Calculation demand collection and task generation method and system and electronic equipment

Abstract

The invention provides a calculation demand collection and task generation method and system and electronic equipment, and the method comprises the following steps: receiving user description information, carrying out the content recognition and integrity analysis of multiple types of user description information through a multi-modal large model, and if the integrity of the user description information is insufficient, generating a task; the multi-modal large model guides the user to perfect the user description information through multi-round guidance supplementary questioning until the integrity of the user description information is met, and user demand information is generated; and analyzing the user demand information, and outputting a calculation demand document. The user description information is analyzed through the multi-modal large model, the user demand information is generated, the user is guided to supplement related information by supplementing guide questions to the user, the description of the user on the demand is perfected, and it is guaranteed that the collected information is comprehensive and accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of computing task generation, and in particular relates to a computing requirement collection and task generation method, system and electronic equipment. Background Art

[0002] With the rapid development of quantum and classical computers, users' computing needs are growing and becoming more diverse, requiring them to generate computing tasks tailored to their needs. However, for quantum computers, users must have a thorough understanding of the computer's characteristics and parameters, and must encode their computing needs into specific quantum codes or quantum circuits before submitting a computing task. This requires a high level of expertise and is prone to errors, requiring significant time and effort to analyze their computing needs and task codes. With classical computers, users have numerous computing needs, such as system simulation, model training, or data analysis. However, it is difficult for users to accurately and completely describe and analyze their computing problems and needs, resulting in the inability to quickly and accurately create computing tasks. This results in low computing efficiency and consumes significant time, effort, and cost. Summary of the Invention

[0003] The purpose of the present invention is to propose a method and system for collecting computing requirements and generating tasks, which solves the problem in related technologies that quantum computers or classical computers have high requirements on user expertise and low efficiency when generating computing tasks.

[0004] To this end, in a first aspect, the present invention provides a method for collecting computing requirements, comprising the following steps:

[0005] Receive user description information and perform content recognition and integrity analysis on multiple types of user description information through a multimodal large model. If the user description information is not complete enough, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the user description information is complete and generates user demand information;

[0006] Analyze user demand information and output computing demand documents.

[0007] Optionally, the receiving of user description information and performing content recognition and integrity analysis on multiple types of user description information through a multimodal large model. If the integrity of the user description information is insufficient, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the integrity of the user description information is satisfied. Generating user demand information includes:

[0008] Receive user description information, perform natural language understanding of the user description information in the multimodal large model, identify key content, and detect the completeness of the user description information. Initiate guided supplementary questions based on the completeness of the user description information to complete a round of dialogue interaction; repeat multiple rounds of dialogue interaction, and integrate and organize the user description information through the multimodal large model to generate structured user demand information.

[0009] Optionally, the user description information includes data information in the form of text, voice, picture or video, which is used to describe the quantum computing task or the classical computing task.

[0010] Optionally, the detecting of the integrity of the user description information and initiating guided supplementary questions based on the integrity of the user description information includes: matching the user description information with a preset requirement template, determining whether there is any description in the user description information that does not conform to the preset requirement template; if there is any description that does not conform to the preset requirement template, it is determined that the user description information is insufficiently complete; the multimodal large model initiates guided supplementary questions for the description that does not conform to the preset requirement template based on the cross-attention mechanism.

[0011] Optionally, parsing the user demand information and outputting a computing demand document includes:

[0012] By comparing user demand information with predefined computing demand templates, the semantic label of each parameter in the user demand information is identified. At the same time, combined with historical tasks and domain knowledge, through comparative analysis and probabilistic model evaluation, the user demand information is supplemented to obtain candidate computing demand solutions, and the final computing demand document is output based on the candidate computing demand solutions.

[0013] Optionally, if multiple candidate computing requirement solutions are parsed based on the user requirement information, the candidate computing requirement solutions are weighted and sorted based on the context information, and a final computing requirement document is output with the optimal parameter combination.

[0014] In a second aspect, a method for generating a computing demand task is provided, comprising the following steps:

[0015] Receiving a computing requirement document, analyzing user description information using a multimodal large model, generating user requirement information, and then parsing the user requirement information;

[0016] Evaluate computing requirements documents, dynamically analyze computing resources that meet computing requirements based on the computing device status of different computing platforms, and obtain computing resource configuration information;

[0017] Template matching is performed based on the computing requirement document and computing resource configuration information to generate executable computing tasks.

[0018] Optionally, evaluating the computing requirements document and dynamically analyzing computing resources that meet the computing requirements based on the computing device status of different computing platforms to obtain computing resource configuration information includes:

[0019] Real-time collection of computing device status on different computing platforms;

[0020] Use the reinforcement learning model to predict the parameters in the computing requirements document and obtain the predicted data;

[0021] Computing resource configuration information is obtained by matching the computing device status and prediction data of different computing platforms.

[0022] Optionally, performing template matching based on the computing requirement document and computing resource configuration information to generate an executable computing task includes:

[0023] The computing requirement document and computing resource configuration information are matched with the pre-set task generation template to generate a preliminary computing task file; the computing task format is checked through rule verification and syntax parsing algorithms to obtain the computing task file.

[0024] In a third aspect, a method for collecting computing requirements and generating tasks is provided, comprising the following steps:

[0025] Receive user description information, perform content recognition and integrity analysis on multiple types of user description information through a multimodal large model, and generate user demand information;

[0026] Analyze user demand information and output computing demand documents;

[0027] Evaluate computing requirements documents, analyze computing resources that meet computing requirements, and obtain computing resource configuration information;

[0028] Generate executable computing tasks based on computing requirements documents and computing resource configuration information.

[0029] In a fourth aspect, a computing requirement collection and task generation apparatus is provided, comprising:

[0030] The multimodal interaction module receives user description information, performs content recognition and integrity analysis on multiple types of user description information through a multimodal large model, and generates user demand information;

[0031] Computing requirement parsing module, used to parse user requirement information and output computing requirement documents;

[0032] Dynamic evaluation module, used to evaluate computing requirement documents, analyze computing resources that meet computing requirements, and obtain computing resource configuration information;

[0033] The computing task generation module generates executable computing tasks based on computing requirement documents and computing resource configuration information.

[0034] In a fifth aspect, an electronic device is provided, comprising a memory and a processor;

[0035] The memory stores computer-executable instructions;

[0036] The processor executes the computer-executable instructions stored in the memory, so that the processor performs the computing requirement collection and task generation method.

[0037] Beneficial effects:

[0038] (1) The present disclosure provides a method, system and electronic device for collecting computing requirements and generating tasks, which analyzes user description information through a multimodal large model to generate user requirement information, and guides users to supplement relevant information by asking supplementary questions to improve their description of requirements, thereby ensuring that the collected information is comprehensive and accurate.

[0039] (2) This disclosure compares user demand information with predefined computing demand templates to identify the semantic labels of each parameter in the user demand information. At the same time, it combines historical tasks and domain knowledge, and through comparative analysis and probability model evaluation, it completes the user demand information and comprehensively analyzes the task requirements, providing a basis for subsequent resource matching and task generation.

[0040] (3) The present disclosure matches the computing device status of different computing platforms collected in real time with the predicted data to achieve accurate prediction of computing resource parameters, match computing resources in real time, and complete dynamic evaluation of computing resource requirements applicable to quantum or classical systems.

[0041] (4) The present disclosure generates executable computing tasks through computing requirement documents and computing resource configuration information. The computing tasks can be directly submitted to the corresponding computing platform for execution, thereby realizing the automatic generation of computing tasks, reducing manual participation, improving the efficiency of computing task generation, and ensuring the quality of task files.

[0042] 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 invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 A method flow chart of an embodiment of a method for collecting computing requirements in the present disclosure;

[0045] Figure 2 A method flow chart of an embodiment of a method for generating a computing demand task in the present disclosure;

[0046] Figure 3 A method flow chart of an embodiment of a method for collecting computing requirements and generating tasks in the present disclosure;

[0047] Figure 4 A system structure diagram of an embodiment of a computing requirement collection and task generation device in the present disclosure;

[0048] Figure 5 A system structure diagram of an embodiment of an electronic device in the present disclosure

[0049] In the figure, 101-multimodal interaction module, 102-computing requirement analysis module, 103-dynamic evaluation module, 104-computing task generation module, 200-electronic device, 201-processor, 202-memory, 203 communication component, 204-bus. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0051] In the specification and claims of this application, as well as in the accompanying drawings, the terms "first," "second," "third," "fourth," and the like are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that such terms are interchangeable where appropriate. For example, first information could be referred to as second information, and similarly, second information could be referred to as first information without departing from the scope of this disclosure.

[0052] The word "if" as used herein may be interpreted as "when" or "when" or "in response to determining," depending on the context.

[0053] Furthermore, as used herein, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise.

[0054] It should be further understood that the terms “comprises” and “includes” indicate the existence of features, steps, operations, elements, components, items, types, and / or groups, but do not preclude the existence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, types, and / or groups.

[0055] The terms "or" and "and / or" as used herein are to be interpreted as inclusive, or mean any one or any combination. Thus, "A, B, or C" or "A, B, and / or C" means "any one of the following: A; B; C; A and B; A and C; B and C; A, B, and C." An exception to this definition occurs only when a combination of elements, functions, steps, or operations are inherently mutually exclusive in some manner.

[0056] In a first aspect, the present disclosure provides Figure 1 A computing requirement collection method shown includes the following steps:

[0057] S101: Receive user description information and perform content recognition and integrity analysis on multiple types of user description information using a multimodal large model. If the integrity of the user description information is insufficient, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the user description information is complete, thereby generating user demand information.

[0058] Among them, the multimodal large model detects the completeness of the user's description information and initiates guided supplementary questions based on the completeness of the user's description information;

[0059] It includes receiving user description information, performing natural language understanding on the user description information through a multimodal large model, identifying key content, and detecting the completeness of the user description information, initiating guided supplementary questions based on the completeness of the user description information, and completing a round of dialogue interaction; repeating multiple rounds of dialogue interaction, and integrating and organizing the user description information through a multimodal large model to generate structured user demand information. In one embodiment, user description information is received through an interactive interface with a dialog box and file upload function, and the user is guided to describe the background, goals and related parameters of the computing needs. When the user fully provides the background, goals and related parameters, it is determined that the completeness of the user description information is met. A semantic analysis model is built into the multimodal large model to extract key information based on the user description information, such as computing scenarios, algorithm names or optimization goals.

[0060] In one embodiment, the multimodal large model is adjusted based on an open source pre-trained multimodal large model to adapt it to the demand analysis of quantum computing tasks, classical computing tasks, and quantum-classical hybrid computing tasks. Pre-trained multimodal large models such as Qwen2.5-Omni can be selected, which supports cross-modal understanding of text, images, audio, and video. In order to improve the capabilities and adaptability of the multimodal large model in quantum computing tasks, classical computing tasks, and quantum-classical hybrid computing tasks, the adjustment methods mainly include: 1. Constructing a dedicated dataset for quantum and classical computing. Integrate the demand descriptions, task cases, related documents, and user interaction records in the fields of quantum computing and classical computing, and annotate the data and information into a structured format. The relevant documents can be operation documents, technical reports, or computing task examples. 2. Adaptive fine-tuning strategy: Based on the existing open source multimodal large model, load the pre-trained weights as a starting point, design a loss function for quantum and classical computing demand analysis, and optimize the accurate analysis of multimodal information, the automatic completion of fuzzy information, and the accuracy of key parameter extraction. While fine-tuning the multimodal model, it is linked with the subsequent computing resource evaluation and task generation modules to ensure that the demand analysis results can be directly used for dynamic resource evaluation and automatic task generation. 3. Experimentation and verification, build a special verification data set to evaluate the accuracy and adaptability of the model in parsing different computing task requirements. For example, for the QAOA (Quantum Approximate Optimization Algorithm) and VQE (Variational Quantum Eigensolver) algorithm requirements, experiments are conducted in the quantum computing field and the classical computing field for materials science computing tasks to test the model's adaptability to the requirements of different scenarios. Then, multiple fine-tuning is performed based on the experimental results until the best effect is achieved.

[0061] User description information includes data in the form of text, voice, images, or video, and is used to describe quantum computing tasks or classical computing tasks. Users can upload user description information in the form of text, voice, images, or videos, for example, through auxiliary files such as instruction documents, algorithm codes, calculation flowcharts, and data sets.

[0062] Detecting the integrity of user description information and initiating guided supplementary questions based on the integrity of the user description information includes: matching the user description information with a preset requirement template, determining whether the user description information contains descriptions that do not conform to the preset requirement template, and if so, determining that the user description information is insufficiently complete. The multimodal large model initiates guided supplementary questions for descriptions that do not conform to the preset requirement template based on a cross-attention mechanism. The preset requirement template contains key information that needs to be collected, such as computing task type, task description, key parameters, algorithm name, or input data, and each key information is clearly described. After extracting the key information from the user description information, the multimodal large model matches it with the key information of the preset requirement template to determine whether there is ambiguous or missing information in the user description information. If there is ambiguous or missing information, the description does not conform to the preset requirement template, and the user description information is determined to be insufficiently complete. The multimodal large model initiates guided supplementary questions for descriptions that do not conform to the preset requirement template based on a cross-attention mechanism, guiding the user to supplement the information so that the key information provided by the user conforms to the key information of the preset requirement template.

[0063] During the process of initiating guided and supplementary questioning with users, a reinforcement learning algorithm is employed to continuously optimize the interaction strategy, adjusting the questioning sequence and depth based on user feedback to ensure that the resulting computing requirement information is comprehensive and accurate. An optional reinforcement learning algorithm is Proximal Policy Optimization (PPO). During each round of interaction with the user, a reward signal is calculated based on user feedback, such as clarity and information completion, and the questioning strategy for the next round is adjusted. The quality of the user's response to the system's questions over multiple rounds of dialogue, such as whether they accurately provide missing information or clarify the requirement description, constitutes a reward signal. Clear responses and supplementary key information generate positive rewards, while ambiguous or incomplete responses, or repeated questions, generate negative rewards. The multimodal large model is integrated with the reinforcement learning algorithm. Based on the pre-trained multimodal large model, reinforcement learning continuously adjusts the parameters of the policy network. This allows the system to more effectively capture and correct for ambiguities or omissions in user descriptions during user interaction, ensuring the integrity of user descriptions.

[0064] The completeness of user descriptions must meet the following requirements: 1. Clear background and objectives. The user's task description context must be extracted, including the problem domain and the primary objective. The problem domain can be quantum computing, classical computing, or hybrid computing. The primary task template can be solving combinatorial optimization problems, chemical simulation, or materials science computing. The actual computational scenario and expected results must be described. 2. Key information must be complete. For quantum computing tasks, key information includes: number of qubits, algorithm type, operating parameters such as the number of iterations and measurements, and device type. Examples of algorithm types include QAOA and VQE, and device types include superconducting quantum computers, optical quantum computers, or quantum simulators. For classical computing tasks, key information includes: data input format and requirements, algorithm or model name, computational scale, and hardware requirements. Hardware requirements include CPU and GPU configurations. 3. Clear details must be described. There should be no ambiguity or ambiguity in the user description. For example, clear values or ranges should be given for algorithm parameters, initial data, and constraints. Sufficient text should be provided for problem background, data sources, and expected results. If the user's description is ambiguous or missing, multiple rounds of interaction can be used to supplement and confirm it, ensuring that the final requirement information is unambiguous. 4. Supplementary files and multimodal data. In addition to text descriptions, users are encouraged to upload relevant auxiliary files such as algorithm flowcharts, experimental data, and explanatory documents. These multimodal data can serve as verification and supplementary explanations, further ensuring the comprehensiveness and accuracy of the information.

[0065] In one embodiment, when a user inputs a description such as "I need to use the QAOA algorithm to solve a quantum combinatorial optimization problem involving 20 variables," the system checks the completeness of the user description and initiates supplementary questions, prompting the user to provide additional information such as the problem scenario, computational scale, initial data, and constraints, and then guides the user to upload auxiliary files. In another embodiment, after the user uploads a problem-solving flowchart and additional data documents related to the problem, the multimodal large model intelligently identifies and parses the node and edge structure in the diagram, extracting key information from the flowchart. Based on this key information and additional data, the system generates a problem-solving document and appends relevant structured information, including core parameters such as the quantum computer type, number of qubits, number of algorithm iterations, and number of measurement settings.

[0066] S102, parsing user demand information and outputting a computing demand document;

[0067] By comparing user demand information with predefined computing demand templates, the semantic label of each parameter in the user demand information is identified. At the same time, combined with historical tasks and domain knowledge, through comparative analysis and probabilistic model evaluation, the user demand information is supplemented to obtain candidate computing demand solutions, and the final computing demand document is output based on the candidate computing demand solutions.

[0068] A hybrid parsing algorithm is used to parse user demand information, including rule-based pattern matching integrated with a semantic understanding model based on deep neural networks. First, the user demand information is compared with predefined computing demand templates to identify the semantic labels for each parameter in the user demand information. For example, for quantum computing tasks, the number of qubits, algorithm type, operating parameters, and input data format are extracted. Algorithm types include QAOA and VQE, and operating parameters include the number of iterations and the number of measurements. During the parsing process, user demand information is supplemented through comparative analysis and probabilistic model evaluation, combining historical tasks and domain knowledge. For example, if the user does not explicitly provide the number of qubits required for the calculation, the computing demand parsing module can automatically infer the appropriate bit range based on the problem scale and auxiliary documents uploaded by the user, thereby forming candidate computing demand solutions. Among them, the probabilistic model options include Bayesian networks, hidden Markov models, and conditional random fields.

[0069] If multiple candidate computing requirement solutions are parsed based on user requirement information, the candidate computing requirement solutions are weighted and sorted according to the context information, and the final computing requirement document is output with the optimal parameter combination.

[0070] Second, as Figure 2 As shown, a method for generating a computing demand task is provided, comprising the following steps:

[0071] S201: Receive a computing requirement document, analyze user description information through a multimodal large model, generate user requirement information, and then parse the user requirement information to obtain the information.

[0072] S202: Evaluate the computing requirements document, dynamically analyze computing resources that meet the computing requirements based on the computing device status of different computing platforms, and obtain computing resource configuration information;

[0073] These include: real-time collection of computing device status across different computing platforms;

[0074] The parameters in the computing requirement document are predicted through the reinforcement learning model to obtain the predicted data; the reinforcement learning model can optionally use the PPO algorithm to predict the required number of quantum bits, error correction requirements, and system running time.

[0075] Computing resource configuration information is obtained by matching the computing device status and predicted data from different computing platforms. The FedAvg algorithm integrates task logs and software and hardware data from different computing platforms to construct a cross-platform computing resource evaluation model to ensure that the prediction results have higher generalization and applicability. During the evaluation process, the dynamic evaluation module collects the computing device status of different computing platforms in real time, matches it with the user's computing needs, and proposes an appropriate computing resource configuration. The computing resource configuration information is fed back to the user in the form of an evaluation report for review. The evaluation report not only includes information such as the recommended computing device model, expected runtime, and success rate, but also provides task splitting and parallel computing solutions.

[0076] S203: Perform template matching based on the computing requirement document and computing resource configuration information to generate an executable computing task.

[0077] This includes: matching the computing requirement document and computing resource configuration information with the pre-set task generation template to generate a preliminary computing task file; performing format checking on the computing task through rule verification and syntax parsing algorithms to obtain a computing task file. The computing task file is a file that meets the target computing platform standards generated based on the computing requirement document and computing resource configuration information, such as a task code that conforms to the QASM (Quantum Assembly Language) format. The task generation template is a template pre-set in the system, covering different types of computing tasks, such as quantum computing tasks, classical computing tasks, and quantum-classical hybrid computing tasks. The computing requirement document and computing resource configuration information are matched with the pre-set task generation template through the template engine, and the parameters are embedded in the preset code framework to generate a preliminary computing task file. The computing task is then format checked through rule verification and syntax parsing algorithms to ensure that the computing task file has no logic and syntax errors before being submitted to the computing platform.

[0078] Thirdly, as Figure 3 As shown, a method for collecting computing requirements and generating tasks is provided, including the following steps:

[0079] S301: Receive user description information and perform content recognition and integrity analysis on multiple types of user description information using a multimodal large model. If the user description information is not complete enough, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the user description information is complete enough, thereby generating user demand information.

[0080] S302, parsing user demand information and outputting a computing demand document;

[0081] S303: Evaluate the computing requirements document, dynamically analyze computing resources that meet the computing requirements based on the computing device status of different computing platforms, and obtain computing resource configuration information;

[0082] S304: Perform template matching based on the computing requirement document and computing resource configuration information to generate an executable computing task.

[0083] Fourthly, Figure 4 As shown, a computing requirement collection and task generation device is provided, comprising:

[0084] The multimodal interaction module 101 receives user description information, performs content recognition and integrity analysis on multiple types of user description information through a multimodal large model, and generates user demand information;

[0085] Among them, the multimodal large model detects the completeness of the user's description information and initiates guided supplementary questions based on the completeness of the user's description information;

[0086] The multimodal interaction module 101 is the entry point for user interaction. It captures and understands the multimodal information input by users through natural language processing, speech-to-text conversion, image content analysis, and video content analysis. In one embodiment, the multimodal interaction module 101 is based on a large multimodal model based on the Transformer architecture. Through in-depth analysis of user descriptions, it achieves a preliminary understanding of user needs, generates user need information, and transmits it to the computational needs analysis module.

[0087] The multimodal interaction module 101 prompts the user to supplement and complete the required computing requirement information during multiple rounds of dialogue with the user. For the collected computing requirement information, the multimodal interaction module 101 will remind the user to check and confirm it during the interaction process. For content that the user does not know or is uncertain about, the multimodal interaction module 101 gives the user corresponding prompts and suggestions based on the previous interaction content to assist the user in confirming the relevant content. The multimodal interaction module 101 not only supports multiple input methods, but also has contextual memory and multi-round dialogue capabilities, and can integrate information across rounds to form structured demand data. The data processing flow of the entire multimodal interaction module 101 includes data preprocessing, feature extraction, semantic matching and information fusion, and ultimately converts multimodal data into structured computing requirement information.

[0088] The computing requirement parsing module 102 is used to parse the user requirement information and output a computing requirement document;

[0089] Computing requirement analysis module 102 employs a hybrid analysis algorithm, combining rule-based pattern matching with a semantic understanding model based on deep neural networks. It first compares user requirement information with predefined computing requirement templates to identify the semantic labels for each parameter in the user requirement information. During the analysis process, it integrates historical tasks and domain knowledge, using comparative analysis and probabilistic model evaluation to complete the user requirement information and generate candidate computing requirement solutions.

[0090] Dynamic evaluation module 103, used to evaluate the computing requirement document, analyze the computing resources that meet the computing requirements, and obtain computing resource configuration information;

[0091] Among them, the dynamic evaluation module 103 matches the computing device status and prediction data of different computing platforms to obtain computing resource configuration information. The dynamic evaluation module 103 integrates task logs and software and hardware data from different computing platforms through the federated averaging algorithm to build a cross-platform computing resource evaluation model to ensure that the prediction results have higher generalization ability and applicability. During the evaluation process, the dynamic evaluation module 103 will collect the computing device status of different computing platforms in real time, match it with the user's computing needs, and propose appropriate computing resource configuration. The computing resource configuration information will be fed back to the user in the form of an evaluation report for review. The evaluation report not only contains information such as the recommended computing device model, expected running time and success rate, but also provides task splitting, parallel computing solutions, etc. For special scenarios, such as quantum chemistry simulation or combinatorial optimization problems, the dynamic evaluation module 103 will also generate special parameter optimization solutions to further reduce the risk of task execution failure.

[0092] The computing task generation module 104 generates executable computing tasks according to the computing requirement document and computing resource configuration information.

[0093] Among them, the computing task generation module 104 matches the computing requirement document and computing resource configuration information with the pre-set task generation template to generate a preliminary computing task file; and performs format check on the computing task through rule verification and syntax parsing algorithm to obtain the computing task file.

[0094] The computing task generation module 104 also embeds a feedback mechanism, which collects actual operation data and error logs during task execution, and uses these data to continuously iteratively train the template engine and parameter generation algorithm to further improve the intelligence level of computing task generation.

[0095] The computing task generation module 104 supports multiple interface forms, such as RESTful API, WebSocket API, etc., can be connected to common computing platforms, realize automatic submission and status feedback of computing tasks, and support priority setting, such as queueing high-priority tasks.

[0096] After the computing task generation module 104 submits the computing task file to the task computing platform, the user can view the task information, such as the task ID, estimated completion time, and real-time progress, through a visual interface.

[0097] Taking the calculation task of the band structure of superconducting materials in a materials science experiment as an example, the implementation steps of an embodiment of a computing requirement collection and task generation device are as follows:

[0098] The multimodal interaction module 101 receives the text description uploaded by the user, "I need to calculate the band structure of a system composed of 25 atoms in a new type of superconducting material", and uploads a tight binding model schematic and related experimental data files. The multimodal interaction module 101 understands the text description, analyzes the keywords "superconducting material", "25 atoms", and "band structure", and analyzes the atomic arrangement and bonding relationship in the schematic diagram through the image analysis function. If, after the multimodal large model analysis, the user's current description is missing information, the multimodal interaction module 101 actively guides and prompts the user to supplement the calculation-related information, such as guiding the user to select a specific quantum algorithm (it is recommended that the user use the VQE algorithm), set the number of iterations, determine the calculation steps, etc. After multiple rounds of interaction, the multimodal interaction module 101 integrates the interaction results into structured computing requirement information.

[0099] The computational requirements analysis module 102 performs in-depth semantic analysis and structural organization of computational requirements information. This module performs semantic analysis of the computational requirements data and uses a predefined materials science computational requirements template to identify the semantic labels for each parameter. If the computational requirements are missing non-critical information required by the computational requirements template, the computational requirements analysis module 102 will independently infer and complete the information based on historical data. The analysis results are output as a structured computational requirements document, which details the computational parameters, computational process, and auxiliary information.

[0100] The dynamic assessment module 103 performs a computing resource assessment based on the computing requirements document. Based on historical data and the real-time hardware and software status of the computing platform, it predicts the computing resources required for the calculation. The dynamic assessment module 103 collects the computing device status of different computing platforms in real time and matches it with the user's computing requirements. It then selects the appropriate computing resource configuration, such as a 100-bit superconducting quantum computer and a GPU server with 96GB of video memory on a specific computing platform, and adopts a distributed quantum-classical hybrid computing solution. Based on the computing resource configuration information, the dynamic assessment module 103 generates an assessment report and provides feedback to the user. The assessment report not only includes information such as the recommended computing device model, estimated runtime, and success rate, but also provides task splitting and parallel computing solutions.

[0101] The computing task generation module 104 generates a computing task file that meets the standards of the target computing platform based on the computing requirements document and computing resource configuration information. Finally, the system calls the computing platform's API interface to send the computing task to the computing device and returns relevant information such as the task ID, estimated completion time, and execution progress to the user interface.

[0102] Fifthly, Figure 5 As shown, an electronic device is provided, characterized in that it includes: a memory, a processor;

[0103] Memory stores computer-executable instructions;

[0104] The processor executes the computer-executable instructions stored in the memory, so that the processor performs the above method.

[0105] In one embodiment, the electronic device 200 includes: at least one processor 201 and a memory 202. Optionally, the electronic device 200 further includes a communication component 203. The processor 201, the memory 202, and the communication component 203 are connected via a bus 204.

[0106] During the specific implementation process, at least one processor 201 executes the computer-executable instructions stored in the memory 202, so that the at least one processor 201 performs the above method.

[0107] The specific implementation process of the processor 201 can be found in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here in this embodiment.

[0108] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0109] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0110] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified into address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.

[0111] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.

Claims

1. A method for collecting computing requirements, characterized in that: The steps include: Receive user description information and perform content recognition and integrity analysis on multiple types of user description information through a multimodal large model. If the user description information is not complete enough, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the user description information is complete and generates user demand information; Analyze user demand information and output computing demand documents.

2. A computing requirement collection method according to claim 1, characterized in that: The receiving of user description information and performing content recognition and integrity analysis on multiple types of user description information through the multimodal large model. If the integrity of the user description information is insufficient, the multimodal large model guides the user to improve the user description information through multiple rounds of guided supplementary questions until the integrity of the user description information is satisfied. The generated user demand information includes: Receive user description information, perform natural language understanding of the user description information in the multimodal large model, identify key content, and detect the completeness of the user description information. Initiate guided supplementary questions based on the completeness of the user description information to complete a round of dialogue interaction; repeat multiple rounds of dialogue interaction, and integrate and organize the user description information through the multimodal large model to generate structured user demand information.

3. A computing requirement collection method according to claim 2, characterized in that: The user description information includes data information in the form of text, voice, picture or video, which is used to describe the quantum computing task or the classical computing task.

4. A computing requirement collection method according to claim 2, characterized in that: The method of detecting the integrity of the user description information and initiating guided supplementary questions based on the integrity of the user description information includes: matching the user description information with a preset demand template, determining whether there is any description in the user description information that does not conform to the preset demand template, and if there is any description that does not conform to the preset demand template, determining that the user description information is insufficiently complete, and initiating guided supplementary questions for the description that does not conform to the preset demand template based on the cross-attention mechanism of the multimodal large model.

5. The method for collecting computing requirements according to claim 1, wherein: The analysis of user demand information and the output of computing demand documents include: By comparing user demand information with predefined computing demand templates, the semantic label of each parameter in the user demand information is identified. At the same time, combined with historical tasks and domain knowledge, through comparative analysis and probabilistic model evaluation, the user demand information is supplemented to obtain candidate computing demand solutions, and the final computing demand document is output based on the candidate computing demand solutions.

6. A computing requirement collection method according to claim 5, characterized in that: If multiple candidate computing requirement solutions are parsed based on user requirement information, the candidate computing requirement solutions are weighted and sorted according to the context information, and the final computing requirement document is output with the optimal parameter combination.

7. A method for generating computing demand tasks, characterized in that: The steps include: Receiving a computing requirement document, analyzing user description information using a multimodal large model, generating user requirement information, and then parsing the user requirement information; Evaluate computing requirements documents, dynamically analyze computing resources that meet computing requirements based on the computing device status of different computing platforms, and obtain computing resource configuration information; Template matching is performed based on the computing requirement document and computing resource configuration information to generate executable computing tasks.

8. A method for generating computing demand tasks according to claim 7, characterized in that: The computing resource configuration information obtained by evaluating the computing requirement document and dynamically analyzing the computing resources that meet the computing requirements based on the computing device status of different computing platforms includes: Real-time collection of computing device status on different computing platforms; Use the reinforcement learning model to predict the parameters in the computing requirements document and obtain the predicted data; Computing resource configuration information is obtained by matching the computing device status and prediction data of different computing platforms.

9. A method for generating computing demand tasks according to claim 7, characterized in that: The template matching based on the computing requirement document and computing resource configuration information to generate an executable computing task includes: The computing requirement document and computing resource configuration information are matched with the pre-set task generation template to generate a preliminary computing task file; the computing task format is checked through rule verification and syntax parsing algorithms to obtain the computing task file.

10. A method for collecting computing requirements and generating tasks, characterized in that: The steps include: Receive user description information and perform content recognition and integrity analysis on multiple types of user description information through a multimodal large model. If the user description information is not complete enough, the multimodal large model guides the user to complete the user description information through multiple rounds of guided supplementary questions until the user description information is complete and generates user demand information; Analyze user demand information and output computing demand documents; Evaluate computing requirements documents, dynamically analyze computing resources that meet computing requirements based on the computing device status of different computing platforms, and obtain computing resource configuration information; Template matching is performed based on the computing requirement document and computing resource configuration information to generate executable computing tasks.

11. A computing requirement collection and task generation device, characterized in that: include: The multimodal interaction module is used to receive user description information, perform content recognition and integrity analysis on multiple types of user description information through a multimodal large model, and generate user demand information; Computing requirement parsing module, used to parse user requirement information and output computing requirement documents; Dynamic evaluation module, used to evaluate computing requirement documents, analyze computing resources that meet computing requirements, and obtain computing resource configuration information; The computing task generation module generates executable computing tasks based on computing requirement documents and computing resource configuration information.

12. An electronic device, characterized in that: Including memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method of claim 10 .

Images