Quantum chip cooperative computing method and device based on over-fusion architecture

Abstract

The application relates to a quantum chip cooperative computing method and device based on an over-amount fusion architecture, and relates to the technical field of quantum computing technology.The application breaks through the scale and accuracy bottleneck of a single quantum chip by using an over-amount fusion architecture and a cooperative computing model, realizes intelligent task decomposition and dynamic optimal scheduling of multiple quantum chips and classical computing power, deeply couples hardware performance parameters and computing task characteristics, maximizes resource utilization, and through real-time coherence monitoring and active fault tolerance maintenance in the whole computing cycle, and based on dynamic weight fusion of credibility perception, the execution success rate of a large-scale complex quantum computing task, the final result accuracy and the overall robustness of the system are improved, and an efficient and reliable system-level solution is provided for practical quantum computing.

Description

Technical Field

[0001] This application relates to the field of quantum computing technology, and in particular to a quantum chip collaborative computing method and device based on a super-quantum fusion architecture. Background Technology

[0002] With the rapid development of information technology, classical computing is gradually approaching the physical limits of Moore's Law, and the bottleneck of computing power is becoming increasingly prominent. Quantum computing, with its superposition and entanglement properties of qubits, has become the core direction for breaking through the constraints of computing power. It has shown potential far exceeding that of classical computing in complex scenarios such as drug development, materials science, and pricing of financial derivatives. In order to give full play to the complementary advantages of quantum computing and classical computing, quantum-supercomputer integrated collaborative computing has become an industry consensus. By combining the parallel processing capability of quantum computing with the efficient numerical computing capability of supercomputers, complex problems can be solved efficiently. Current quantum-supercomputer integrated computing solutions are mostly simple superpositions of quantum computers and supercomputers, achieving task interaction and data feedback through traditional information transmission methods. Essentially, they are still two independent hardware devices operating in parallel, resulting in numerous technical shortcomings. Firstly, the depth of collaboration is insufficient. Task division often adopts a serial process: after the supercomputer completes parameter preparation and optimization, it is handed over to the quantum computer to execute quantum circuits, and then the results are retrieved for iterative optimization. A significant amount of time is spent on data transmission and task switching, failing to achieve real-time collaboration and making it difficult to leverage the overall computing power advantage. This leads to low efficiency in large-scale, complex task processing. Secondly, existing architectures lack a unified hardware integration and resource scheduling mechanism. The quantum measurement and control system and the classical computing system are independent, increasing system complexity and integration difficulty, and resulting in insufficient computing power for quantum bit manipulation. Furthermore, quantum bits are extremely sensitive to environmental interference, exhibiting prominent decoherence problems. Traditional collaborative models cannot provide real-time error correction and environmental control support, further restricting computational accuracy and stability. Simultaneously, existing technologies struggle to adapt to the collaborative needs of quantum chips using multiple technology routes. Different types of quantum chips, such as superconducting and ion traps, each have their advantages but significant differences in characteristics, lacking a unified supercomputer-supercomputer integrated architecture to achieve resource integration and efficient scheduling. Furthermore, issues such as inconsistent heterogeneous instruction sets and inconsistencies in the mapping between quantum state data and classical memory lead to prominent computing power silos, making it difficult to efficiently integrate quantum computing power into general computing systems and limiting the practical application and large-scale deployment of quantum-hyper-convergence technology. Summary of the Invention

[0003] Therefore, it is necessary to provide a quantum chip collaborative computing method and device based on a super-quantum fusion architecture that can achieve deep integration, real-time collaboration and intelligent scheduling of quantum chips and classical computing systems, addressing the aforementioned technical problems.

[0004] Firstly, a collaborative computing method for quantum chips based on a super-fusion architecture is provided, the method comprising: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model to obtain a decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. Based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and a computing task allocation scheme is output based on the computing power evaluation parameters. According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. In response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task's solution request.

[0005] Optionally, before fusing the computation results and generating the final computation result in response to the completion of the parallel computing operation, the method further includes: During the parallel computing operation, relevant parameters of the quantum chip cluster are collected, including quantum state parameters, environmental parameters, and performance parameters. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet the preset standard, the relevant parameters are adjusted based on the preset adjustment mechanism until the coherence of the qubits meets the preset standard.

[0006] Optionally, building a hyperconverged architecture includes: Deploy the quantum chip cluster, the classical computing power cluster, and the target control node; Construct heterogeneous interconnect modules; Based on the heterogeneous interconnection module, the corresponding interface of the target control node is interconnected with the quantum measurement and control port of the quantum chip cluster and the high-speed I / O port of the classical computing power cluster, respectively. Clock synchronization calibration and signal noise reduction are performed on the interconnect links, and the super-converged architecture is generated upon completion of the execution.

[0007] Optionally, constructing a computational task decomposition model includes: Determine the dimensions for decomposing computational tasks, and the quantifiable computational task features corresponding to those dimensions. Based on the benchmark test computation task set, execute the corresponding computation tasks and obtain the computation task execution data; Based on the quantifiable computational task characteristics, an initial computational task decomposition function is constructed, and the initial weight parameters in the initial computational task decomposition function are determined using the computational task execution data, and an initial computational task pattern knowledge base is constructed. In the actual operation of the super-fusion architecture, the predicted data of the computation task decomposition results generated based on the current computation task decomposition function, as well as the corresponding actual execution result data of the computation task, are monitored. The predicted data is compared with the actual execution result data to generate feedback data for the initial computation task decomposition function; Based on the feedback data of the initial computation task decomposition function, the weight parameters in the initial computation task decomposition function are dynamically adjusted using an online learning algorithm to obtain the computation task decomposition model, and the initial computation task pattern knowledge base is updated.

[0008] Optionally, constructing a computing power evaluation model includes: Determine the feature set of computing tasks and the feature set of hardware units, run a benchmark test program that includes multiple computing task modes, obtain actual performance data of executing computing tasks on the target hardware unit, and generate a training dataset; Construct an initial computing power evaluation model, which includes one or more feature transformation sub-functions and corresponding adjustable weight parameters; Based on the training dataset, the adjustable weight parameters and coefficients in the feature transformation sub-function of the initial computing power evaluation model are adjusted by an optimization algorithm until the initial computing power evaluation model meets the preset standard, thus obtaining the target computing power evaluation model. The target computing power evaluation model is deployed to the super-converged architecture. During the operation of the super-converged architecture, computing task allocation schemes and corresponding actual execution effect data are collected. Based on a preset time period, the newly collected data is used to incrementally learn the target computing power evaluation model in order to update the parameters of the target computing power evaluation model.

[0009] Optionally, in response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model, and the decomposition results include: The computational task corresponding to the computational task solution request is formalized to obtain a weighted directed acyclic graph, and the constraints and computational objectives of the computational task are obtained. Traverse the weighted directed acyclic graph, identify and label the quantum computing task-priority sub-modules and classical computing task-priority sub-modules in the weighted directed acyclic graph, and use the graph partitioning mechanism to divide the unlabeled sub-module set to obtain alternative partitioning schemes; Determine the first and second evaluation coefficients of the target sub-modules in the alternative partitioning schemes for quantum computing tasks and classical computing tasks, respectively. In response to the fact that both the first evaluation coefficient and the second evaluation coefficient are greater than a preset evaluation threshold, the target sub-module is recursively processed to decompose the target sub-module into the collaborative computing task sub-module; Based on the quantum computing task priority submodule, the classical computing task priority submodule, and the collaborative computing task submodule, the task decomposition model is determined; Based on the constraints and the calculation objective, the adjustable coefficients in the task decomposition model are adjusted, and the decomposition result is determined according to the adjusted task decomposition model.

[0010] Optionally, based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and based on the computing power evaluation parameters, a computing task allocation scheme is output, including: Obtain the set of computational tasks from the decomposition results, as well as the real-time status information of the available computing units in the super-fusion architecture; Based on the computing power evaluation model, determine the comprehensive computing power evaluation parameters corresponding to the target computing tasks and target computing units that meet the resource type requirements of the target computing tasks in the computing task set; Based on the comprehensive computing power evaluation parameters, a computing power evaluation matrix is ​​constructed, and optimization objectives and constraints are determined. Based on the computing power evaluation matrix, the optimization objectives, and the constraints, a task allocation optimization model is constructed. The task allocation optimization model is solved to obtain the computation task allocation scheme.

[0011] Optionally, in response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet a preset standard, the relevant parameters are adjusted based on a preset adjustment mechanism until the coherence of the qubits meets the preset standard, including: During the parallel computing operation of the quantum chip cluster, relevant parameters of each quantum chip participating in the computation are periodically collected. Based on the relevant parameters, the coherence index of the qubit is determined, and based on the evaluation index, it is determined whether the coherence index of the qubit meets the preset standard. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet the preset standard, the relevant parameters are adjusted based on the preset adjustment mechanism until the coherence of the qubits meets the preset standard. The preset adjustment mechanism includes at least one of the following operations: adjusting environmental parameters to a preset range, optimizing quantum state parameters and quantum gate operation parameters, and reducing the quantum chip load.

[0012] Optionally, in response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task solution request, including: The computing units that perform parallel computing operations collect raw output data and corresponding execution metadata. The execution metadata includes at least the task completion identifier and health score of each computing unit. The raw data of the quantum computing unit is preprocessed to obtain the expected value and statistical variance of the quantum result. The raw output data of the classical computing unit is read to obtain the classical result and correlate it with the error estimate. Based on the preset computation task result type, the corresponding fusion algorithm is called to perform fusion processing on the parallel computation results to obtain the fusion result; The fusion result is subjected to a consistency check. In response to the successful consistency check, the fusion result is defined as a computation result used to respond to the computation task's solution request.

[0013] Secondly, a quantum chip collaborative computing device based on a super-fusion architecture is provided, the device comprising: The first analysis and processing module is used to construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. The second analysis and processing module is used to respond to the computation task solution request, decompose the computation task corresponding to the computation task solution request based on the task decomposition model, and obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task, wherein the cooperative computation task is decomposed into quantum subcomputation task and classical subcomputation task. The third analysis and processing module is used to determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units based on the computing power evaluation model, and output a computing task allocation scheme based on the computing power evaluation parameters. The fourth analysis and processing module is used to distribute the quantum-first computing task and the quantum computing task in the collaborative computing task to the quantum chip cluster, and distribute the classical-first computing task and the classical sub-computing task in the collaborative computing task to the classical computing power cluster, according to the computing task allocation scheme, and perform parallel computing operations. The fifth analysis and processing module is used to perform fusion processing on the parallel computing results in response to the completion of the parallel computing operation, and generate computing results to respond to the computing task solution request.

[0014] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model to obtain a decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. Based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and a computing task allocation scheme is output based on the computing power evaluation parameters. According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. In response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task's solution request.

[0015] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, performs the following steps: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model to obtain a decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. Based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and a computing task allocation scheme is output based on the computing power evaluation parameters. According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. In response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task's solution request.

[0016] Fifthly, a computer program product is provided, the computer program product comprising a computer program, which, when executed by a processor, performs the following steps: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model to obtain a decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. Based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and a computing task allocation scheme is output based on the computing power evaluation parameters. According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. In response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task's solution request.

[0017] The aforementioned quantum chip collaborative computing method and apparatus based on a super-fusion architecture, the method comprising: constructing a super-fusion architecture, a computational task decomposition model, and a computational power evaluation model, wherein the super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster; responding to a computational task solution request, decomposing the computational task corresponding to the computational task solution request based on the task decomposition model to obtain a decomposition result, wherein the decomposition result includes at least one of the following: a quantum-first computational task, a classical-first computational task, and a collaborative computational task, wherein the collaborative computational task is decomposed into a quantum computing task and a classical sub-computational task; based on the computational power evaluation model, determining the computational power evaluation parameters of each computational task in the decomposition result on different computing units, and outputting a computational task allocation scheme based on the computational power evaluation parameters; and allocating the quantum computing tasks in the quantum-first computational task and the collaborative computational task according to the computational task allocation scheme. The tasks are distributed to the quantum chip cluster, and the classical-priority computing tasks and the classical sub-computing tasks in the collaborative computing tasks are distributed to the classical computing power cluster and parallel computing operations are performed. In response to the completion of the parallel computing operations, the parallel computing results are fused to generate computing results for responding to the computing task solution request. This application breaks through the scale and accuracy bottleneck of a single quantum chip through a super-fusion architecture and collaborative computing model, realizes intelligent task decomposition and dynamic optimal scheduling of multiple quantum chips and classical computing power, deeply couples hardware performance parameters with computing task characteristics, maximizes resource utilization, and improves the success rate of large-scale complex quantum computing tasks, the accuracy of final results and the overall robustness of the system by implementing real-time coherence monitoring and active fault tolerance maintenance throughout the computing cycle and based on trustworthiness-aware dynamic weight fusion, thus providing an efficient and reliable system-level solution for practical quantum computing. Attached Figure Description

[0018] Figure 1 This is an application environment diagram of a quantum chip collaborative computing method based on a super-fusion architecture in one embodiment; Figure 2 This is a flowchart illustrating a quantum chip collaborative computing method based on a hyper-fusion architecture in one embodiment. Figure 3 This is a structural block diagram of a quantum chip collaborative computing device based on a hyper-fusion architecture in one embodiment; Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in 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 in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It should be understood that, in the description of this application, unless the context explicitly requires it, words such as "including" or "comprising" throughout the specification should be interpreted as including rather than exclusive or exhaustive; that is, meaning "including but not limited to".

[0021] It should also be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0022] It should be noted that the terms "S1," "S2," etc., are used only for descriptive purposes and do not specifically refer to the order or sequence, nor are they intended to limit this application. They are merely for the convenience of describing the method of this application and should not be construed as indicating the sequential order of the steps. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0023] The quantum chip collaborative computing method based on hyper-fusion architecture provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with a data processing platform set on server 104 via a network. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. Server 104 can be implemented as a standalone server or a server cluster consisting of multiple servers.

[0024] In one embodiment, such as Figure 2 As shown, a collaborative computing method for quantum chips based on a hyper-fusion architecture is provided, which is then applied to... Figure 1 Taking the terminal in the example, the explanation includes the following steps: S1: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster.

[0025] It should be noted that the hyperscale fusion architecture refers to a system-level computing architecture used to deeply integrate multiple quantum computing units with classical high-performance computing units through software and hardware co-design, achieving hyperscale computing capabilities that surpass the physical limitations of a single chip, and forming a unified computing resource pool through task-level intelligent fusion scheduling and execution; a quantum chip cluster refers to a collection of multiple physically independent quantum processing units, which can be homogeneous (such as multiple superconducting chips) or heterogeneous (such as hybrid superconducting and ion trap chips), interconnected through a network, and can be uniformly scheduled to collaboratively complete computing tasks; a classical computing power cluster refers to a cluster composed of traditional high-performance computing resources (such as CPU servers, GPU / FPGA accelerated computing nodes), which is responsible for executing purely classical computing tasks, providing auxiliary computing such as parameter optimization for quantum computing, and serving as the control center and data aggregation point of the entire fusion architecture.

[0026] S2: In response to the computation task solution request, based on the task decomposition model, the computation task corresponding to the computation task solution request is decomposed to obtain the decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task.

[0027] S3: Based on the computing power evaluation model, determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units, and output the computing task allocation scheme based on the computing power evaluation parameters.

[0028] It should be noted that the computing task allocation scheme refers to a specific execution plan generated by the scheduling system based on the task decomposition results and computing power evaluation results. This scheme clearly specifies which specific physical computing unit (a quantum chip or classical server) each subtask will be assigned to for execution.

[0029] S4: According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed.

[0030] It should be noted that parallel computing operation refers to the process in which quantum chip clusters and classical computing power clusters execute their respective assigned sub-tasks simultaneously (in parallel) according to the allocation scheme. During this process, there may be cross-cluster collaborative operations (such as quantum chips waiting for classical nodes to send parameters).

[0031] S5: In response to the completion of the parallel computing operation, the parallel computing results are fused to generate a computing result for responding to the computing task solution request.

[0032] It should be noted that fusion processing refers to the process of integrating, correcting and merging the results from different computing units, which may be partial or intermediate, according to the task logic and preset algorithms after parallel computing is completed, in order to generate a complete, unified and more accurate final computing result.

[0033] In some specific implementations, constructing a hyperconverged architecture includes: The quantum chip cluster, the classical computing power cluster, and the target control node are deployed. The target control node is used to coordinate the link construction, initiate requests such as clock synchronization calibration and signal noise reduction processing. Construct a heterogeneous interconnect module, which includes a high-performance network switching core (such as using InfiniBand HDR / NDR or custom optoelectronic hybrid switching technology), a tightly coupled quantum control interface, and a classical computing node interface. Based on the heterogeneous interconnection module, the corresponding interface of the target control node is interconnected with the quantum measurement and control port of the quantum chip cluster and the high-speed I / O port of the classical computing power cluster, respectively. Clock synchronization calibration and signal noise reduction are performed on the interconnect links. In response to the completion of the execution, the super-converged architecture is generated. Clock synchronization calibration and signal noise reduction are common methods, and the specific process will not be described in detail here.

[0034] In some specific implementations, constructing a computational task decomposition model includes: The computational task decomposition dimensions and the corresponding quantifiable computational task characteristics are determined. The task decomposition dimensions may include time spent, fidelity / error rate, communication overhead, and fitness. The corresponding quantifiable computational task characteristics may include time characteristics, such as the number of operation gates, gate depth, and time complexity of classical algorithms; fidelity / error rate characteristics, such as the required entanglement degree, sensitivity to coherence time, and error propagation; communication characteristics, such as the amount of input / output data and the required state transition type (such as quantum state tomography and classical parameter transfer); and fitness characteristics, such as computational complexity and accuracy requirements. Based on the benchmark computing task set, the corresponding computing tasks are executed and the computing task execution data is obtained. The benchmark computing task set covers typical quantum algorithms (such as VQE, QAOA, Grover, quantum simulation) and hybrid operations. These tasks and their different decomposition variants are run on the corresponding hardware platforms, such as full quantum execution, full classical simulation, and hybrid execution at different split points. The computing task execution data may include actual execution time, final result fidelity, communication latency and adaptability, etc. Based on the quantifiable computational task characteristics, an initial computational task decomposition function is constructed. Using the computational task execution data, initial weight parameters in the initial computational task decomposition function are determined. An initial computational task pattern knowledge base is then constructed to associate the corresponding computational pattern with its optimal resource preference (quantum / classical / cooperative) on specific hardware. The computational task decomposition function includes: in, Represents the i-th submodule Allocate to computing resources j The cost value, , , and All of these represent dynamic weighting coefficients, which can be cost conversion coefficients. Indicates the execution time. Indicates the error rate. Indicates communication transmission cost, Let represent the fit coefficient, where the formula for calculating the fit coefficient is: , , All are weighting coefficients. LZ = E1 × fzd × YL + E2 × (1 - XQ) + E3 × ys, JD = E5 × fzd × YN + E4 × XQ + E6 × YS, where LZ is the quantum fitness of the task, JD is the classical fitness of the task, E1~E6 are all weighting coefficients, fzd is the computational complexity, YL is the quantum dependency coefficient, XQ is the accuracy requirement, ys is the quantum resource fitness coefficient, YS is the classical resource fitness coefficient, and YN is the classical dependency coefficient. In the actual operation of the super-converged architecture, the system monitors the predicted data of the computation task decomposition results generated based on the current computation task decomposition function, as well as the corresponding actual execution result data of the computation task. That is, in actual operation, the model is deployed as a monitoring agent. For each decomposed and executed task, the system records its predicted decomposition scheme, predicted costs in each dimension, and actual execution results. The predicted data is compared with the actual execution result data to generate feedback data for the initial computation task decomposition function; Based on the feedback data from the initial computational task decomposition function, the weight parameters in the initial computational task decomposition function are dynamically adjusted using an online learning algorithm to obtain the computational task decomposition model. The initial computational task pattern knowledge base is then updated, i.e., the predicted results are periodically compared with the actual results to form feedback data. Using this data, the weight coefficients in the cost function are fine-tuned through online learning algorithms (such as incremental learning and Bayesian updates). At the same time, based on the success or failure of task execution, the associated records in the task pattern knowledge base are updated or strengthened, making the model's understanding of the actual performance of the current hardware system increasingly accurate.

[0035] In some specific implementations, constructing a computing power assessment model includes: Determine the feature set of computing tasks and the feature set of hardware units, run benchmark test programs including various computing task modes, obtain actual performance data of computing tasks executed on target hardware units, and generate training datasets. The feature set of computing tasks may include the number of bits required and the maximum degree of entanglement in quantum computing tasks, and the algorithm complexity in classical tasks. The feature set of hardware units may include static performance indicators and dynamic state indicators, such as the peak computing power, memory bandwidth, chip temperature and available memory of classical CPUs / GPUs. An initial computing power evaluation model is constructed, which includes one or more feature transformation sub-functions and corresponding adjustable weight parameters, wherein the feature transformation sub-functions are used to convert the original features into a unified cost item; Based on the training dataset, the adjustable weight parameters and coefficients in the feature transformation sub-function of the initial computing power evaluation model are adjusted by an optimization algorithm until the initial computing power evaluation model meets the preset standard, thereby obtaining the target computing power evaluation model. The optimization algorithm may include gradient descent, Bayesian optimization, and other algorithms, and the preset standard may be that the evaluation accuracy reaches a preset value. The target computing power evaluation model is deployed to the super-converged architecture. During the operation of the super-converged architecture, computing task allocation schemes and corresponding actual execution effect data are collected. Based on a preset time period, the newly collected data is used to incrementally learn the target computing power evaluation model to update the parameters of the target computing power evaluation model, so as to realize the dynamic update and optimization of the model.

[0036] In some specific implementations, in response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model, and the decomposition results include: The computational task corresponding to the computational task solution request is formalized to obtain a weighted directed acyclic graph, and the constraints and computational objectives of the computational task are obtained. Specifically, the computational task is formalized as a quadruple T=(G,R,C,H), where G=(V,E) represents the computational graph of the task, V represents the computational operation, E represents the data flow dependency, R represents the resource constraints of the computational task, such as the maximum allowed time, the maximum number of qubits, etc., C represents the success rate or accuracy threshold of the computational task, and H represents the metadata of the computational task, which is used to describe the type (such as VQE, QAOA, quantum simulation) and parameters. The weighted directed acyclic graph (DAG) is traversed, identifying and labeling quantum computing task-priority submodules and classical computing task-priority submodules. A graph partitioning mechanism is then used to divide the unlabeled submodule set, yielding alternative partitioning schemes. Specifically, the task computation graph G is traversed, identifying computational patterns such as clearly quantum-suited parts (e.g., large-scale unitary transforms, quantum Fourier transforms), clearly classical-suited parts (e.g., post-processing of large-scale linear equation solving, parameter optimization loops), and gray areas with ambiguous boundaries. For the gray areas and the entire graph, a graph partitioning algorithm (e.g., spectral clustering) is applied. Based on a heuristic cutting algorithm, multiple possible task partitioning schemes are generated. Each scheme divides the vertex set V into multiple subsets {Vc}. For each subset Vc, the cost value of allocating it to quantum resources, classical resources, or splitting it into cooperative tasks is calculated using the above-mentioned task decomposition function. If the difference between the cost value allocated to quantum resources and the cost value allocated to classical resources is large, the allocation scheme with the smaller cost value is defined as the corresponding priority task. If the cost value allocated to quantum resources is significantly lower than the cost value allocated to classical resources, it is defined as a quantum-priority computation task, and so on. The first evaluation coefficient and the second evaluation coefficient of the target sub-module in the alternative partitioning scheme for quantum computing tasks and classical computing tasks are determined respectively, wherein the first evaluation coefficient and the second evaluation coefficient are the cost values ​​mentioned above. In response to the fact that both the first evaluation coefficient and the second evaluation coefficient are greater than the preset evaluation threshold, the target sub-module is recursively processed to decompose the target sub-module into the collaborative computing task sub-module. The preset evaluation threshold can be set according to actual needs. When the internal structure of VC is complex, that is, when allocating the whole to a single resource is too costly, collaborative decomposition is triggered to split the module into smaller quantum computing task modules and classical computing task modules, and the overall cost of collaborative execution is calculated through the task decomposition function mentioned above. Based on the quantum computing task priority submodule, the classical computing task priority submodule, and the collaborative computing task submodule, the task decomposition model is determined; Based on the constraints and the computational objective, the adjustable coefficients in the task decomposition model are adjusted. The decomposition result is determined according to the adjusted task decomposition model. The constraints are resource constraints, and the computational objective is an accuracy objective, namely R and C mentioned above. Specifically, the total cost of all candidate decomposition schemes is compared. Under the premise of satisfying the total constraints R and C, the scheme with the lowest total cost is selected as the final decomposition result. The output includes the decomposition result containing three types of task labels and their corresponding computational graph sub-modules.

[0037] In some specific implementations, based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and based on the computing power evaluation parameters, a computing task allocation scheme is output, including: Obtain the set of computational tasks {JSRW} from the decomposition results. i}, and the real-time status information of the available computing units in the super-converged architecture, which may include basic performance parameters, dynamic coherence parameters, environmental parameters, computing resource utilization, link performance (such as real-time bandwidth, data packet loss rate), etc. Based on the computing power evaluation model, the target computing tasks and target computing units (JSDY) that meet the resource type requirements of the target computing tasks are determined in the computing task set. i The corresponding comprehensive computing power evaluation parameters, wherein the calculation formula is: in, Represents the i-th computational task On the j-th computational unit Comprehensive computing power evaluation parameters For benchmark performance scoring, For the fit coefficient, , and All are weighting coefficients. To implement the delay factor, To ensure mission fidelity, For real-time load rates, this applies to quantum computing tasks. =(N1 / N2)× ,in, 1 represents the baseline performance score of the quantum computing task, N1 is the number of available qubits per unit, and N2 is the number of qubits required for the task. To maintain the average process fidelity of the relevant quantum gates, For the depth of the quantum circuit of the mission, For a classical computing task, the current decoherence time of a key qubit is given. =(F1 / F2)×min(1,BW1 / BW2), where, The benchmark performance scores for classic computing tasks are as follows: F1 represents the peak floating-point computing power per unit, F2 represents the estimated floating-point operations required by the task, BW1 represents the available memory bandwidth per unit, and BW2 represents the estimated memory bandwidth requirement of the task. =1+a1M+a2U, where a1 and a2 are weighting coefficients, M is the architecture feature matching gain value, and U is the architecture mismatch penalty value; Based on the comprehensive computing power evaluation parameters, a computing power evaluation matrix is ​​constructed, and optimization objectives and constraints are determined. Based on the computing power evaluation matrix, the optimization objectives, and the constraints, a task allocation optimization model is constructed. Specifically, for each task-unit pair, the comprehensive computing power evaluation parameters are... This forms an m×n evaluation matrix Z, where m is the number of computational tasks and n is the number of units. The optimization objective is... , As decision variables, =1 indicates that task i is assigned to unit j. Constraints may include that each task can be assigned to at most one unit, and that the task load of each unit does not exceed its capacity. Solving the task allocation optimization model yields the computational task allocation scheme. Specifically, a mixed-integer programming solver, a heuristic algorithm (such as a genetic algorithm), or a priority-based greedy algorithm (selecting tasks at each step while satisfying constraints) is used. The optimization problem is solved by allocating the highest feasible task (unit pair) to the highest feasible task, and the final output is a specific computational task allocation scheme that explicitly lists each unit pair. Which specific destination should it be scheduled to? Execute above.

[0038] In some specific embodiments, before fusing the computation results and generating the final computation result in response to the completion of the parallel computing operation, the method further includes: During the parallel computing operation, relevant parameters of the quantum chip cluster are collected. These relevant parameters include quantum state parameters, environmental parameters, and performance parameters. Quantum state parameters may include quantum state purity, specific Bell state fidelity, etc. Environmental parameters may include dilution refrigerator temperature, residual magnetic field strength, microwave noise spectral density, etc. Performance parameters may include the real-time T1 (energy relaxation time), T2 (dephase time), and single / double qubit gate process fidelity of the qubits. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet a preset standard, the relevant parameters are adjusted based on a preset adjustment mechanism until the coherence of the qubits meets the preset standard, including: During the parallel computing operation of the quantum chip cluster, relevant parameters of each quantum chip participating in the computation are periodically collected. Based on the relevant parameters, the coherence index of the qubit is determined, and based on the evaluation index, it is determined whether the coherence index of the qubit meets the preset standard. The method for determining the coherence index of the qubit is: HEA = min(Q1 / Q2, T2 / R2, P3), where HEA is the coherence evaluation value, Q1 is the currently measured energy relaxation time, Q2 is the nominal or factory reference value of the qubit, T2 is the currently measured dephase time, R2 is the corresponding nominal reference value, and P3 is the average quantum gate process fidelity. When HEA is less than 1 and / or any key parameter is lower than the safety threshold, such as the temperature being lower than the safety threshold, it is judged to not meet the preset standard. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet the preset standard, the relevant parameters are adjusted based on the preset adjustment mechanism until the coherence of the qubits meets the preset standard. The preset adjustment mechanism includes at least one of the following operations: adjusting the environmental parameters to a preset range, optimizing the quantum state parameters and quantum gate operation parameters, and reducing the quantum chip load. That is, if the temperature is lower than the safety threshold, the environmental parameters are adjusted, and so on.

[0039] In some specific implementations, in response to the completion of parallel computing operations, the parallel computing results are fused to generate computing results for responding to the computing task solution request, including: The computing units that perform parallel computing operations collect raw output data and corresponding execution metadata. The execution metadata includes at least the task completion identifier and health score of each computing unit. The raw data of the quantum computing unit is preprocessed to obtain the expected value and statistical variance of the quantum result. The raw output data of the classical computing unit is read to obtain the classical result and correlate it with the error estimate. Based on the preset computation task result type, the corresponding fusion algorithm is invoked to fuse the parallel computation results, resulting in a fused result. The fusion algorithm is as follows: in, This represents the sub-calculation result returned by the i-th calculation unit. This represents the total number of parallel computing units. Indicates the fusion result. This represents the overall health score of the i-th computing unit when performing a computing task. This represents the overall health score of the j-th computing unit when performing a computing task. It is obtained by normalizing and weighting the real-time state information of the computing unit (such as quantum gate fidelity, coherence time retention rate of the quantum chip, temperature and noise stability, etc.). This represents the statistical variance of the i-th sub-result. This represents the statistical variance of the j-th sub-outcome. This indicates the uncertainty of the fusion result and is used for quantification. The smaller the overall statistical error range, the higher the confidence level of the fusion result; The fusion result is subjected to consistency verification, that is, the difference between the results of different weight schemes or different subsets is compared. If the difference exceeds the preset tolerance, an alarm is triggered or redundant verification calculation is started. In response to the consistency verification passing, the fusion result is defined as the calculation result used to respond to the calculation task solution request. That is, the fusion result and the uncertainty metric of the fusion result are encapsulated together to generate the final calculation result.

[0040] The aforementioned quantum chip collaborative computing method based on a super-fusion architecture includes: constructing a super-fusion architecture, a computational task decomposition model, and a computational power evaluation model, wherein the super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster; responding to a computational task solution request, decomposing the computational task corresponding to the computational task solution request based on the task decomposition model to obtain a decomposition result, wherein the decomposition result includes at least one of the following: a quantum-first computational task, a classical-first computational task, and a collaborative computational task, wherein the collaborative computational task is decomposed into a quantum computing task and a classical sub-computational task; based on the computational power evaluation model, determining the computational power evaluation parameters of each computational task in the decomposition result on different computing units, and outputting a computational task allocation scheme based on the computational power evaluation parameters; and distributing the quantum-first computational task and the quantum computing task in the collaborative computational task according to the computational task allocation scheme. The quantum chip cluster distributes the classical priority computing task and the classical sub-computing tasks in the collaborative computing task to the classical computing power cluster and performs parallel computing operations. In response to the completion of the parallel computing operations, the parallel computing results are fused to generate computing results for responding to the computing task's solution request. This application breaks through the scale and accuracy bottleneck of a single quantum chip through a super-fusion architecture and collaborative computing model, realizing intelligent task decomposition and dynamic optimal scheduling of multiple quantum chips and classical computing power. It deeply couples hardware performance parameters with computing task characteristics, maximizes resource utilization, and improves the success rate of large-scale complex quantum computing tasks, the accuracy of final results, and the overall robustness of the system by implementing real-time coherence monitoring and active fault tolerance maintenance throughout the computing cycle and based on trustworthiness-aware dynamic weight fusion. This provides an efficient and reliable system-level solution for practical quantum computing.

[0041] It should be understood that, although Figure 2 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 2 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0042] In one embodiment, such as Figure 3 As shown, a quantum chip collaborative computing device based on a hyper-fusion architecture is provided, comprising: The first analysis and processing module is used to construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. The second analysis and processing module is used to respond to the computation task solution request, decompose the computation task corresponding to the computation task solution request based on the task decomposition model, and obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task, wherein the cooperative computation task is decomposed into quantum subcomputation task and classical subcomputation task. The third analysis and processing module is used to determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units based on the computing power evaluation model, and output a computing task allocation scheme based on the computing power evaluation parameters. The fourth analysis and processing module is used to distribute the quantum-first computing task and the quantum computing task in the collaborative computing task to the quantum chip cluster, and distribute the classical-first computing task and the classical sub-computing task in the collaborative computing task to the classical computing power cluster, according to the computing task allocation scheme, and perform parallel computing operations. The fifth analysis and processing module is used to perform fusion processing on the parallel computing results in response to the completion of the parallel computing operation, and generate computing results to respond to the computing task solution request.

[0043] Specific limitations regarding the quantum chip collaborative computing device based on the hyper-fusion architecture can be found in the limitations of the quantum chip collaborative computing method based on the hyper-fusion architecture mentioned above, and will not be repeated here. Each module in the aforementioned quantum chip collaborative computing device based on the hyper-fusion architecture can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0044] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a quantum chip collaborative computing method based on a superscalar fusion architecture. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0045] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0046] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: S1: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. S2: In response to the computation task solving request, based on the task decomposition model, the computation task corresponding to the computation task solving request is decomposed to obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. S3: Based on the computing power evaluation model, determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units, and output the computing task allocation scheme based on the computing power evaluation parameters; S4: According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. S5: In response to the completion of the parallel computing operation, the parallel computing results are fused to generate a computing result for responding to the computing task solution request.

[0047] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: S1: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. S2: In response to the computation task solving request, based on the task decomposition model, the computation task corresponding to the computation task solving request is decomposed to obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. S3: Based on the computing power evaluation model, determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units, and output the computing task allocation scheme based on the computing power evaluation parameters; S4: According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. S5: In response to the completion of the parallel computing operation, the parallel computing results are fused to generate a computing result for responding to the computing task solution request.

[0048] In one embodiment, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, performs the following steps: S1: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. S2: In response to the computation task solving request, based on the task decomposition model, the computation task corresponding to the computation task solving request is decomposed to obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. S3: Based on the computing power evaluation model, determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units, and output the computing task allocation scheme based on the computing power evaluation parameters; S4: According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. S5: In response to the completion of the parallel computing operation, the parallel computing results are fused to generate a computing result for responding to the computing task solution request.

[0049] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0050] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0051] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application.

Claims

1. A quantum chip collaborative computing method based on a hyper-fusion architecture, characterized in that, The method includes: Construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model to obtain a decomposition result. The decomposition result includes at least one of the following: quantum-first computation task, classical-first computation task, and cooperative computation task. The cooperative computation task is decomposed into quantum computing task and classical sub-computation task. Based on the computing power evaluation model, the computing power evaluation parameters of each computing task in the decomposition results on different computing units are determined, and a computing task allocation scheme is output based on the computing power evaluation parameters. According to the computing task allocation scheme, the quantum priority computing task and the quantum sub-computing task in the collaborative computing task are distributed to the quantum chip cluster, and the classical priority computing task and the classical sub-computing task in the collaborative computing task are distributed to the classical computing power cluster, and parallel computing operations are performed. In response to the completion of the parallel computing operation, the parallel computing results are fused to generate computing results for responding to the computing task's solution request.

2. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 1, characterized in that, Before fusing the computation results and generating the final computation result in response to the completion of the parallel computing operation, the method further includes: During the parallel computing operation, relevant parameters of the quantum chip cluster are collected, including quantum state parameters, environmental parameters, and performance parameters. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet the preset standard, the relevant parameters are adjusted based on the preset adjustment mechanism until the coherence of the qubits meets the preset standard.

3. The quantum chip collaborative computing method based on a hyper-fusion architecture according to claim 2, characterized in that, Building a hyperconverged architecture includes: Deploy the quantum chip cluster, the classical computing power cluster, and the target control node; Construct heterogeneous interconnect modules; Based on the heterogeneous interconnection module, the corresponding interface of the target control node is interconnected with the quantum measurement and control port of the quantum chip cluster and the high-speed I / O port of the classical computing power cluster, respectively. Clock synchronization calibration and signal noise reduction are performed on the interconnect links, and the super-converged architecture is generated upon completion of the execution.

4. The quantum chip collaborative computing method based on a hyper-fusion architecture according to claim 3, characterized in that, Constructing a computational task decomposition model includes: Determine the dimensions for decomposing computational tasks, and the quantifiable computational task features corresponding to those dimensions. Based on the benchmark test computation task set, execute the corresponding computation tasks and obtain the computation task execution data; Based on the quantifiable computational task characteristics, an initial computational task decomposition function is constructed, and the initial weight parameters in the initial computational task decomposition function are determined using the computational task execution data, and an initial computational task pattern knowledge base is constructed. In the actual operation of the super-fusion architecture, the predicted data of the computation task decomposition results generated based on the current computation task decomposition function, as well as the corresponding actual execution result data of the computation task, are monitored. The predicted data is compared with the actual execution result data to generate feedback data for the initial computation task decomposition function; Based on the feedback data of the initial computation task decomposition function, the weight parameters in the initial computation task decomposition function are dynamically adjusted using an online learning algorithm to obtain the computation task decomposition model, and the initial computation task pattern knowledge base is updated.

5. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 4, characterized in that, The construction of the computing power assessment model includes: Determine the feature set of computing tasks and the feature set of hardware units, run a benchmark test program that includes multiple computing task modes, obtain actual performance data of executing computing tasks on the target hardware unit, and generate a training dataset; Construct an initial computing power evaluation model, which includes one or more feature transformation sub-functions and corresponding adjustable weight parameters; Based on the training dataset, the adjustable weight parameters and coefficients in the feature transformation sub-function of the initial computing power evaluation model are adjusted by an optimization algorithm until the initial computing power evaluation model meets the preset standard, thus obtaining the target computing power evaluation model. The target computing power evaluation model is deployed to the super-converged architecture. During the operation of the super-converged architecture, computing task allocation schemes and corresponding actual execution effect data are collected. Based on a preset time period, the newly collected data is used to incrementally learn the target computing power evaluation model in order to update the parameters of the target computing power evaluation model.

6. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 5, characterized in that, In response to a computation task solution request, the computation task corresponding to the computation task solution request is decomposed based on the task decomposition model, and the decomposition results include: The computational task corresponding to the computational task solution request is formalized to obtain a weighted directed acyclic graph, and the constraints and computational objectives of the computational task are obtained. Traverse the weighted directed acyclic graph, identify and label the quantum computing task-priority sub-modules and classical computing task-priority sub-modules in the weighted directed acyclic graph, and use the graph partitioning mechanism to divide the unlabeled sub-module set to obtain alternative partitioning schemes; Determine the first and second evaluation coefficients of the target sub-modules in the alternative partitioning schemes for quantum computing tasks and classical computing tasks, respectively. In response to the fact that both the first evaluation coefficient and the second evaluation coefficient are greater than a preset evaluation threshold, the target sub-module is recursively processed to decompose the target sub-module into the collaborative computing task sub-module; Based on the quantum computing task priority submodule, the classical computing task priority submodule, and the collaborative computing task submodule, the task decomposition model is determined; Based on the constraints and the calculation objective, the adjustable coefficients in the task decomposition model are adjusted, and the decomposition result is determined according to the adjusted task decomposition model.

7. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 6, characterized in that, Based on the aforementioned computing power evaluation model, the computing power evaluation parameters for each computing task in the decomposition results on different computing units are determined. Based on the computing power evaluation parameters, a computing task allocation scheme is output, including: Obtain the set of computational tasks from the decomposition results, as well as the real-time status information of the available computing units in the super-fusion architecture; Based on the computing power evaluation model, determine the comprehensive computing power evaluation parameters corresponding to the target computing tasks and target computing units that meet the resource type requirements of the target computing tasks in the computing task set; Based on the comprehensive computing power evaluation parameters, a computing power evaluation matrix is ​​constructed, and optimization objectives and constraints are determined. Based on the computing power evaluation matrix, the optimization objectives, and the constraints, a task allocation optimization model is constructed. The task allocation optimization model is solved to obtain the computation task allocation scheme.

8. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 7, characterized in that, In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet a preset standard, the relevant parameters are adjusted based on a preset adjustment mechanism until the coherence of the qubits meets the preset standard, including: During the parallel computing operation of the quantum chip cluster, relevant parameters of each quantum chip participating in the computation are periodically collected. Based on the relevant parameters, the coherence index of the qubit is determined, and based on the evaluation index, it is determined whether the coherence index of the qubit meets the preset standard. In response to the fact that the coherence of the qubits corresponding to the quantum chip cluster does not meet the preset standard, the relevant parameters are adjusted based on the preset adjustment mechanism until the coherence of the qubits meets the preset standard. The preset adjustment mechanism includes at least one of the following operations: adjusting environmental parameters to a preset range, optimizing quantum state parameters and quantum gate operation parameters, and reducing the quantum chip load.

9. The quantum chip collaborative computing method based on a super-fusion architecture according to claim 8, characterized in that, In response to the completion of parallel computing operations, the parallel computing results are fused to generate computing results for responding to the computing task's solution request, including: The computing units that perform parallel computing operations collect raw output data and corresponding execution metadata. The execution metadata includes at least the task completion identifier and health score of each computing unit. The raw data of the quantum computing unit is preprocessed to obtain the expected value and statistical variance of the quantum result. The raw output data of the classical computing unit is read to obtain the classical result and correlate it with the error estimate. Based on the preset computation task result type, the corresponding fusion algorithm is called to perform fusion processing on the parallel computation results to obtain the fusion result; The fusion result is subjected to a consistency check. In response to the successful consistency check, the fusion result is defined as a computation result used to respond to the computation task's solution request.

10. A quantum chip collaborative computing device based on a hyper-fusion architecture, characterized in that, The device includes: The first analysis and processing module is used to construct a super-fusion architecture, a computing task decomposition model, and a computing power evaluation model. The super-fusion architecture includes at least a quantum chip cluster and a classical computing power cluster. The second analysis and processing module is used to respond to the computation task solution request, decompose the computation task corresponding to the computation task solution request based on the task decomposition model, and obtain the decomposition result. The decomposition result includes at least one of the following: quantum priority computation task, classical priority computation task and cooperative computation task, wherein the cooperative computation task is decomposed into quantum subcomputation task and classical subcomputation task. The third analysis and processing module is used to determine the computing power evaluation parameters of each computing task in the decomposition results on different computing units based on the computing power evaluation model, and output a computing task allocation scheme based on the computing power evaluation parameters. The fourth analysis and processing module is used to distribute the quantum-first computing task and the quantum computing task in the collaborative computing task to the quantum chip cluster, and distribute the classical-first computing task and the classical sub-computing task in the collaborative computing task to the classical computing power cluster, according to the computing task allocation scheme, and perform parallel computing operations. The fifth analysis and processing module is used to perform fusion processing on the parallel computing results in response to the completion of the parallel computing operation, and generate computing results to respond to the computing task solution request.

Images