Hybrid computing workflow platform based on directed acyclic graph and terminal equipment

Abstract

The invention discloses a hybrid computing workflow platform and terminal equipment based on a directed acyclic graph, and the hybrid computing workflow platform based on the directed acyclic graph comprises a graphical workflow construction module, a workflow calculation module and a workflow calculation module, the system comprises a task scheduling execution engine module, a resource awareness and distribution mechanism module, a node plug-in interface management module, a calculation result feedback mechanism module, a template multiplexing and version control module, a quantum-classical calculation interface fusion module and an AI driving process optimization module. According to the workflow platform, multi-level algorithm integration from atomic-scale quantum chemistry modeling to molecular-scale classical dynamics simulation and then to machine learning prediction can be realized, and flexible scheduling of quantum calculation, material simulation, MD, QMC, AI and the like on a unified platform is supported. The problems of task splitting and low collaboration efficiency in an existing calculation process are solved.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a hybrid computing workflow platform based on directed acyclic graphs and a terminal device. Background Technology

[0002] With the increasing application of cross-scale scientific computing in materials design, drug development, and quantum-classical hybrid simulations, researchers urgently need to integrate molecular dynamics, quantum chemical calculations, artificial intelligence models, and quantum computing resources. However, the workflow platforms in these technologies suffer from the following shortcomings: 1. In current research activities, multiple computational steps are typically handled by different models or programs. For example, obtaining molecular structures from classical simulations, then performing electronic structure calculations or quantum algorithm solutions at higher precision, and finally analyzing various properties. The data interfaces between these steps differ significantly, conversions are cumbersome, and manual intervention is often required, increasing operational complexity and the probability of errors, making large-scale automation and process reproducibility difficult. 2. Various computational models and algorithm tools typically rely on scripts or command-line interfaces, requiring users to have certain programming skills, mathematical modeling knowledge, and experience in configuring high-performance computing resources. Researchers without a professional background find it difficult to quickly build cross-scale, cross-algorithm joint computational workflows, hindering the widespread application of computational technologies and multidisciplinary integration. 3. Different computational tasks have significantly different resource requirements, some relying on general-purpose computing processors, while others are highly dependent on graphics acceleration or quantum computing resources. Currently, common systems struggle to automatically and intelligently match and schedule computing resources based on task characteristics, leading to uneven resource allocation or frequent system bottlenecks that hinder efficient task execution. 4. In high-throughput screening and data-assisted design tasks, it is often necessary to continuously adjust input parameters or update models based on intermediate computation results. Existing systems lack task monitoring and feedback update mechanisms, making it difficult to support adaptive learning and continuous optimization, thus affecting model performance. 5. Scientific computing workflows often exhibit high generality, such as stages like structure processing, model building, computation execution, and result analysis. However, existing platforms lack template-based management mechanisms, making it difficult to easily save, migrate, share, or reuse existing workflow platforms, which is detrimental to team collaboration and knowledge accumulation. Summary of the Invention

[0003] The present invention is proposed to address at least one of the aforementioned problems. According to a first aspect of the invention, a hybrid computing workflow platform based on a directed acyclic graph is provided, the workflow platform comprising: a graphical workflow construction module, a task scheduling and execution engine module, a resource awareness and allocation mechanism module, a node plugin interface management module, a computation result feedback mechanism module, a template reuse and version control module, a quantum-classical computing interface fusion module, and an AI-driven process optimization module.

[0004] The graphical workflow building module is used to provide a drag-and-drop visual interface and generate a structured workflow template based on the user-input visual workflow arrangement. The workflow template represents a directed acyclic graph formed by task nodes and their dependencies.

[0005] The task scheduling and execution engine module is used to parse the workflow template, construct a directed acyclic graph (DAG) object and its dependencies, determine the task execution order and identify parallelizable tasks based on the DAG object and its dependencies using a topological sorting algorithm, and schedule tasks in a sequential and / or parallel manner.

[0006] The resource awareness and allocation mechanism module is used to monitor the resource status of heterogeneous computing resources in real time, and allocate computing resources using a multi-dimensional weighted algorithm based on the resource status and the computing type of the nodes. The heterogeneous computing resources include at least two of the following: CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), and quantum simulator. The multi-dimensional factors include: task priority weight, model type, and data size.

[0007] The node plugin interface management module is used to define a unified task node interface standard, and supports command line parameter configuration, input / output file path management and parameter verification.

[0008] The calculation result feedback mechanism module is used to automatically collect task execution results and write them into the database to achieve a closed loop of data flow.

[0009] The template reuse and version control module is used to save and share standard process templates, and supports version rollback, parameter overwrite and data migration.

[0010] The quantum-classical computing interface fusion module is used to interface with quantum computing resources and coordinate the scheduling and execution of quantum computing resources and classical computing tasks.

[0011] The AI-driven process optimization module is used to perform intelligent workflow recommendation, parameter optimization, node priority evaluation, and result error prompts based on historical workflow data.

[0012] In one embodiment of the present invention, the visual workflow orchestration includes: node dragging, parameter filling, and node connection; the graphical workflow construction module is specifically used to: detect in real time whether a loop is formed when the user drags the node, fills in the parameter, and connects the node through the drag-and-drop visual interface; if a loop is detected, highlight the error and prevent saving, so as to ensure that the generated structured workflow template conforms to the directed acyclic graph structure.

[0013] In one embodiment of the present invention, the task scheduling and execution engine module further includes a fault tolerance mechanism unit, a task retry strategy unit, a runtime environment management unit, and a log management unit; the fault tolerance mechanism unit is used to perform multiple backups on tasks marked as critical tasks; the task retry strategy unit is used to automatically retry tasks up to a predetermined number of times when a transient failure occurs; the runtime environment management unit is used to automatically switch the task runtime environment according to the computing resource status; and the log management unit is used to collect and store all task logs in real time.

[0014] In one embodiment of the present invention, when the resource awareness and allocation mechanism module uses a multi-dimensional weighted algorithm to allocate computing resources, it is specifically used to: implement a strategy of prioritizing the allocation of computing resources to CPUs with low-latency connections for quantum algorithm tasks; implement a strategy of prioritizing the allocation of big data tasks to local storage nodes; and automatically degrade non-critical tasks when resources are overloaded.

[0015] In one embodiment of the present invention, when the calculation result feedback mechanism module automatically collects task results and writes them into the database, it is specifically used to: automatically trigger retry, skip or switch to different execution path operations when the task results meet preset conditions, automatically convert the result data into a standard format and write it into the database for subsequent nodes to call, and provide data visualization charts.

[0016] In one embodiment of the present invention, the template reuse and version control module is further used to set the visibility scope of the template and assign role permissions; wherein, the visibility scope includes: private, team and organization-wide levels, and the roles include: viewer, editor and administrator.

[0017] In one embodiment of the present invention, the quantum computing resources include: a local quantum simulator, a remote quantum cloud platform, and an actual quantum processor.

[0018] In one embodiment of the present invention, when the AI-driven process optimization module performs intelligent workflow recommendation, parameter optimization, node priority evaluation, and result error prompting based on historical workflow data, it is specifically used to: employ an active learning mechanism to train a model using the historical workflow data, provide users with intelligent workflow recommendation, node priority evaluation, and parameter optimization suggestions, and provide error prompts when the task result deviates from the expected threshold; wherein, the model includes a collaborative filtering model for workflow recommendation, a regression model for node priority evaluation and parameter optimization, and a time series model for result error prediction.

[0019] According to a second aspect of the present invention, a terminal device is provided, comprising: a hybrid computing workflow platform based on any of the above-described directed acyclic graphs.

[0020] According to the embodiments of the present invention, the hybrid computing workflow platform and terminal device based on directed acyclic graphs have the following beneficial effects:

[0021] 1. This invention enables the integration of multi-level algorithms, from atomic-scale quantum chemical modeling to molecular-scale classical dynamics simulation and machine learning prediction. It supports flexible scheduling of quantum computing, materials simulation, molecular dynamics (MD), quantum Monte Carlo (QMC), and artificial intelligence (AI) on a unified platform. Its core capability lies in connecting previously dispersed multi-physics computing modules into an end-to-end computing flow through a directed graph, solving the problems of task fragmentation and inefficient collaboration in existing computing processes.

[0022] 2. This invention enables users without a programming background to quickly build complex cross-scale workflows through graphical drag-and-drop node configuration and visual management of task dependencies, thus lowering the barrier to entry for high-performance computing and quantum algorithms.

[0023] 3. The resource awareness and allocation mechanism module can detect the load status of various types of nodes such as CPU, GPU, FPGA, and quantum simulator in real time, and automatically allocate the best computing resources according to task characteristics, which can effectively improve resource utilization and scheduling intelligence.

[0024] 4. This invention can embed quantum computing tasks as optional node modules into the overall process, supporting variable quantum algorithms and allowing them to be executed in parallel or cascaded with classical tasks. Users can insert quantum nodes as needed and connect to local simulators or cloud-based quantum computing platforms through a unified interface. This not only fully demonstrates the accuracy advantages of quantum computing in specific scenarios but also ensures the configurability and robustness of the overall process.

[0025] 5. This invention supports outputting the prediction results of computational nodes (such as band structure, optical spectrum, structural configuration, electron density, and charge transfer) to external experimental modules (such as material characterization platforms and molecular synthesis laboratories), realizing a closed-loop optimization process of "computational prediction – experimental verification – data feedback – model update". This is conducive to forming a continuously evolving high-precision model system and improving the reliability and value of theoretical simulation in guiding experiments.

[0026] 6. This invention also supports users in saving any completed workflow as a template. The template embeds the configuration of computing nodes, dependencies, and parameter settings, facilitating rapid reuse and inheritance in subsequent projects. The template can be imported and exported, supporting internal team collaboration, cross-project migration, and reproduction, which significantly helps in building standardized computing processes and improving collaborative development efficiency.

[0027] 7. This application can automatically record the input and output data, operation logs, and error messages of all computing nodes, and bind them to user accounts to form a complete task history. This not only facilitates problem tracking and reproduction, but also provides crucial support for the archiving, auditing, and knowledge accumulation of scientific research results. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the structure of a hybrid computing workflow platform based on a directed acyclic graph provided in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the visualization interface of a hybrid computing workflow platform based on a directed acyclic graph provided in an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram of the structure of a terminal device provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0033] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0034] It should be understood that the invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0035] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0036] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0037] The first aspect of this invention provides a hybrid computing workflow platform based on directed acyclic graphs, such as... Figure 1 As shown, the hybrid computing workflow platform 10 based on directed acyclic graphs includes: a graphical workflow building module 100, a task scheduling and execution engine module 200, a resource awareness and allocation mechanism module 300, a node plugin interface management module 400, a computation result feedback mechanism module 500, a template reuse and version control module 600, a quantum-classical computing interface fusion module 700, and an AI-driven process optimization module 800.

[0038] The graphical workflow building module 100 provides a drag-and-drop visual interface and generates a structured workflow template based on the user-input visual workflow arrangement. The workflow template represents a directed acyclic graph (DAG) formed by task nodes and their dependencies.

[0039] It should be noted that the graphical workflow building module 100 serves as the front-end entry point for user interaction with the platform, enabling non-professional users to build, debug, and optimize complex computational processes without programming. Furthermore, using a directed acyclic graph (DAG) as the foundational structure for workflow modeling ensures the acyclic nature and parallelizability of task dependencies.

[0040] As an example, the graphical workflow building module 100 also supports viewing suggestions for module integration and has convenient functions such as search, highlighting, and parameter template prompts.

[0041] The task scheduling and execution engine module 200 is used to parse workflow templates, construct directed acyclic graph objects and their dependencies, determine the task execution order and identify parallelizable tasks based on the directed acyclic graph objects and their dependencies and through a topological sorting algorithm, and schedule tasks in a sequential and / or parallel manner.

[0042] It should be noted that the task scheduling and execution engine module 200 is the core module of the platform, which transforms workflow templates into actual execution sequences.

[0043] As an example, the task scheduling execution engine module 200 can first parse the workflow template and construct a DAG object in memory. Then, it can determine the task execution order and identify groups of tasks that can be parallelized by using a topology sorting algorithm.

[0044] The resource awareness and allocation mechanism module 300 is used to monitor the resource status of heterogeneous computing resources in real time, and allocate computing resources according to the resource status and the computing type of the node using a multi-dimensional weighted algorithm. The heterogeneous computing resources include at least two of CPU, GPU, FPGA and quantum simulator. The multi-dimensional factors include: task priority weight, model type and data size.

[0045] As an example, the resource awareness and allocation mechanism module 300 monitors CPU utilization, GPU memory usage, FPGA logic unit utilization, quantum device queue length, etc. in real time.

[0046] It should be noted that the priority weights can be determined by the user-defined task importance, and the model types include commonly used computational models.

[0047] The node plugin interface management module 400 is used to define a unified task node interface standard, supporting command line parameter configuration, input / output file path management, and parameter validation.

[0048] As an example, the interface standard includes: unified input / output management, parameter configuration, and lifecycle hook interfaces.

[0049] It should be noted that all task nodes are designed based on a unified interface and connect to local simulators or cloud-based quantum computing platforms through this unified interface. This supports command-line parameter configuration, input / output file path management, and parameter verification, enabling rapid integration of various software packages and algorithms. It supports universal JSON / YAML configuration descriptors, allowing users to quickly build new nodes based on the SDK (Software Development Kit), thus improving the platform's scalability.

[0050] The calculation result feedback mechanism module 500 is used to automatically collect task execution results and write them into the database, realizing a closed loop of data flow.

[0051] It should be noted that the calculation result feedback mechanism module 500 can construct a data loop for the workflow, used to capture, transmit, and utilize task results. Specifically, after the task is completed, the output directory can be automatically scanned, and the result files can be uploaded to the central database.

[0052] The template reuse and version control module 600 is used to save and share standard process templates, and supports version rollback, parameter overwrite and data migration.

[0053] As an example, it allows users to save and share standard process templates, supporting version rollback, parameter overriding, and rapid migration, thus improving work efficiency and reusability. Simultaneously, users can manage the scope of template sharing through the permission system, enabling organizational collaboration and task module consolidation.

[0054] Specifically, the template reuse and version control module 600 is also used to set the visibility scope of templates and assign role permissions; the visibility scope includes: private, team and organization-wide levels, and the roles include: viewer, editor and administrator.

[0055] The quantum-classical computing interface fusion module 700 is used to interface with quantum computing resources and coordinate the scheduling and execution of quantum computing resources and classical computing tasks.

[0056] Specifically, quantum computing resources include: local quantum simulators, remote quantum cloud platforms, and actual quantum processors.

[0057] The AI-driven process optimization module 800 is used to perform intelligent workflow recommendations, parameter optimization, node priority evaluation, and result error prompts based on historical workflow data.

[0058] It should be noted that intelligent workflow recommendations can be generated based on users' historical behavior and task objectives.

[0059] As an example, for parameter-sensitive tasks, Bayesian optimization can be used to automatically search for the optimal value, reducing manual trial and error.

[0060] It is worth mentioning that this application supports a variety of variational quantum algorithms, such as VQE (Variational Quantum Eigensolver), UCCSD (Unitary Coupled Cluster Singles and Doubles), and QSCI (Quantum Selected Configuration Interaction). These variational quantum algorithms can be executed in parallel or cascaded with classical tasks.

[0061] The hybrid computing workflow platform 10 based on directed acyclic graphs in this invention has the following advantages: 1. This invention can achieve multi-level algorithm integration from atomic-scale quantum chemical modeling to molecular-scale classical dynamics simulation and machine learning prediction, supporting flexible scheduling of quantum computing, materials simulation, MD, QMC, AI, etc. on a unified platform. Its core capability lies in connecting the originally scattered multi-physics computing modules into an end-to-end computing process through a directed graph, solving the problems of task fragmentation and inefficient collaboration in existing computing processes. 2. This invention enables users without programming backgrounds to quickly build complex cross-scale workflows through graphical drag-and-drop node configuration and task dependency visualization management, lowering the barrier to entry for high-performance computing and quantum algorithms. 3. The resource awareness and allocation mechanism module 300 can detect the load status of multiple types of nodes such as CPU, GPU, FPGA, and quantum simulator in real time, and automatically allocate the best computing resources according to task characteristics, effectively improving resource utilization and scheduling intelligence. 4. This invention can embed quantum computing tasks as optional node modules into the overall process, supporting variable quantum algorithms and executing them in parallel or cascaded with classical tasks. Users can insert quantum nodes as needed and connect to local simulators or cloud-based quantum computing platforms through a unified interface. This not only fully demonstrates the precision advantages of quantum computing in specific scenarios but also ensures the configurability and robustness of the overall process. 5. This invention supports outputting the prediction results of computing nodes (such as band structure, optical spectrum, structural configuration, electron density, and charge transfer) to external experimental modules (such as material characterization platforms and molecular synthesis laboratories), realizing a closed-loop optimization process of "computational prediction – experimental verification – data feedback – model update." This is beneficial for forming a continuously evolving high-precision model system and improving the reliability and value of theoretical simulations in guiding experiments. 6. This invention also allows users to save any completed workflow as a template. The template embeds computing node configurations, dependencies, and parameter settings, facilitating rapid reuse and inheritance in subsequent projects. Templates can be imported and exported, supporting internal team collaboration, cross-project migration, and reproduction, significantly helping to build standardized computing processes and improve collaborative development efficiency. 7. This invention can automatically record the input and output data, running logs, and error messages of all computing nodes and bind them to user accounts, forming a complete task history. It not only facilitates problem tracking and reproduction, but also provides crucial support for the archiving, auditing, and knowledge accumulation of scientific research results.

[0062] In some embodiments, the visual workflow orchestration includes: node dragging, parameter filling, and node connection; the graphical workflow building module 100 is specifically used to: detect in real time whether a loop is formed when the user drags, fills in parameters, and connects nodes through the drag-and-drop visual interface; if a loop is detected, it highlights the error and prevents saving, so as to ensure that the generated structured workflow template conforms to the directed acyclic graph structure.

[0063] As an example, users can add, drag, and connect nodes using the mouse, and the platform can automatically verify the acyclicity of the directed acyclic graph. If a cycle is found, it will be highlighted as an error. Users can also click on each node to bring up a parameter panel where they can fill in the parameters.

[0064] In this embodiment, non-professional users can generate relevant structured workflow templates without coding, thus lowering the barrier to entry.

[0065] In some embodiments, the task scheduling execution engine module 200 further includes a fault tolerance mechanism unit, a task retry strategy unit, a runtime environment management unit, and a log management unit; the fault tolerance mechanism unit is used to perform multiple backups on tasks marked as critical tasks; the task retry strategy unit is used to automatically retry tasks up to a predetermined number of times when a transient failure occurs; the runtime environment management unit is used to automatically switch the task runtime environment according to the computing resource status; and the log management unit is used to collect and store all task logs in real time.

[0066] As an example, a transient failure could be a network interruption. When a network interruption occurs, the task retry strategy unit automatically retryes the task three times, attempting to restart the task.

[0067] As another example, the fault tolerance mechanism unit also features a rollback and restart function for failed nodes. When a task permanently fails, the task scheduling execution engine module 200 rolls back to the most recent checkpoint and restarts the upstream tasks in the dependency chain. Specifically, for quantum computing tasks, the rollback includes resetting the state of the quantum device.

[0068] It should be noted that the task can automatically switch the running environment according to the computing resource status, and supports mainstream scheduling systems such as Kubernetes and Slurm.

[0069] In this embodiment, the fault tolerance mechanism covers hardware failures, software anomalies, and network problems, improving task completion rates. Logs include execution time, resource consumption, and error stack traces, facilitating root cause analysis through the log management unit.

[0070] In some embodiments, when the resource awareness and allocation mechanism module 300 uses a multi-dimensional weighted algorithm to allocate computing resources, it is specifically used to: implement a strategy of prioritizing the allocation of computing resources to CPUs with low-latency connections for quantum algorithm tasks; implement a strategy of prioritizing the allocation of big data tasks to local storage nodes; and automatically degrade non-critical tasks when resources are overloaded.

[0071] As an example, this resource allocation strategy improves the overall system performance and stability. By providing low-latency CPU connections for quantum algorithm tasks, it ensures the accuracy and efficiency of quantum computing; by allocating local storage nodes for big data tasks, it reduces I / O bottlenecks; and by intelligently degrading non-critical tasks under high load, it ensures the continuity and reliability of the core computing process. This multi-dimensional optimization mechanism effectively balances the utilization efficiency of heterogeneous computing resources and achieves optimized offloading of high-load tasks.

[0072] In some embodiments, when the calculation result feedback mechanism module 500 automatically collects task results and writes them to the database, it is specifically used to: automatically trigger retry, skip or switch to different execution path operations when the task results meet preset conditions, automatically convert the result data into a standard format and write it to the database for subsequent nodes to call, and provide data visualization charts.

[0073] As an example, the calculation result feedback mechanism module 500 also supports chart display, data visualization, and triggering the next step based on the result conditions. It should be noted that the platform typically supports standard data formats such as HDF5 (Hierarchical Data Format Version 5), CSV (Comma-Separated Values), and JSON (JavaScript Object Notation), facilitating subsequent data analysis and visualization.

[0074] In this embodiment, the calculation result feedback mechanism module 500 can realize the automated closed-loop control of the calculation process, dynamically adjust the execution path through the condition triggering mechanism, and avoid manual intervention and data conversion errors; at the same time, the standardized data storage and visualization capabilities improve the traceability and interpretability of cross-scale calculation results.

[0075] In some embodiments, when the AI-driven process optimization module 800 performs intelligent workflow recommendation, parameter optimization, node priority evaluation, and result error prompting based on historical workflow data, it is specifically used to: adopt an active learning mechanism to train a model through historical workflow data to provide users with intelligent workflow recommendation, node priority evaluation, and parameter optimization suggestions, and to provide error prompts when the task result deviates from the expected threshold.

[0076] The models include a collaborative filtering model for workflow recommendation, a regression model for node priority evaluation and parameter optimization, and a time series model for result error prediction.

[0077] In this embodiment, the AI-driven process optimization module 800 improves the intelligence level and computational reliability of workflow construction through a multi-model collaborative active learning mechanism. This not only lowers the usage threshold for non-professional users and reduces the cost of manual trial and error, but also dynamically optimizes computational parameters and task scheduling strategies, thereby improving the efficiency and accuracy of cross-scale hybrid computing.

[0078] Next, taking water molecules with different bond lengths as an example, this invention will specifically describe the hybrid computational workflow platform 10 based on directed acyclic graphs of this invention:

[0079] like Figure 2 As shown, the platform first imports the structure file of a water molecule, then performs geometry optimization to obtain a stable configuration, and calculates its single-point energy using three methods: HF (Hartree-Fock), QSCI, and DFT (Density Functional Theory). Based on this, the platform automatically integrates the calculation results, generates energy comparison data, and performs multi-dimensional visualization analysis combining charge distribution, molecular orbitals, potential energy surfaces, and electron density. Finally, it provides an intuitive evaluation of the computational accuracy and efficiency of different quantum chemical methods and outputs a comprehensive description of the physicochemical properties of water molecules under multiple theoretical models, verifying the automation, integration, and visualization capabilities of this workflow platform in molecular simulation and computational chemistry tasks.

[0080] In addition, the present invention also provides a terminal device, such as Figure 3 As shown, the terminal device 1 includes the aforementioned hybrid computing workflow platform 10 based on a directed acyclic graph.

[0081] The terminal device 1 provided in this embodiment includes: a processor, a memory, a transceiver, a communication interface and / or a system bus; the memory and the communication interface are connected to the processor and the transceiver through the system bus and complete mutual communication; the memory is used to store computer programs; the communication interface is used to communicate with other devices; and the processor and the transceiver are used to run a hybrid computing workflow platform 10 based on a directed acyclic graph.

[0082] The system bus mentioned above can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. The communication interface is used to enable communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). Memory may include RAM (Random Access Memory) and may also include NVM (non-volatile memory), such as at least one disk drive.

[0083] The processors mentioned above can be general-purpose processors, including CPUs, NPs (Network Processors), etc.; they can also be DSPs (Digital Signal Processors), ASICs (Application Specific Integrated Circuits), FPGAs, or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0084] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.

[0085] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0087] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0088] Similarly, it should be understood that, in order to streamline this application and aid in understanding one or more of the various inventive aspects, features of this application may sometimes be grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of this application. However, this approach should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with features fewer than all features of a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0089] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or elements of any method or apparatus so disclosed may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0090] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0091] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can take the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0092] It should be noted that the above embodiments are illustrative of this application and not limiting of it, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware.

[0093] The above description is merely a specific embodiment or illustration of the embodiments of this application. The scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. A hybrid computing workflow platform based on directed acyclic graphs, characterized in that, The workflow platform includes: A graphical workflow building module is used to provide a drag-and-drop visual interface and generate a structured workflow template based on the user-input visual workflow arrangement. The workflow template represents a directed acyclic graph formed by task nodes and their dependencies. The task scheduling and execution engine module is used to parse the workflow template, construct a directed acyclic graph object and its dependencies, determine the task execution order and identify parallelizable tasks based on the directed acyclic graph object and its dependencies and through a topological sorting algorithm, and schedule tasks in a sequential and / or parallel manner. The resource awareness and allocation mechanism module is used to monitor the resource status of heterogeneous computing resources in real time, and allocate computing resources according to the resource status and the computing type of the node using a multi-dimensional weighted algorithm. The heterogeneous computing resources include at least two of CPU, GPU, FPGA and quantum simulator. The multi-dimensional factors include: task priority weight, model type and data size. The node plugin interface management module is used to define a unified task node interface standard, and supports command line parameter configuration, input and output file path management and parameter validation. The calculation result feedback mechanism module is used to automatically collect task execution results and write them into the database to achieve a closed loop of data flow. The template reuse and version control module is used to save and share standard process templates, and supports version rollback, parameter overwrite and data migration; A quantum-classical computing interface fusion module is used to interface with quantum computing resources and coordinate the scheduling and execution of the quantum computing resources and classical computing tasks. The AI-driven process optimization module is used to perform intelligent workflow recommendations, parameter optimization, node priority evaluation, and result error prompts based on historical workflow data.

2. The hybrid computing workflow platform based on directed acyclic graphs according to claim 1, characterized in that, The visual workflow orchestration includes: node dragging, parameter entry, and node connection; The graphical workflow construction module is specifically used to: detect in real time whether a loop is formed when the user drags and drops the node, fills in the parameter, and connects the node through the drag-and-drop visual interface. If a loop is detected, it will highlight the error and prevent saving, so as to ensure that the generated structured workflow template conforms to the directed acyclic graph structure.

3. The hybrid computing workflow platform based on directed acyclic graphs according to claim 2, characterized in that, The task scheduling and execution engine module also includes a fault tolerance mechanism unit, a task retry strategy unit, a runtime environment management unit, and a log management unit; The fault tolerance mechanism unit is used to perform multiple backups for tasks marked as critical tasks; The task retry strategy unit is used to automatically retry the task up to a predetermined number of times when a transient failure occurs; The runtime environment management unit is used to automatically switch the task runtime environment according to the computing resource status; The log management unit is used to collect and store all task logs in real time.

4. The hybrid computing workflow platform based on directed acyclic graphs according to claim 2, characterized in that, When the resource awareness and allocation mechanism module uses a multi-dimensional weighted algorithm to allocate computational resources, it is specifically used for: A strategy is implemented to prioritize the allocation of computing resources to CPUs with low-latency connections for quantum algorithm tasks; Implement a strategy that prioritizes allocating big data tasks to local storage nodes; When resources are overloaded, non-critical tasks are automatically degraded.

5. The hybrid computing workflow platform based on directed acyclic graphs according to claim 1, characterized in that, The calculation result feedback mechanism module, when automatically collecting task results and writing them to the database, is specifically used for: When the task result meets the preset conditions, it automatically triggers retry, skip, or redirects to different execution paths, and automatically converts the result data into a standard format and writes it into the database for subsequent nodes to call, and provides data visualization charts.

6. The hybrid computing workflow platform based on directed acyclic graphs according to claim 5, characterized in that, The template reuse and version control module is also used to set the visibility range of the template and assign role permissions; The visibility scope includes: private, team, and organization-wide levels, and the roles include: viewer, editor, and administrator.

7. The hybrid computing workflow platform based on directed acyclic graphs according to claim 5, characterized in that, The quantum computing resources include: a local quantum simulator, a remote quantum cloud platform, and an actual quantum processor.

8. The hybrid computing workflow platform based on directed acyclic graphs according to claim 1, characterized in that, The AI-driven process optimization module, when performing intelligent workflow recommendation, parameter optimization, node priority evaluation, and result error prompting based on historical workflow data, is specifically used for: An active learning mechanism is adopted to train the model through the historical workflow data, providing users with intelligent workflow recommendations, node priority evaluation, parameter optimization suggestions, and error prompts when the task results deviate from the expected threshold. The model includes a collaborative filtering model for workflow recommendation, a regression model for node priority evaluation and parameter optimization, and a time series model for result error prediction.

9. A terminal device, characterized in that, include: A hybrid computing workflow platform based on a directed acyclic graph as described in any one of claims 1-8.

Images