CN112488563A - Determination method and device for force calculation parameters - Google Patents
Determination method and device for force calculation parameters Download PDFInfo
- Publication number
- CN112488563A CN112488563A CN202011460037.4A CN202011460037A CN112488563A CN 112488563 A CN112488563 A CN 112488563A CN 202011460037 A CN202011460037 A CN 202011460037A CN 112488563 A CN112488563 A CN 112488563A
- Authority
- CN
- China
- Prior art keywords
- service request
- computing power
- parameter
- determining
- calculation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application discloses a method and a device for determining a calculation force parameter, relates to the field of network communication, and is used for improving the efficiency of determining the calculation force parameter. The determination method comprises the following steps: acquiring a computing power service request of a user; determining the parameter type of the computing power service request according to the computing power service request; determining a force calculation parameter based on the force calculation service request and the current neural network model; the current neural network model corresponds to the parameter type of the computing power service request. Compared with the method for manually determining the computing power parameter corresponding to the service request in the prior art, the method needs the user to have certain professional ability, and is low in efficiency.
Description
Technical Field
The present application relates to the field of network communications, and in particular, to a method and an apparatus for determining computational power parameters.
Background
In recent years, Artificial Intelligence (AI) has become a general technology in contemporary society and an indispensable technology for intelligent society in the future. Algorithms and data are three elements of AI. The calculation force is used as the basis of AI, and directly influences the application and deployment of AI services. The AI industry, which has emerged today, and the underlying intelligent services all place high demands on computing power.
The calculation parameters corresponding to different service requests are different, and the content and the form of the service requests are various. Therefore, when the description of the service request is relatively abstract, the user is required to manually analyze the content of the service request, so as to determine the calculation parameters corresponding to the service request. However, the method for manually determining the calculation force parameters corresponding to the service request requires that the user has certain professional ability, and is low in efficiency.
Disclosure of Invention
The application provides a determination method and a determination device for force calculation parameters, which are used for improving the efficiency of determining the force calculation parameters.
In order to achieve the purpose, the technical scheme is as follows:
in a first aspect, the present application provides a method of determining a computational force parameter. The determination method comprises the following steps: and acquiring a computing service request of a user. Then, according to the computing power service request, determining the type of the computing power service request, and determining a target computing power parameter based on the current neural network model corresponding to the type of the computing power service request.
According to the determination method of the computing power parameter, the computing power parameter is determined based on the computing power service request and the current neural network model by obtaining the computing power service request of the user. Compared with the method for manually determining the computing power parameter corresponding to the service request in the prior art, the method needs the user to have certain professional ability, and is low in efficiency.
In a second aspect, the present application provides an apparatus for determining a computational power parameter. The device includes: the acquiring unit is used for acquiring a computing power service request of a user; the determining unit is used for determining the type of the computing power service request according to the computing power service request acquired by the acquiring unit; the determining unit is further configured to determine a calculation power parameter based on the calculation power service request acquired by the acquiring unit and the current neural network model; the current neural network model corresponds to the type of computing service request described above.
In a third aspect, the present application provides an computational force parameter determination apparatus comprising a memory and a processor. The memory is coupled to the processor. The memory is for storing computer program code comprising computer instructions. The computational force parameter determination apparatus, when executed by a processor, performs the computational force parameter determination method as described in the first aspect and any of its possible design forms.
In a fourth aspect, the present application provides a computer-readable storage medium having stored therein instructions, which, when run on a computing power parameter determining apparatus, cause the apparatus to perform the computing power parameter determining method according to the first aspect and any possible design thereof.
In a fifth aspect, the present application provides a computer program product comprising computer instructions which, when run on a computational power parameter determining device, cause the computational power parameter determining device to perform the computational power parameter determining method according to the first aspect and any of its possible design approaches.
For a detailed description of the second to fifth aspects and their various implementations in this application, reference may be made to the detailed description of the first aspect and its various implementations; moreover, the beneficial effects of the second aspect to the fifth aspect and the various implementation manners thereof may refer to the beneficial effect analysis of the first aspect and the various implementation manners thereof, and are not described herein again.
These and other aspects of the present application will be more readily apparent from the following description.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a first flowchart illustrating a method for determining a calculation force parameter provided in the present application;
FIG. 2 is a schematic diagram of a first neural network expert system provided in the present application;
FIG. 3 is a schematic structural diagram of a second neural network expert system provided in the present application;
FIG. 4 is a second flowchart illustrating a method for determining a calculation force parameter according to the present application;
fig. 5 is a schematic hardware structure diagram of a computational power parameter determination device provided in an embodiment of the present application;
fig. 6 is a schematic structural diagram of a determination apparatus for computing force parameters according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless otherwise specified.
In recent years, Artificial Intelligence (AI) has become a general technology in contemporary society and an indispensable technology for intelligent society in the future. Algorithms and data are three elements of AI. The calculation force is used as the basis of AI, and directly influences the application and deployment of AI services. The AI industry, which has emerged today, and the underlying intelligent services all place high demands on computing power.
The calculation parameters corresponding to different service requests are different, and the content and the form of the service requests are various. Therefore, when the description of the service request is relatively abstract, the user is required to manually analyze the content of the service request, so as to determine the calculation parameters corresponding to the service request. However, the method for manually determining the calculation force parameters corresponding to the service request requires that the user has certain professional ability, and is low in efficiency.
In order to solve the above problems, the present application provides a method for determining a calculation power parameter, which determines the calculation power parameter based on a calculation power service request of a user and a current neural network model by obtaining the calculation power service request. Compared with the method for manually determining the computing power parameter corresponding to the service request in the prior art, the method needs the user to have certain professional ability, and is low in efficiency.
The execution subject of the determination method of the computational power parameter provided by the embodiment of the application is a determination device (hereinafter, simply referred to as a determination device) of the computational power parameter. The determining device may be integrated with the computing power network system, or may be provided independently, which is not limited in this application.
The computing power network system is a network system which can issue the current computing power state and the network state as routing information to a network, and the network routes a computing task message to a corresponding computing node to realize optimal user experience, optimal computing resource utilization rate and optimal network efficiency. The computing task is dynamically and flexibly scheduled by building the dynamic routing capability of the computing task in the computing network system according to the service requirement and based on real-time multidimensional factors such as computing resource performance, network performance, cost and the like, so that the resource utilization rate and the network utilization efficiency are improved, and the service user experience is improved. And facing to an edge computing scene, edge computing network formation can be realized through a computing network system, edge-edge cooperation is realized, the characteristics of multiple instances and multiple copies of services are utilized, the nearby access of a user and the load balance of the services are realized, the problems of complex deployment, low efficiency, low resource reuse rate and the like are solved, and the edge computing scale deployment is assisted.
The method for determining the computational power parameter provided in the embodiment of the present application is described below.
As shown in fig. 1, the method for determining the calculation force parameter includes:
s101, the determining device obtains a computing power service request of a user.
Alternatively, the determination means may receive a user-entered computing service request.
Optionally, the computing service request may include a rendering service, a training task, an inference service, a video playback, a supercomputing class, or other services.
The computing power service request comprises a computing power parameter, a service requirement parameter of the user or an abstract requirement parameter of the user.
The computing power parameter is a parameter corresponding to a bottom hardware resource required by the computing power network system to meet the computing power service.
Optionally, the computational parameters may include parameters of a Central Processing Unit (CPU), parameters of a Graphics Processing Unit (GPU), stored parameters, parameters of network indexes, and the like.
The service requirement parameter of the user is the specific requirement of the user on the computing service.
For example, the service requirement parameters of the user may include a requirement parameter of the user for a service, a requirement parameter for a model image, a requirement parameter for an algorithm, a requirement parameter for time, and a requirement parameter for a running time.
The abstract requirement parameter of the user is abstract description of the computing power service request of the user.
Illustratively, the abstract requirement parameter of the user is a neural network model for training face recognition, or the abstract requirement parameter of the user is a subsequent rendering of a movie.
S102, the determining device determines the type of the computing power service request according to the computing power service request.
The type of the computing power service request comprises a computing power parameter type, a service requirement type or an abstract requirement type.
Alternatively, the determination means may determine the type of the computing service request based on a keyword included in the computing service request.
The keywords include a calculation power parameter keyword, a business requirement keyword, and an abstract requirement keyword.
Optionally, the computational parameter keywords include CPU, GPU, tensor operation, storage, encoding and decoding, and network index.
For example, as shown in table 1, when a keyword CPU is included in a computing power service request, a parameter of the CPU 10 Tera Operations Per Second (TOPS) is determined as a computing power parameter; when the computing power service request contains a keyword GPU, determining a parameter 12 (TFLOPs) of the GPU as a computing power parameter; when the computing power service request comprises keyword tensor operation, determining a parameter 8TFLOPS of a tensor operation embedded neural Network Processor (NPU)/Tensor Processing Unit (TPU) as a computing power parameter; when the computing power service request comprises keyword storage, determining 10 kilomega of the stored parameters as computing power parameters; when the computing power service request contains a keyword network index, determining that the parameter delay of the network index is less than 50 milliseconds (ms) as a computing power parameter; and when the computing power service request comprises the keyword coding and decoding, determining that the coding and decoding meet the standard as the computing power parameter.
TABLE 1
And determining the type of the computing power service request as the type of the computing power parameters, wherein the number of the computing power parameters included in the computing power service request is greater than or equal to a first preset threshold value.
The first preset threshold may be determined according to actual conditions, which is not limited in this application.
Further, if the number of the key force calculation parameters in the force calculation service request is greater than or equal to a second preset threshold, determining that the type of the force calculation service request is the type of the force calculation parameters.
The second preset threshold may be determined according to actual conditions, which is not limited in this application.
The key force calculation parameters are force calculation parameters which are necessary for meeting force calculation business in the force calculation parameters.
For example, the key computation parameter may be a parameter of the CPU, a parameter of the GPU, a stored parameter, or a parameter of a network index.
The type of the computing power service request is judged to be the computing power parameter type through the key computing power parameter, so that the accuracy of determining the type of the computing power service request to be the computing power parameter type can be improved.
Optionally, the service requirement keyword includes a service, a model mirror image, an algorithm, time, a running time, and a data volume.
For example, as shown in table 2, when it is determined that the computing power service request includes a keyword service, it is determined that video rendering or model training of a demand parameter for the service is a service demand parameter; when determining that the computational power service request contains a keyword model mirror image, determining a demand parameter ray tracing and global light rendering KeyShot or a residual error network ResNet of the model mirror image as a service demand parameter; when the computing power service request is determined to contain a keyword algorithm, determining a required parameter resolution 1080 of the algorithm to be progressive e scanning (p) or a 152-layer network and 6 thousand parameters as service requirement parameters; when determining that the computing power service request contains keyword time, determining a demand parameter for time as a service demand parameter for one month or one week; when determining that the operation times of the keywords are contained in the computing power service request, determining the requirement parameters of the operation times once or hundred thousand times as service requirement parameters; and when determining that the computing power service request contains the data volume of the keywords, determining a demand parameter 10G video or a 10G video for the data volume as a service demand parameter.
TABLE 2
And when the number of the service demand parameters included in the calculation force service request is determined to be greater than or equal to a third preset threshold value, determining the type of the calculation force service request as the service demand type.
The third preset threshold may be determined according to actual conditions, which is not limited in this application.
Further, if the number of the key service requirement parameters in the computing service request is greater than or equal to a fourth preset threshold, determining that the type of the computing service request is the service requirement parameter type.
The fourth preset threshold may be determined according to actual conditions, which is not limited in this application.
The key business requirement parameters are the business requirement parameters which are necessary for meeting the computing power business in the computing power parameters.
The key business requirement parameters can be requirement parameters for businesses, requirement parameters for model images, requirement parameters for algorithms, requirement for time and requirement for operation times.
The type of the computing power service request is judged to be the type of the service demand parameter through the key service demand parameter, so that the accuracy of determining the type of the computing power service request to be the type of the service demand parameter can be improved.
And when determining that the number of the computing power parameters included in the computing power service request is smaller than a first preset threshold value and the number of the service demand parameters is smaller than a second preset threshold value and the abstract demand parameters are included, determining that the type of the computing power service request is the abstract demand type.
The abstract requirement parameter is abstract description of the user on the computing power service request.
Illustratively, the computing power service request of the user is a neural network model for training face recognition, the computing power service request does not include a computing power parameter and does not include a service requirement parameter, and the type of the computing power service request is an abstract requirement type if the neural network model for training face recognition is the abstract requirement parameter.
It should be noted that, when the computing power service request includes the computing power parameter, the service requirement parameter, and the abstract requirement parameter, and the number of the computing power parameters is greater than or equal to the first preset threshold, and the number of the service requirement parameters is greater than or equal to the third preset threshold, the type of the computing power service request may be the type of the computing power parameter, may also be the type of the service requirement, and may also be the type of the abstract requirement.
Optionally, when the type of the computing power service request of the user is the computing power parameter type, the determining device determines that the computing power parameter included in the computing power service request is the target computing power parameter. In the case where the type of the computation power parameter is the business requirement type or the abstract requirement type, S103 is continuously performed.
S103, the determining device determines the calculation force parameters based on the calculation force service request and the current neural network model.
The neural network model includes a first neural network model and a second neural network model.
The first neural network model corresponds to the type of the computing power service request being a service demand type. That is, when the type of the computing power service request is the service demand type, the current neural network model is the first neural network model.
The first neural network model is used for predicting force parameters according to business demand parameters.
Alternatively, the first neural network model may be a neural network model in a first neural network expert system.
As shown in fig. 2, the first neural network expert system 20 includes a first neural network model 21, a first rule base 22, and a first output interpretation module 23.
The input to the first neural network expert system 20 is a business requirement parameter in the computational business request and the output of the first neural network expert system 20 is a computational parameter.
The first rule base 22 is used for saving expert knowledge corresponding to the business requirement parameters and the computing power parameters and training the first neural network model 21.
Optionally, the first rule base 22 may store expert knowledge of the calculation power parameter and the business requirement parameter obtained from actual experience. Expert knowledge can be from a large set of instances actually running, or from a large number of rules derived from experts or literature in the field, or a mixture of both.
Illustratively, as shown in table 3, the first rule base 22 includes service requirement parameters of recommended classical AlexNet model configuration, 80w nodes, 6096w parameters, 3.8% of convolutional layer parameters, 96.2% of full connection layer parameters, recommended storage of 10G, recommended completion time of 1 week, recommended network configuration of non-real time, corresponding computing parameters of 10TOPS for logical operation, 12TFLOPS for matrix operation, 8TFLOPS for tensor operation, 10G for storage, and non-real time for network; the service demand parameters are 8 paths of videos, the definition is 1080p, the service demand parameters are stored as 2T, the network delay is less than 200ms, the uplink bandwidth is xxMB, the downlink bandwidth is xxMb, the coding and decoding capability is xxx standard, the corresponding calculation parameters are that the logical operation is 3TO PS, the matrix operation is 5TFLOPS, the service demand parameters are stored as 25T, the network delay is less than 200ms, the uplink bandwidth is xxMB, the downlink bandwidth is xxMb, the coding and decoding are 2 decoding engines and 4 coding engines.
TABLE 3
The first output interpretation module 23 is configured to translate the result of the prediction obtained by the first neural network model 21 to obtain the calculation force parameter.
Illustratively, when the type of the computing power service request is determined to be the service demand type, the service demand parameter in the computing power service request is substituted into the first neural network model to obtain a first prediction result, and the first prediction result is interpreted through a first output interpretation mechanism to obtain the computing power parameter.
The second neural network model corresponds to the type of the computing power service request being an abstract requirement type. Namely, when the type of the computing service request is the abstract requirement type, the current neural network model is the second neural network model.
The second neural network model is used for predicting force parameters according to the abstract demand parameters.
Alternatively, the second neural network model may be a neural network model in a second neural network expert system.
As shown in fig. 3, the second neural network expert system 30 includes a second neural network model 31, a second rule base 32, and a second output interpretation module 33.
The input to the second neural network expert system 30 is the abstract demand parameter in the computing power service request and the output of the second neural network expert system 30 is the computing power parameter.
The second rule base 32 is used for saving expert knowledge corresponding to the abstract requirement parameters and the computing power parameters and training the second neural network model 31.
Optionally, the second rule base 32 may store expert knowledge corresponding to the calculation force parameter and the business requirement parameter obtained from actual experience. Expert knowledge can be from a large set of instances actually running, or from a large number of rules derived from experts or literature in the field, or a mixture of both.
Illustratively, as shown in table 4, the second rule base 32 includes abstract requirement parameters of training a network model for image recognition, training data 8G, and model uncertainty, and corresponding computation parameters of 10TOPS for logical operation, 12TFLOPS for matrix operation, 8TFLOPS for tensor operation, 10G for storage, and non-real-time network; the abstract requirement parameter is deployment of a cloud Augmented Reality (AR)/Virtual Reality (VR) game, the total content of the game is 200G, the corresponding calculation force parameter is that the logic calculation is 3TOPS, the matrix operation is 5TFLOPS, the storage is 25T, the network delay is less than 200ms, the uplink bandwidth is xxMB, the downlink bandwidth is xxMB, and the encoding and decoding are 2 decoding engines and 4 encoding engines.
TABLE 4
The second output interpretation module 33 is configured to translate the predicted result obtained by the second neural network model 31 to obtain the force calculation parameter.
Illustratively, when the type of the computing power service request is determined to be an abstract requirement type, the abstract requirement parameter in the computing power service request is substituted into the second neural network model to obtain a second prediction result, and the second prediction result is interpreted through a second output interpretation mechanism to obtain the computing power parameter.
Optionally, with reference to fig. 1, as shown in fig. 4, after S103, the method for determining the calculated force parameter further includes:
and S104, the determining device determines that the difference value between the calculation force parameter and the target calculation force parameter is less than or equal to a preset threshold value.
The target calculation force parameter is a calculation force parameter meeting the calculation force service request.
Optionally, after the determining device determines the calculation power parameter, the calculation power parameter may be compared with the target calculation power parameter, and when a difference between the calculation power parameter and the target calculation power parameter is less than or equal to a preset threshold, it is determined that the calculation power parameter meets the calculation power service request of the user.
And S105, optimizing the current neural network model based on the calculation force parameters and the calculation force service request.
An implementable approach optimizes a current neural network model based on the computing power parameter and the business requirement parameter in the computing power business request.
Illustratively, when the type of the computing power service request is determined to be the service demand type, the computing power parameter is determined based on the service demand parameter in the computing power service request and the first neural network model, and when the difference value between the computing power parameter and the target computing power parameter is smaller than or equal to a preset threshold value, the first rule base is updated according to the determined computing power parameter and the service demand parameter in the computing power service request.
And training the first neural network model by using the updated first rule base to obtain the trained first neural network model.
For example, as shown in table 5, the updated first rule base includes that the abstract requirement parameters of the user are network model training data 8G for training an image recognition, the model is uncertain, the service requirement parameters are recommended classical AlexNet model configuration, 80w nodes, 6096w parameters, convolutional layer parameters 3.8%, full connection layer parameters 96.2%, recommended storage is 10G, recommended completion time is 1 week, recommended network configuration is non-real-time, the corresponding computation force parameters are logical operation of 10TOPS, matrix operation of 12TFLOPS, tensor operation of 8TFLOPS, storage of 10G, network non-real-time, and the computation force parameters satisfy the requirements of the user.
TABLE 5
And when the next determination device determines the force calculation parameters based on the service requirement parameters in the force calculation service request and the current neural network model, the trained first neural network model is the current neural network model.
In another implementation, the current neural network model is optimized based on the computing power parameter and the abstract requirement parameter in the computing power service request.
Illustratively, when the type of the computing power service request is determined to be an abstract requirement type, the computing power parameter is determined based on the abstract requirement parameter in the computing power service request and the second neural network model.
And the difference value between the calculation force parameter and the target calculation force parameter is less than or equal to a preset threshold value. And updating the second rule base according to the determined calculation parameters and the abstract requirement parameters in the calculation service request.
And training the second neural network model by using the updated second rule base to obtain the trained second neural network model.
And when the next determination device determines the force calculation parameters based on the abstract requirement parameters in the force calculation service request and the current neural network model, the trained second neural network model is the current neural network model.
According to the determination method of the computing power parameter, the computing power parameter is determined based on the computing power service request and the current neural network model by obtaining the computing power service request of the user. Compared with the method for manually determining the computing power parameter corresponding to the service request in the prior art, the method needs the user to have certain professional ability, and is low in efficiency.
The scheme provided by the embodiment of the application is mainly introduced from the perspective of a method. To implement the above functions, it includes hardware structures and/or software modules for performing the respective functions. Those of skill in the art would readily appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as hardware or combinations of hardware and computer software. Whether a function is performed as hardware or computer software drives hardware depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
As shown in fig. 5, an embodiment of the present application provides a determination apparatus 500 for computing force parameters. The computing power parameter determining device may include at least one processor 501, a communication line 502, a memory 503, and a communication interface 504.
Specifically, the processor 501 is configured to execute computer-executable instructions stored in the memory 503, so as to implement steps or actions of the terminal.
The processor 501 may be a chip. For example, the Field Programmable Gate Array (FPGA) may be a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), a system on chip (So C), a Central Processing Unit (CPU), a Network Processor (NP), a digital signal processing circuit (DSP), a Microcontroller (MCU), a Programmable Logic Device (PLD) or other integrated chips.
A communication line 502 for transmitting information between the processor 501 and the memory 503.
The memory 503 is used for storing and executing computer execution instructions, and is controlled by the processor 501 to execute.
The memory 503 may be separate and coupled to the processor via a communication link 502. The memory 503 may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The non-volatile memory may be a read-only memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an electrically Erasable EPROM (EEPROM), or a flash memory. Volatile memory can be Random Access Memory (RAM), which acts as external cache memory. By way of example, and not limitation, many forms of RAM may be used, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), double data rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM). It should be noted that the memory of the systems and devices described herein is intended to comprise, without being limited to, these and any other suitable types of memory.
A communication interface 504 for communicating with other devices or a communication network. The communication network may be an ethernet, a Radio Access Network (RAN), or a Wireless Local Area Network (WLAN).
It is noted that the configuration shown in fig. 5 does not constitute a limitation of the determination device of the calculated force parameter, and the determination device of the calculated force parameter may include more or less components than those shown in fig. 5, or combine some components, or arrange different components, in addition to the components shown in fig. 5.
As shown in fig. 6, the present embodiment provides a calculation force parameter determination device 60. The calculation force parameter determination device may include an acquisition unit 61 and a determination unit 62.
The obtaining unit 61 is configured to obtain a computing service request of a user. For example, in conjunction with fig. 1, the acquisition unit 61 may be configured to perform S101.
A determining unit 62, configured to determine the type of the computing power service request acquired by the acquiring unit 61. For example, in conjunction with fig. 1, the determination unit 62 may be configured to perform S102.
And the determining unit 62 is further configured to determine the computational power parameter based on the computational power service request acquired by the acquiring unit 61 and the current neural network model. For example, in conjunction with fig. 1, the determination unit 62 may be configured to perform S103.
It should be understood that, in the various embodiments of the present application, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
In actual implementation, the obtaining unit 61 and the determining unit 62 may be implemented by the processor 501 shown in fig. 5 calling the program code in the memory 503. The specific implementation process may refer to the description of the determination method part of the calculation force parameter shown in fig. 1, and is not described herein again.
Another embodiment of the present application further provides a computer-readable storage medium, which stores therein computer instructions, which, when executed on a computing power parameter determining device, cause the computing power parameter determining device to perform the steps performed by the computing power parameter determining device in the method flow shown in the above method embodiment.
In another embodiment of the present application, there is also provided a computer program product comprising instructions that, when executed on a computing power parameter determining device, cause the computing power parameter determining device to perform the steps performed by the computing power parameter determining device in the method flow shown in the above method embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, devices and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus, and method may be implemented in other ways. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. A method for determining a computational force parameter, the method comprising:
acquiring a computing power service request of a user;
determining the type of the computing power service request according to the computing power service request;
determining a force calculation parameter based on the force calculation service request and a current neural network model; the current neural network model corresponds to a type of the computing power service request.
2. The method of claim 1, wherein determining the type of the computing service request comprises:
and determining the type of the computing power service request according to the keywords contained in the computing power service request.
3. The method of determining according to claim 1, further comprising:
determining that the difference value between the calculation force parameter and the target calculation force parameter is less than or equal to a preset threshold value; the target calculation force parameter is a calculation force parameter meeting the calculation force service request;
optimizing the current neural network model based on the computational power parameters and the computational power service request.
4. An apparatus for determining a computational power parameter, the apparatus comprising:
the acquiring unit is used for acquiring a computing power service request of a user;
the determining unit is used for determining the type of the computing power service request according to the computing power service request acquired by the acquiring unit;
the determining unit is further configured to determine a force calculation parameter based on the force calculation service request acquired by the acquiring unit and a current neural network model; the current neural network model corresponds to a type of the computing power service request.
5. The determination apparatus according to claim 4, wherein the determination unit is specifically configured to:
and determining the type of the computing power service request according to the keywords contained in the computing power service request.
6. The determination apparatus according to claim 4, wherein the determination unit is further configured to:
determining that the difference value between the calculation force parameter and the target calculation force parameter is less than or equal to a preset threshold value; the target calculation force parameter is a calculation force parameter meeting the calculation force service request;
optimizing the current neural network model based on the computational power parameters and the computational power service request.
7. An computational force parameter determination device, characterized in that the computational force parameter determination device comprises a memory and a processor; the memory and the processor are coupled; the memory for storing computer program code, the computer program code comprising computer instructions; when the processor executes the computer instructions, the computational force parameter determination apparatus performs the computational force parameter determination method according to any one of claims 1 to 3.
8. A computer-readable storage medium, having stored therein instructions, which, when run on a computational force parameter determination device, cause the device to perform the computational force parameter determination method of any one of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011460037.4A CN112488563B (en) | 2020-12-11 | 2020-12-11 | Method and device for determining calculation force parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011460037.4A CN112488563B (en) | 2020-12-11 | 2020-12-11 | Method and device for determining calculation force parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112488563A true CN112488563A (en) | 2021-03-12 |
| CN112488563B CN112488563B (en) | 2023-06-06 |
Family
ID=74916169
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011460037.4A Active CN112488563B (en) | 2020-12-11 | 2020-12-11 | Method and device for determining calculation force parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112488563B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114062891A (en) * | 2021-10-11 | 2022-02-18 | 深圳市德斯戈智能科技有限公司 | Calculation force testing device based on AI chip |
| CN115511388A (en) * | 2022-11-10 | 2022-12-23 | 浩鲸云计算科技股份有限公司 | Business demand calculation capacity method based on rule measurement and calculation model |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190050388A1 (en) * | 2016-02-22 | 2019-02-14 | Orphanalytics Sa | Method and device for detecting style within one or more symbol sequences |
| CN110532092A (en) * | 2019-08-21 | 2019-12-03 | 云湾科技(嘉兴)有限公司 | Reso urce matching method, device and equipment |
| CN110659134A (en) * | 2019-09-04 | 2020-01-07 | 腾讯云计算（北京）有限责任公司 | Data processing method and device applied to artificial intelligence platform |
| CN110782343A (en) * | 2018-07-30 | 2020-02-11 | 中移(苏州)软件技术有限公司 | Calculation power circulation method and system based on block chain in distributed network |
| CN111131516A (en) * | 2019-12-31 | 2020-05-08 | 中国联合网络通信集团有限公司 | Service configuration method and device |
| CN111126594A (en) * | 2019-11-25 | 2020-05-08 | 北京邮电大学 | Neural network model dynamic segmentation method and device based on edge calculation |
| CN111144561A (en) * | 2018-11-05 | 2020-05-12 | 杭州海康威视数字技术股份有限公司 | Neural network model determining method and device |
| CN111401516A (en) * | 2020-02-21 | 2020-07-10 | 华为技术有限公司 | Neural network channel parameter searching method and related equipment |
| CN111818585A (en) * | 2020-07-30 | 2020-10-23 | 中国联合网络通信集团有限公司 | User access method and access network equipment |
| CN111866775A (en) * | 2020-07-28 | 2020-10-30 | 中国联合网络通信集团有限公司 | Service arranging method and device |
| CN112035247A (en) * | 2020-08-12 | 2020-12-04 | 博泰车联网(南京)有限公司 | Resource scheduling method, vehicle machine and computer storage medium |
-
2020
- 2020-12-11 CN CN202011460037.4A patent/CN112488563B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190050388A1 (en) * | 2016-02-22 | 2019-02-14 | Orphanalytics Sa | Method and device for detecting style within one or more symbol sequences |
| CN110782343A (en) * | 2018-07-30 | 2020-02-11 | 中移(苏州)软件技术有限公司 | Calculation power circulation method and system based on block chain in distributed network |
| CN111144561A (en) * | 2018-11-05 | 2020-05-12 | 杭州海康威视数字技术股份有限公司 | Neural network model determining method and device |
| CN110532092A (en) * | 2019-08-21 | 2019-12-03 | 云湾科技(嘉兴)有限公司 | Reso urce matching method, device and equipment |
| CN110659134A (en) * | 2019-09-04 | 2020-01-07 | 腾讯云计算（北京）有限责任公司 | Data processing method and device applied to artificial intelligence platform |
| CN111126594A (en) * | 2019-11-25 | 2020-05-08 | 北京邮电大学 | Neural network model dynamic segmentation method and device based on edge calculation |
| CN111131516A (en) * | 2019-12-31 | 2020-05-08 | 中国联合网络通信集团有限公司 | Service configuration method and device |
| CN111401516A (en) * | 2020-02-21 | 2020-07-10 | 华为技术有限公司 | Neural network channel parameter searching method and related equipment |
| CN111866775A (en) * | 2020-07-28 | 2020-10-30 | 中国联合网络通信集团有限公司 | Service arranging method and device |
| CN111818585A (en) * | 2020-07-30 | 2020-10-23 | 中国联合网络通信集团有限公司 | User access method and access network equipment |
| CN112035247A (en) * | 2020-08-12 | 2020-12-04 | 博泰车联网(南京)有限公司 | Resource scheduling method, vehicle machine and computer storage medium |
Non-Patent Citations (3)
| Title |
|---|
| HAIFENG LU 等: "Optimization of lightweight task offloading strategy for mobile edge computing based on deep reinforcement learning", 《FUTURE GENERATION COMPUTER SYSTEMS》, vol. 102, pages 847 - 861 * |
| ZHI YANG 等: "A Cost-Based Resource Scheduling Paradigm in Cloud Computing", 《2011 12TH INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED COMPUTING, APPLICATIONS AND TECHNOLOGIES》, pages 417 - 422 * |
| 李建飞 等: "算力网络中面向业务体验的算力建模", 《中兴通讯技术》, vol. 26, no. 5, pages 34 - 38 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114062891A (en) * | 2021-10-11 | 2022-02-18 | 深圳市德斯戈智能科技有限公司 | Calculation force testing device based on AI chip |
| CN115511388A (en) * | 2022-11-10 | 2022-12-23 | 浩鲸云计算科技股份有限公司 | Business demand calculation capacity method based on rule measurement and calculation model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112488563B (en) | 2023-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109783237B (en) | Resource allocation method and device | |
| CN110069341B (en) | Method for scheduling tasks with dependency relationship configured according to needs by combining functions in edge computing | |
| US10652360B2 (en) | Access scheduling method and apparatus for terminal, and computer storage medium | |
| CN109191287B (en) | Block chain intelligent contract fragmentation method and device and electronic equipment | |
| CN114915630B (en) | Task allocation method, network training method and device based on Internet of Things equipment | |
| CN112488563A (en) | Determination method and device for force calculation parameters | |
| CN110231984B (en) | Multi-workflow task allocation method and device, computer equipment and storage medium | |
| CN111126594A (en) | Neural network model dynamic segmentation method and device based on edge calculation | |
| CN111752691A (en) | AI (artificial intelligence) calculation graph sorting method, device, equipment and storage medium | |
| CN113645637A (en) | Method and device for unloading tasks of ultra-dense network, computer equipment and storage medium | |
| CN112202829A (en) | Social robot scheduling system and scheduling method based on micro-service | |
| CN112559078B (en) | Method and system for hierarchically unloading tasks of mobile edge computing server | |
| CN115955685A (en) | Multi-agent cooperative routing method, equipment and computer storage medium | |
| CN114461299B (en) | Unloading decision determining method and device, electronic equipment and storage medium | |
| Yang et al. | Coded computing in unknown environment via online learning | |
| CN113271606B (en) | Service scheduling method for ensuring stability of cloud native mobile network and electronic equipment | |
| CN110866127A (en) | Method for establishing index and related device | |
| CN109739513B (en) | Dynamic scheduling method and device for service requests under multi-edge cloud | |
| CN117707797B (en) | Task scheduling method and device based on distributed cloud platform and related equipment | |
| CN117707797A (en) | Task scheduling method and device based on distributed cloud platform and related equipment | |
| CN112019368B (en) | VNF migration method, VNF migration device and VNF migration storage medium | |
| Liu et al. | Computation Offloading and Task Scheduling with Fault-Tolerance for Minimizing Redundancy in Edge Computing | |
| CN115840634B (en) | Service execution method, device, equipment and storage medium | |
| CN115361284A (en) | Deployment adjustment method of virtual network function based on SDN | |
| CN117149424A (en) | Resource capacity expansion method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |