A data processing method, system, device and computer readable storage medium

Abstract

The application discloses a data processing method, system, device and computer readable storage medium, relates to the technical field of data processing, is applied to a data processing device, loads a webpage user interface of a browser; obtain the data to be processed based on the webpage user interface; the data to be processed is sent to the target computing power stick; obtain the processing result of the target computing power stick to the data to be processed;Wherein, the target computing power stick is connected with the data processing device through the USB interface;The processing result of the data to be processed includes the result obtained after the target computing power stick calls the target neural network model to process the data to be processed.The computing power stick can obtain the processing result of the data to be processed based on the model inference ability, realizes that the data processing device is not upgraded or replaced without the condition, provides the model inference ability for the data processing device through connecting the computing power stick, conveniently improves the data interaction ability of the data processing device, and is plug and play, easy to operate and expand.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a data processing method, system, device, and computer-readable storage medium. Background Technology

[0002] Currently, users can use data processing devices for data processing, such as servers, PCs (Personal Computers), smart wearable devices, testing equipment, and in-vehicle control systems. However, due to limitations in the performance and hardware of these devices, the ways to process data using them are limited and cannot meet user needs. Therefore, it is necessary to upgrade or replace these data processing devices.

[0003] Upgrading or replacing data processing equipment would increase costs, and the upgraded or replaced equipment might still fail to meet user needs, resulting in a poor user experience.

[0004] In conclusion, how to conveniently improve the data processing capabilities of data processing equipment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a data processing method that can, to a certain extent, solve the technical problem of how to conveniently improve the data processing capabilities of data processing equipment. This application also provides a data processing system, an electronic device, and a computer-readable storage medium.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] A data processing method, applied to a data processing device, includes:

[0008] Load the browser's webpage user interface;

[0009] Data to be processed is obtained based on the webpage user interface.

[0010] The data to be processed is sent to the target computing power bar;

[0011] Obtain the processing result of the target computing power stick on the data to be processed;

[0012] The target computing stick is connected to the data processing device via a USB interface; the processing result of the data to be processed includes the result obtained after the target computing stick calls the target neural network model to process the data to be processed.

[0013] In an exemplary embodiment, it also includes:

[0014] Determine the user terminal connected to the data processing device;

[0015] Collect computing power resource information, random access memory information, and read-only memory information of the user terminal;

[0016] Based on the computing power resource information, the random access memory information, and the read-only memory information, a neural network model is deployed in the user terminal to configure the user terminal as a computing power stick.

[0017] In an exemplary embodiment, after obtaining the processing result of the target computing power stick on the data to be processed, the method further includes:

[0018] The processing results of the data to be processed are used to extract instructions;

[0019] If a data acquisition instruction is extracted from the processing result of the data to be processed, then data is collected according to the data acquisition instruction to obtain the associated processing data;

[0020] The associated processing data is transmitted to the target computing power bar;

[0021] Receive the processing results of the target computing power stick on the associated processing data.

[0022] In an exemplary embodiment, before obtaining the data to be processed based on the web page user interface, the method further includes:

[0023] The target computing power stick supports the model in the web user interface according to the set method;

[0024] Obtain the selection instructions transmitted by the user through the webpage user interface;

[0025] Based on the user's selection command, the model displayed on the webpage user interface is selected to obtain model selection information;

[0026] The model selection information is sent to the target computing power stick so that the target computing power stick determines the target neural network model according to the model selection information.

[0027] In an exemplary embodiment, before loading the browser's webpage user interface, the method further includes:

[0028] Obtain the intranet interface configuration information of the target computing stick based on the USB virtual network card protocol;

[0029] The USB Ethernet is detected, and in response to the detection of USB Ethernet, a network interface is created according to the intranet interface configuration information.

[0030] Based on the IP address agreed upon with the target computing power stick, a connectivity verification is performed on the target computing power stick;

[0031] If the connectivity verification is successful, the browser accesses the IP address of the target computing stick to obtain the device information and certificate fingerprint returned by the target computing stick, so as to establish a communication link with the target computing stick.

[0032] In an exemplary embodiment, sending the data to be processed to the target computing power stick includes:

[0033] The data to be processed is encrypted to obtain encrypted data to be processed;

[0034] The encrypted data to be processed is sent to the target computing power stick, so that the target computing power stick can decrypt the encrypted data to be processed to obtain the data to be processed.

[0035] A data processing method applied to a computing power stick includes:

[0036] The system acquires data to be processed sent by a data processing device, the data to be processed including data acquired by the data processing device based on the web page user interface of a loaded browser;

[0037] The target neural network model is invoked to process the data to be processed, and the processing result of the data to be processed is obtained;

[0038] Send the processing result of the data to be processed to the data processing device;

[0039] The computing power stick is connected to the data processing device via a USB interface.

[0040] A data processing system, applied to a data processing device, comprising:

[0041] The loading module is used to load the browser's webpage user interface;

[0042] The data acquisition module is used to acquire data to be processed based on the web page user interface.

[0043] The data to be processed sending module is used to send the data to be processed to the target computing power stick;

[0044] The processing result acquisition module is used to acquire the processing result of the target computing power stick on the data to be processed;

[0045] The target computing stick is connected to the data processing device via a USB interface; the processing result of the data to be processed includes the result obtained after the target computing stick calls the target neural network model to process the data to be processed.

[0046] An electronic device, comprising:

[0047] Memory, used to store computer programs;

[0048] A processor for implementing the steps of any of the above-described data processing methods when executing the computer program.

[0049] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the data processing methods described above.

[0050] This application provides a data processing method applied to a data processing device, which includes loading a browser's webpage user interface; acquiring data to be processed based on the webpage user interface; sending the data to be processed to a target computing stick; and acquiring the processing result of the target computing stick on the data to be processed. The target computing stick is connected to the data processing device via a USB interface. The processing result of the data to be processed includes the result obtained after the target computing stick calls a target neural network model to process the data. In this application, a data processing device is connected to a computing stick via a USB interface, enabling plug-and-play functionality. The computing stick contains a neural network model. Because the neural network model has inference capabilities, after the data processing device sends the data to be processed to the computing stick, the computing stick can obtain the processing result based on the model's inference capabilities. This allows the data processing device to be provided with model inference capabilities without requiring upgrades or replacements, conveniently improving its data processing capacity. Furthermore, interaction with the computing stick is achieved solely through a web-based user interface, making it easy for users to operate. By simply expanding the neural network model in the computing stick, its model inference capabilities can be extended, thereby expanding the data processing capacity of the data processing device, demonstrating good applicability. This application also provides a data processing system, electronic device, and computer-readable storage medium that solves the corresponding technical problems. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0052] Figure 1 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0053] Figure 2 This is a system architecture diagram based on a web-based user interface;

[0054] Figure 3 This is a flowchart for data processing based on mobile phones;

[0055] Figure 4 System architecture diagrams for PC and mobile devices;

[0056] Figure 5 This is a flowchart for data processing based on associated data;

[0057] Figure 6 Another flowchart of a data processing method provided in an embodiment of this application;

[0058] Figure 7 This is a schematic diagram of the structure of a data processing system provided in an embodiment of this application;

[0059] Figure 8 This is a structural diagram of a computer device 20 according to an exemplary embodiment. Detailed Implementation

[0060] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] Please see Figure 1 , Figure 1 This is a flowchart of a data processing method provided in an embodiment of this application.

[0062] This application provides a data processing method applied to a data processing device, which may include the following steps:

[0063] Step S101: Load the browser's webpage user interface.

[0064] Step S102: Obtain the data to be processed based on the web user interface.

[0065] In practical applications, data processing devices can interact with computing power sticks via voice wake-up. For example, after receiving a user's wake-up command, the data processing device continues to receive the user's data processing commands. If the data processing command indicates interaction with the computing power stick, the device parses the data to be processed from the command and sends it to the target computing power stick for processing. During this process, the user can wake up the data processing device to invoke the computing power stick, which is simple and convenient, but may have a weak user experience and be difficult for the user to understand. Therefore, a web user interface (Web UI) can be set in the browser of the data processing device, such as... Figure 2 As shown, this allows users to access the computing power unit through a web-based user interface. For example, the user's voice can be captured and parsed to obtain data to be processed. Alternatively, data can be retrieved through input boxes on the web interface. Correspondingly, the data processing device can load a browser's web interface and retrieve the data to be processed based on it. The type of data to be processed can be flexibly determined according to the application scenario; for example, it can be user-inputted voice data, user-inputted image data, test data, device logs, text, etc.

[0066] In an exemplary embodiment, the first level of the Web UI design may include index.html (main entry point), styles, js, and components; styles may include base.css (global styles) and components.css (model interface styles); js may include app.js (application initialization), api.js (REST API encapsulation), ws.js (WebSocket communication), models.js (model management), chat.js (LLM dialogue), cv.js (CV control), asr.js (ASR call), and tts.js (TTS playback); components may include chat-panel.html (LLM interactive interface), cv-panel.html (CV control interface), asr-panel.html (speech recognition interface), tts-panel.html (speech synthesis interface), and doc-panel.html (document processing interface). This achieves front-end and back-end separation, where the data processing device's browser only loads HTML and JS, and the computing power stick Web Server only provides static files and inference APIs. The API design can be as shown in Table 1.

[0067] Table 1 API Design

[0068]

[0069] In an exemplary embodiment, the computing power stick can automatically select the neural network model required for data processing. However, considering the diverse types of neural network models in the computing power stick, to facilitate the selection of the neural network model required by the user for data processing, before acquiring the data to be processed based on the web user interface, the data processing device can also display the models supported by the target computing power stick in the web user interface according to a set method for the user to select. The device then acquires the selection command transmitted by the user through the web user interface; this user selection command can be a selection command input by the user's voice or a command generated after the user clicks on a model in the web user interface. Based on the user selection command, the device selects the model displayed in the web user interface to obtain model selection information. The model selection information is then sent to the target computing power stick so that the target computing power stick determines the target neural network model according to the model selection information. During this process, considering that the web user interface is loaded by a browser, the data processing device can interact with the computing power stick via HTTP POST or WebSocket.

[0070] As can be seen from this implementation process, before the data processing device sends the data to be processed to the target computing power stick connected to the data processing device, it can also display the models supported by the target computing power stick in a set manner in the web user interface; obtain the selection instructions transmitted by the user through the web user interface; select the model displayed in the web user interface based on the user's selection instructions to obtain model selection information; and send the model selection information to the target computing power stick so that the target computing power stick determines the target neural network model according to the model selection information. This realizes the ability for users to perceive and call the computing power stick through the data processing device by using a web user interface set in the browser, thus improving the user experience.

[0071] Step S103: Send the data to be processed to the target computing power bar.

[0072] Step S104: Obtain the processing result of the target computing power stick on the data to be processed.

[0073] In practical applications, after acquiring the data to be processed, the data processing device can send the data to the target computing stick connected to it. The computing stick refers to a portable AI (Artificial Intelligence) acceleration device used to accelerate deep learning inference tasks. Since the computing stick is equipped with a neural network model, the target computing stick can call the target neural network model to process the data to be processed and obtain the processing result. The data processing device can then obtain the processing result and apply it, such as displaying the processing result to the user or controlling the data processing device itself based on the processing result.

[0074] Understandably, the computing resources and supported neural network models in the computing stick can be flexibly determined according to the application scenario. For example, the computing stick can have built-in computing resources such as CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), and NPU (Neural Processing Unit), and can support neural network models such as LLM (Large Language Model), CV (Computer Vision), ASR (Automatic Speech Recognition), TTS (Text To Speech), and VLM (Vision-Language Model).

[0075] It should be noted that the data processing device can connect to multiple computing sticks or only a single computing stick. When multiple computing sticks are connected, they can form a distributed inference cluster, which can improve the overall computing power of the system and can be scaled up to high computing power scenarios. In this case, the target computing stick can be flexibly selected according to the load and data processing capabilities of each computing stick, which will not be elaborated here.

[0076] It should also be noted that the configuration of the data processing device and computing stick can be flexibly determined according to the application scenario. For example, the computing stick can run Linux and support RNDIS (Remote Network Driver Interface Specification), USB Gadget (USB Composite Device Framework) / CDC ECM (Communications DeviceClass – Ethernet Control Model), while the data processing device can run Windows / Linux / macOS / Android systems, etc. In terms of security, the data processing device and computing stick can be configured with certificates or pre-shared keys and can perform time synchronization. Regarding storage, the computing stick can have a model storage directory / opt / models / and a persistent log directory / var / log / compute-stick, etc.

[0077] In an exemplary embodiment, the data processing device can connect to the computing stick via Bluetooth, WiFi (Wireless Fidelity) network, or via USB (Universal Serial Bus) interface. This application does not specifically limit the connection method between the data processing device and the computing stick.

[0078] In specific application scenarios, most data processing devices lack computing power resources. Users need to install drivers, SDKs (Software Development Kits), and local inference environments to enhance the model inference capabilities of the data processing devices. This operation is complex and difficult for users to implement. To avoid this situation, the data processing device can pair with the computing stick via a USB interface and establish network communication to achieve plug-and-play functionality, eliminating complex network configuration. That is, before loading the web interface of the browser, the data processing device can obtain the configuration information of the target computing stick based on the USB virtual network card protocol. For example, the target computing stick can start USB RNDIS or the CDC ECM mode of USB Gadget and configure the internal network interface. The target computing stick is connected to the USB interface of the data processing device, that is, after the computing stick is inserted into the USB port of the data processing device, the computing stick starts the USB Gadget (CDC ECM) and configures the internal network interface (e.g., usb0). Example commands (Linux side) can be modprobe g_ether or create a CDC via the gadgetfs script. ECM; The data processing device detects the USB Ethernet. If the USB Ethernet is detected, it creates a usb0 network interface according to the internal network interface configuration information. During this process, the browser app can listen for the link-up (connection establishment) event in the ConnectivityManager. An IP address is agreed upon with the target computing stick. For example, the data processing device uses DHCP (Dynamic Host Configuration Protocol) with the server: the data processing device's IP is 192.168.7.1, and the computing stick obtains 192.168.7.2 via DHCP. Alternatively, both the data processing device and the computing stick agree on a static IP address, such as the data processing device being 192.168.7.1 and the compute being 192.168.7.2. Based on the IP address, connectivity verification is performed on the target computing stick. For example, the data processing device initiates a ping to 192.168.7.2, and the computing stick checks in / sbin / ifconfig usb0. If it fails, the USBlink / dmesg log is checked to confirm kernel module loading. In response to successful connectivity verification, the target computing stick's IP address is accessed through the browser. For example, the Control Agent calls a GET request. http: / / 192.168.7.2:8080 / api / v1 / hello, retrieves the device information and certificate fingerprint returned by the target computing stick to establish a communication link with it. During this process, the data processing device can also use mTLS (mutually linked TLS) to establish a communication link with the computing stick to ensure the security of the communication link. If mTLS is used, the certificate verification process follows the flow: certificate exchange → mutual verification → establishment of TLS session.

[0079] This application provides a data processing method applied to a data processing device, which includes loading a browser's webpage user interface; acquiring data to be processed based on the webpage user interface; sending the data to be processed to a target computing stick; and acquiring the processing result of the target computing stick on the data to be processed. The target computing stick is connected to the data processing device via a USB interface. The processing result of the data to be processed includes the result obtained after the target computing stick calls a target neural network model to process the data. In this application, the data processing device is connected to a computing stick via a USB interface, enabling plug-and-play functionality. The computing stick is equipped with a neural network model. Since the neural network model has inference capabilities, after the data processing device sends the data to be processed to the computing stick, the computing stick can obtain the processing result of the data based on the model's inference capabilities. This allows the data processing device to be provided with model inference capabilities by connecting to the computing stick without upgrading or replacing the data processing device. This conveniently improves the data processing capabilities of the data processing device. Furthermore, the interaction with the computing stick can be achieved simply through a loaded browser's web interface, making it easy for users to operate. By simply expanding the neural network model in the computing stick, the model inference capabilities of the computing stick can be expanded, thereby expanding the data processing capabilities of the data processing device, demonstrating good applicability.

[0080] Based on the above embodiments, the computing power stick in this application can be a pre-configured computing power stick, or it can be a computing power stick obtained by configuring the user terminal according to the configuration requirements of the computing power stick. The user terminal can be a user device, computer, etc., to utilize the computing power of the user terminal and reduce user costs. Please refer to [link to relevant documentation]. Figure 3 The data processing method provided in this application embodiment, applied to a data processing device, may include the following steps:

[0081] Step S301: Determine the user terminal connected to the data processing device.

[0082] Step S302: Collect computing power resource information, random access memory information and read-only memory information of user terminals.

[0083] Step S303: Deploy a neural network model in the user terminal according to the computing power resource information, random access memory information and read-only memory information, so as to configure the user terminal as a computing power stick.

[0084] In the exemplary embodiment, since the deployment of the neural network model requires computing resources and memory, it is necessary to collect the computing resource information, random access memory (RAM) information, and read-only memory (ROM) information of the user terminal. Then, based on the computing resource information, RAM information, and ROM information, the neural network model is deployed in the user terminal to configure the user terminal as a computing power stick. During this process, the data processing device can network with the user terminal through VPNService, which is not specifically limited here. At this time, the system architecture of the data processing device and the computing power stick is as follows: Figure 4 As shown, the architecture can be as follows:

[0085] flowchart LR;

[0086] A [PC Web UI] --> HTTP Request | B [Mobile App Web Server];

[0087] B --> |Model API Call| C [Model Manager];

[0088] C -->|Calling computing power| D[CPU];

[0089] C --> | Call computing power | E[GPU];

[0090] C -->|Call Computing Power| F[DSP / NPU];

[0091] C-->G [Model Shop: Download / Hot Switch];

[0092] D-->H[LLM model];

[0093] E-->I[CV / VLM model];

[0094] F-->J[ASR / TTS model];

[0095] H-->K[Reasoning Result];

[0096] I-->K;

[0097] J-->K;

[0098] K --> |HTTP Response| A.

[0099] In specific application scenarios, a model store interface can be deployed on the user terminal, allowing users to download models such as LLM, CV, ASR, and TTS. Computing resources can be controlled, supporting CPU for lightweight model inference and low-power mode, GPU for large model FP16 / INT8 inference acceleration, and DSP / NPU for LLM / ASR / TTS voice acceleration. Furthermore, the user terminal can be configured to provide a RESTful API to the WebUI, enabling the data processing device to recommend suitable models to the user based on the acquired CPU, GPU, DSP, RAM, and ROM information, combined with computing power information, ensuring that the model can perform inference on the user terminal side.

[0100] For ease of understanding, assuming that the ROM size is determined based on the model file (weights) and system / temporary space (>=1.5×weights) and additional reservation (≥4GiB), then based on common device tiers (RAM / ROM), we provide recommended LLM / CV / ASR / TTS deployment methods, ROM storage requirements, and concurrency / context suggestions.

[0101] Tier A – 8 GiB RAM / 32 GiB ROM; Recommended LLM: Qwen3-0.6B (INT8 or 4-bit), estimated memory ≈ 1.4–1.8 GiB (very ample), ROM: weights (INT8) ≈ 0.559 GiB → recommended minimum ROM 4–8 GiB, usable for model + caching. Recommended CV / ASR / TTS: YOLOv8n / Whisper-tiny / Tiny VITS (quantization), context & concurrency recommendation max_input_tokens ≤ 1024, concurrency 1 (or 2 concurrency is very limited), deployment backend: ONNX Runtime Mobile / GGML (CPU) or NNAPI (if NPU supported). Expected experience: LLM query response < 0.3s (small Qwen3), CV inference image 50–300ms. It should be noted that 7B is not feasible in this tier (unless extremely aggressive swap / offload is performed, resulting in a poor experience).

[0102] Tier B—12–16 GiB RAM / 64 GiB ROM; Recommended LLM: LLaMA-like 7B (INT8) (preferred); Qwen3-0.6B / higher can also be deployed while retaining concurrency capabilities. 7B INT8 estimated ≈ 10.8 GiB (Note: This assumes an efficient INT8 kernel and available memory ≈ 12–16 GiB); ROM: weights (INT8) ≈ 6.519 GiB → Recommended available ROM ≥ 16–24 GiB (including model + swap + cache). Recommended CV / ASR / TTS: YOLOv8s, Whisper-base (quantization), VITSsmall (quantization). Context & Concurrency: max_input_tokens recommended ≤2048; recommended concurrency: 1 master + a small number of background tasks; backend deployment: ONNX Runtime with vendor INT8 kernel (if QCS8550 supports Hexagon / NNAPI-optimized ONNX), or TensorRT-like vendor runtime (if available). Expected Experience: Theoretically, 7B single-time 100-token computation time is ~0.09s, actual time including I / O is approximately 0.2–0.5s (depending on runtime). Note that for stable 7B operation, the system must reserve ≥11GiB for processes (the system / OS must have sufficient reserves).

[0103] Tier C—24GiB RAM / 128GiB ROM; Recommended LLM: 13B (4-bit optimized) or 7B (higher concurrency); 13B 4-bit total estimated ≈ 10.08GiB (suitable for 24GiB), 13B INT8 estimated ≈ 19.16GiB (feasible, but close to the limit), ROM: 13B weights (INT8) ≈ 12.107GiB → Recommended ROM ≥ 32–40GiB. Recommended CV / ASR / TTS: YOLOv8m, Whisper-base, VITS medium (quantization), Context & Concurrency: Supports max_input_tokens 2048–4096; Concurrency 2 (depending on context / KV), Deployment Backend: ONNX Runtime + vendor EP or allocate dedicated GPU / Hexagon NPU kernels (preferably using INT8 / 4-bit). Expected performance: 13B on 48T with 100 tokens theoretically ~0.16s, actual ~0.4–1.0s; depending on quantization efficiency and IO impact. Note that this level allows for more RAG / vector retrieval / local indexing.

[0104] Compared with existing computing power sticks, the effects of this embodiment are shown in Table 2. It reduces the cost of purchasing computing power sticks, reuses the computing power of user terminals, and the model is scalable and easy to deploy.

[0105] Table 2 Comparison of Results

[0106]

[0107] Step S304: Load the browser's web page user interface and obtain the data to be processed based on the web page user interface.

[0108] Step S305: Send the data to be processed to the target computing power bar.

[0109] Step S306: Obtain the processing result of the target computing power stick on the data to be processed.

[0110] To facilitate understanding the interaction between data processing devices and user terminals, let's assume the mobile device deploys Qwen 7B + Whisper ASR + CosyVoice TTS. This enables an AI PC document assistant: the user uploads a document on the PC, the mobile device parses and summarizes it, and the PC generates a voice reading based on the mobile device's results. Alternatively, let's assume the mobile device deploys Qwen-VL + YOLOv8. This enables AI multimodal inference: the user uploads an image on the PC, the mobile device performs object detection and question answering, and the PC's web UI displays the annotation results and natural language descriptions.

[0111] Based on the above embodiments, the computing power stick may also acquire data using data processing devices for further data processing. To meet this application scenario, please refer to [link / reference needed]. Figure 5 The data processing method provided in this application embodiment, applied to a data processing device, may include the following steps:

[0112] Step S501: Load the browser's webpage user interface.

[0113] Step S502: Obtain the data to be processed based on the web user interface.

[0114] Step S503: Send the data to be processed to the target computing power bar.

[0115] Step S504: Obtain the processing result of the target computing power stick on the data to be processed.

[0116] Step S505: Extract instructions from the processing results of the data to be processed.

[0117] Step S506: If a data acquisition instruction is extracted from the processing result of the data to be processed, then data acquisition is performed according to the data acquisition instruction to obtain the associated processing data.

[0118] Step S507: Transfer the associated processing data to the target computing power bar.

[0119] Step S508: Receive the processing results of the target computing power stick on the associated processing data.

[0120] In an exemplary embodiment, if the computing power stick needs to acquire further data after processing the data to be processed, a data acquisition instruction can be set in the processing result of the data to be processed. Accordingly, the data processing device can extract the instruction from the processing result of the data to be processed. If the data acquisition instruction is extracted from the processing result of the data to be processed, data is collected according to the data acquisition instruction to obtain the associated processing data. The associated processing data is transmitted to the target computing power stick. The processing result of the target computing power stick on the associated processing data is obtained.

[0121] To facilitate understanding of this embodiment, let's assume the data to be processed is "querying the brand of a vehicle". After the data processing device transmits this data to the computing power stick, the computing power stick needs to acquire an image of the vehicle. Therefore, a data acquisition instruction for acquiring the vehicle image can be set in the processing result of the data to be processed. Correspondingly, after parsing the processing result of the data to be processed, the data processing device needs to acquire an image of the vehicle using an image sensor to obtain associated processing data, and transmit the associated processing data to the computing power stick. The computing power stick then processes the image of the vehicle to obtain the brand information of the vehicle, and returns the brand information of the vehicle as the processing result of the associated data to the data processing device.

[0122] This implementation process demonstrates that after acquiring the processing results of the target computing power stick on the data to be processed, the data processing device can also extract instructions from those results. If a data acquisition instruction is extracted from the processing results, data is collected according to the instruction to obtain associated processing data. This associated processing data is then transmitted to the target computing power stick, and the processing results from the target computing power stick on the associated processing data are received. This achieves the goal of guiding the data processing device to acquire the associated data required for data processing by setting data acquisition instructions, thereby enriching the data on the computing power stick, ensuring the data processing effect of the computing power stick, and ultimately improving the data interaction capabilities of the data processing device.

[0123] In an exemplary embodiment, the data to be processed can be input questions, uploaded images, uploaded audio, etc., and the processing results can include text answers, image annotations, and voice playback. For example, a user can transmit audio to a computing stick via a WebUI. The computing stick performs speech recognition and processes the data based on the intent recognition results. If the intent is control-related, it returns the control intent to the data processing device, such as opening or closing a music player, closing or closing an application, or adjusting the temperature. If it is chat-related, it calls an LLM (Local Language Management) to conduct the chat and returns the text result and a TTS (Text-to-Speech) voice broadcast of the chat result. Of course, users can also call an image generation model to perform image operations, such as image generation and image retrieval.

[0124] In an exemplary embodiment, to ensure data transmission security between the data processing device and the computing power stick, the data processing device can perform data transmission with the computing power stick using encryption and decryption operations. For example, during the process of the data processing device sending the data to be processed to the target computing power stick, the data to be processed can be encrypted to obtain encrypted data to be processed; the encrypted data to be processed is then sent to the target computing power stick so that the target computing power stick can decrypt the encrypted data to be processed to obtain the data to be processed. The encryption and decryption method can be flexibly determined according to the application scenario.

[0125] In an exemplary embodiment, the computing stick can support local caching of models and data, enabling inference even without network access or when the vehicle's infotainment system is offline. Furthermore, the neural network model in the computing stick can be configured offline or downloaded in real-time. It can be downloaded by the computing stick itself via the network, or downloaded by a data processing device and then transmitted to the computing stick for installation; this application does not impose specific limitations on this. During the configuration of the neural network model on the computing stick, the installation package and verification code of the neural network model can be obtained. The installation package is verified using the verification code. If the verification passes, the neural network model is deployed based on the installation package. Then, format detection and compatibility checks are performed on the neural network model. If the checks pass, the neural network model can be converted and quantized. The quantized neural network model is then deployed and registered. Finally, hot reloading or service restart ensures the neural network model functions correctly. During this process, historical versions of the neural network model can also be saved. If a historical version needs to be applied, version rollback can be performed to ensure the application stability of the neural network model. Furthermore, signature verification can be enforced during POST / models / deploy. If no signature is provided, the computing stick can refuse to deploy the model.

[0126] Please see Figure 6 , Figure 6 Another flowchart of a data processing method provided in an embodiment of this application.

[0127] This application provides a data processing method applied to a computing power stick, which may include the following steps:

[0128] Step S601: Obtain the data to be processed sent by the data processing device.

[0129] Step S602: Call the target neural network model to process the data to be processed and obtain the processing result of the data to be processed.

[0130] Step S603: Send the processing result of the data to be processed to the data processing device.

[0131] The data processing method for computing power sticks provided in this application can be described in the above embodiments, and will not be repeated here.

[0132] In an exemplary embodiment, to ensure data privacy, the computing stick can run an inference process within a namespace or container while invoking the target neural network model to process the data, and restrict file system and network access to the inference process; the target neural network model is invoked through the inference process to process the data. Furthermore, the computing stick can connect to more sensors or peripherals, such as USB cameras, radar, and LiDAR, to expand its AI capabilities.

[0133] It's important to note that the computing power stick does not depend on any specific model framework and supports: LLM (Qwen, LLaMA, ChatGLM, InternLM); CV (YOLO, SAM, PaddleOCR); ASR (Wenet, Whisper); TTS (VITS, FastSpeech2), etc. During context storage management, the computing power stick can assign a session_id to each session, storing it locally in SQLite or Redis. During token calculation, it supports real-time statistics of input / output tokens based on different models and provides an ` / api / tokens` query interface. When adding a new model to the computing power stick, simply upload the model file to ` / models`; register the model in `models.json`; the web UI will automatically update the dropdown list.

[0134] In an exemplary embodiment, if the computing power stick needs to process multiple data sets simultaneously, it can prioritize data processing. For example, it can prioritize real-time high-priority tasks (ADAS / wake-up / security), followed by interactive voice, and then entertainment / batch processing. To ensure the processing of high-priority data, the computing power stick can reserve resources for high-priority tasks, such as reserving a certain share of NPU / GPU / CPU capacity (e.g., 25% NPU capacity). It can also automatically perform dynamic data migration when resources are scarce. Furthermore, it can dynamically adjust the weight of the data to be processed based on real-time telemetry.

[0135] In an exemplary embodiment, the computing power stick also needs to handle unexpected situations. For example, in the event of insufficient memory / OOM, it can attempt to unload the least-recently-used model, return `resource_exhausted`, log the event, and attempt to restart the runtime. In the event of model incompatibility, it can revert to the previous stable version and notify the application (`model_deployed:false, reason:compatibility_error`). In high-temperature conditions, it can reduce the frequency; for example, when the temperature > threshold (e.g., 80°C), it can automatically reduce the running frequency / reduce the model accuracy and pause non-critical tasks. In the event of an LLM crash, it can capture a core dump, automatically restart the LLM service, and roll back the current session state to the last checkpoint (if applicable).

[0136] In an exemplary embodiment, the computing stick can also support multi-party collaboration, such as multiple computing sticks being connected to a PC via a USB hub; a web UI displays all computing sticks and automatically load balances the load to support multi-task parallel inference. Furthermore, SoftAP (Wi-Fi hotspot mode) can be enabled on the computing stick, providing the same web UI and API, allowing the PC to access the computing stick without a USB cable.

[0137] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a data processing system provided in an embodiment of this application.

[0138] This application provides a data processing system, applied to a data processing device, which may include:

[0139] Loading module 101 is used to load the browser's web page user interface;

[0140] The data acquisition module 102 is used to acquire data to be processed based on a web page user interface;

[0141] The data to be processed sending module 103 is used to send the data to be processed to the target computing power stick;

[0142] The processing result acquisition module 104 is used to acquire the processing result of the target computing power stick on the data to be processed.

[0143] The target computing stick is connected to the data processing device via a USB interface; the processing result of the data to be processed includes the result obtained after the target computing stick calls the target neural network model to process the data to be processed.

[0144] The data processing system provided in this application embodiment, applied to a data processing device, may further include:

[0145] A connection module is used to determine the user terminal connected to the data processing device.

[0146] The terminal information acquisition module collects computing power resource information, random access memory information, and read-only memory information of the user terminal.

[0147] The configuration module is used to deploy a neural network model in the user terminal according to the computing power resource information, random access memory information, and read-only memory information, so as to configure the user terminal as a computing power stick.

[0148] The data processing system provided in this application embodiment, applied to a data processing device, may further include:

[0149] The instruction extraction module is used to extract instructions from the processing results of the target computing power stick after the result acquisition module obtains the processing results of the data to be processed.

[0150] The data acquisition module is used to extract data acquisition instructions from the processing results of the data to be processed, and then perform data acquisition according to the data acquisition instructions to obtain the associated processing data.

[0151] The associated data transmission module is used to transmit associated processing data to the target computing power stick and to receive the processing results of the associated processing data from the target computing power stick.

[0152] The data processing system provided in this application embodiment, applied to a data processing device, may further include:

[0153] The display module is used to display the models supported by the target computing power stick in the web interface according to the set method before the data acquisition module acquires the data to be processed based on the web interface.

[0154] The user selection instruction acquisition module is used to acquire the selection instructions transmitted by the user through the web interface.

[0155] The model selection module is used to select models displayed on the webpage user interface based on user selection instructions and obtain model selection information.

[0156] The model selection result sending module is used to transmit model selection information to the target computing stick, so that the target computing stick can determine the target neural network model according to the model selection information.

[0157] The data processing system provided in this application embodiment, applied to a data processing device, may further include:

[0158] The configuration information acquisition module is used to acquire the target computing stick's internal network interface configuration information based on the USB virtual network card protocol before the loading module loads the browser's web page user interface.

[0159] Create a module for detecting USB Ethernet. In response to the detection of USB Ethernet, create a network interface according to the intranet interface configuration information.

[0160] The verification module is used to verify the connectivity of the target computing stick based on the IP address agreed upon with the target computing stick.

[0161] The access module is used to access the IP address of the target computing stick through a browser if the connectivity verification is successful, obtain the device information and certificate fingerprint returned by the target computing stick, and establish a communication link with the target computing stick.

[0162] This application provides a data processing system applied to a data processing device, wherein the data to be processed transmission module may include:

[0163] The data to be processed sending unit is used to encrypt the data to be processed to obtain encrypted data to be processed; and to send the encrypted data to be processed to the target computing power stick so that the target computing power stick can decrypt the encrypted data to be processed to obtain the data to be processed.

[0164] This application provides a data processing system applied to a computing power stick, which may include:

[0165] The data receiving module is used to acquire the data to be processed sent by the data processing device.

[0166] The processing module is used to call the target neural network model to process the data to be processed and obtain the processing result of the data to be processed;

[0167] The result sending module is used to send the processing results of the data to be processed to the data processing device.

[0168] Furthermore, embodiments of this application also provide a computer device. Figure 8This is a structural diagram of a computer device 20 according to an exemplary embodiment. The content of the diagram should not be considered as any limitation on the scope of this application.

[0169] Figure 8 This is a schematic diagram of the structure of a computer device 20 provided in an embodiment of this application. The computer device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the data processing method disclosed in any of the foregoing embodiments. Furthermore, the computer device 20 in this embodiment may specifically be a server.

[0170] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the computer device 20; the communication interface 24 can create a data transmission channel between the computer device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0171] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222 and video data 223, etc., and the storage method can be temporary storage or permanent storage.

[0172] The operating system 221 manages and controls the various hardware devices on the computer device 20 and the computer program 222 to enable the processor 21 to perform operations and processing on the data 223 in the memory 22. The operating system 221 can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including computer programs capable of performing the data processing methods executed by the computer device 20 as disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include various data collected by the computer device 20.

[0173] It should be noted that the computer equipment can specifically be a module capable of realizing positioning and communication functions or a terminal device containing a module, etc. The terminal device can specifically be a mobile terminal and / or a smart device, etc. The mobile terminal can specifically be at least one of mobile phones, tablets, laptops, etc. The smart device can specifically be at least one of smartwatches, smart refrigerators, smart speakers, smart washing machines, smart TVs, etc. The module can specifically be any one of 2G communication modules, 3G communication modules, 4G communication modules, 5G communication modules, NB-IoT communication modules, etc.

[0174] Furthermore, this application also discloses a computer storage medium storing a computer program. When the computer program is loaded and executed by a processor, it implements the data processing method steps disclosed in any of the foregoing embodiments.

[0175] This application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the data processing method described in any of the preceding embodiments.

[0176] For descriptions of relevant parts in the data processing system, electronic device, and computer-readable storage medium provided in this application's embodiments, please refer to the detailed description of the corresponding parts in the data processing method provided in this application's embodiments; they will not be repeated here. Furthermore, parts of the technical solutions provided in this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0177] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0178] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data processing method, characterized in that, Applied to data processing equipment, including: Load the browser's webpage user interface; Data to be processed is obtained based on the webpage user interface. The data to be processed is sent to the target computing power bar; Obtain the processing result of the target computing power stick on the data to be processed; The target computing stick is connected to the data processing device via a USB interface; the processing result of the data to be processed includes the result obtained after the target computing stick calls the target neural network model to process the data to be processed; Before loading the browser's webpage user interface, the following steps are also included: Obtain the intranet interface configuration information of the target computing stick based on the USB virtual network card protocol; The USB Ethernet is detected, and in response to the detection of USB Ethernet, a network interface is created according to the intranet interface configuration information. Based on the IP address agreed upon with the target computing power stick, a connectivity verification is performed on the target computing power stick; If the connectivity verification is successful, the browser accesses the IP address of the target computing stick to obtain the device information and certificate fingerprint returned by the target computing stick, so as to establish a communication link with the target computing stick.

2. The data processing method according to claim 1, characterized in that, Also includes: Determine the user terminal connected to the data processing device; Collect computing power resource information, random access memory information, and read-only memory information of the user terminal; Based on the computing power resource information, the random access memory information, and the read-only memory information, a neural network model is deployed in the user terminal to configure the user terminal as a computing power stick.

3. The data processing method according to claim 1, characterized in that, After obtaining the processing result of the target computing power stick on the data to be processed, the method further includes: The processing results of the data to be processed are used to extract instructions; If a data acquisition instruction is extracted from the processing result of the data to be processed, then data is collected according to the data acquisition instruction to obtain the associated processing data; The associated processing data is transmitted to the target computing power bar; Receive the processing results of the target computing power stick on the associated processing data.

4. The data processing method according to claim 1, characterized in that, Before obtaining the data to be processed based on the webpage user interface, the process also includes: The target computing power stick supports the model in the web user interface according to the set method; Obtain the selection instructions transmitted by the user through the webpage user interface; Based on the user's selection command, the model displayed on the webpage user interface is selected to obtain model selection information; The model selection information is transmitted to the target computing power stick so that the target computing power stick determines the target neural network model according to the model selection information.

5. The data processing method according to claim 1, characterized in that, Sending the data to be processed to the target computing power stick includes: The data to be processed is encrypted to obtain encrypted data to be processed; The encrypted data to be processed is sent to the target computing power stick, so that the target computing power stick can decrypt the encrypted data to be processed to obtain the data to be processed.

6. A data processing method, characterized in that, Applied to target computing power, including: The system acquires data to be processed sent by a data processing device, the data to be processed including data acquired by the data processing device based on the web page user interface of a loaded browser; The target neural network model is invoked to process the data to be processed, and the processing result of the data to be processed is obtained; Send the processing result of the data to be processed to the data processing device; The target computing stick is connected to the data processing device via a USB interface; Before acquiring the data to be processed sent by the data processing device, the process also includes: The target computing stick's intranet interface configuration information, configured based on the USB virtual network card protocol, is sent to the data processing device. This enables the data processing device to detect the USB Ethernet. Upon detecting the USB Ethernet, the device creates a network interface according to the intranet interface configuration information. Based on the agreed-upon IP address, it performs connectivity verification on the target computing stick. If the connectivity verification is successful, the device accesses the target computing stick's IP address through the browser to obtain the device information and certificate fingerprint returned by the target computing stick, thereby establishing a communication link with the target computing stick.

7. A data processing system, characterized in that, Applied to data processing equipment, including: The loading module is used to load the browser's webpage user interface; The data acquisition module is used to acquire data to be processed based on the web page user interface. The data to be processed sending module is used to send the data to be processed to the target computing power stick; The processing result acquisition module is used to acquire the processing result of the target computing power stick on the data to be processed; The target computing stick is connected to the data processing device via a USB interface; the processing result of the data to be processed includes the result obtained after the target computing stick calls the target neural network model to process the data to be processed; This also includes: The configuration information acquisition module is used to acquire the intranet interface configuration information of the target computing stick based on the USB virtual network card protocol before the loading module loads the web page user interface of the browser; A module is created for detecting USB Ethernet. In response to the detection of USB Ethernet, a network interface is created according to the intranet interface configuration information. The verification module is used to perform connectivity verification on the target computing power stick based on the IP address agreed upon with the target computing power stick; The access module is used to access the IP address of the target computing stick through the browser if the connectivity verification is successful, obtain the device information and certificate fingerprint returned by the target computing stick, and establish a communication link with the target computing stick.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the data processing method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the data processing method as described in any one of claims 1 to 6.

Images