim 9, wherein: The acquisition unit comprises: A segmentation subunit, configured to divide the target audio into a plurality of audio segments; A second encoding subunit is configured to encode the multiple audio segments to obtain a second preset number of audio words corresponding to each audio segment; and The second mapping subunit is configured to map each audio word into an original audio word vector, wherein the multiple original word vectors include the second preset number of original audio word vectors corresponding to each of the multiple audio clips.

13. The device according to claim 12, wherein: Among the multiple original word-unit vectors, the second preset number of original audio word-unit vectors corresponding to the same audio segment are divided into the same group.

14. The device according to any one of claims 9 to 13, wherein: The reconstruction unit is configured to perform autoregressive decoding based on the second comprehensive vector using a word unit reconstruction unit based on a Transformer structure to obtain the multiple target word unit vectors.

15. The device according to any one of claims 9 to 13, wherein: The multiple original word-element vectors have a preset vector length, and the reconstruction unit is configured to segment the second comprehensive vector based on the preset vector length to obtain the multiple target word-element vectors.

16. The device according to any one of claims 9 to 13, wherein: The large model is based on a Transformer structure and is configured to perform autoregressive decoding based on the multiple first synthesis vectors to obtain one or more second synthesis vectors.

17. An electronic device comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-8.

19. A computer program product comprising a computer program, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images