Method for measuring computing capability of intelligent computing center by means of 1-degree computational power, and device

Abstract

The present disclosure relates to the technical field of computational power. Provided are a method for measuring the computing capability of an intelligent computing center by means of 1-degree computational power, and a device. The method comprises: step S1, acquiring an initial half-precision floating-point computational power number, an initial video memory bandwidth and an initial video memory capacity of a processor of a first intelligent computing center; step S2, calculating a first ratio of the initial half-precision floating-point computational power number to a preset half-precision floating-point computational power number, a second ratio of the initial video memory bandwidth to a preset video memory bandwidth, and a third ratio of the initial video memory capacity to a preset video memory capacity; and step S3, setting a product of a weighted sum of the first ratio, the second ratio and the third ratio, a preset unit of floating point operations, and a preset time unit as an initial 1-degree computational power value of the first intelligent computing center.

Description

Method and device for calculating computing power of 1-degree computing power measurement intelligent computing center

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] The present disclosure claims priority to Chinese Patent Application No. 202411224087.0, filed on September 03, 2024 in China, the contents of which are incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present disclosure relates to the technical field of computing power, in particular to a method and device for calculating computing power of a 1-degree computing power measurement intelligent computing center. BACKGROUND

[0004] Computing power is the ability of a computer device or computing / data center to process information, and is the ability of computer hardware and software to jointly perform certain computing requirements. It is the computing power that achieves target result output by processing information data, and is a new type of productivity that integrates information computing power, network carrying capacity, and data storage power. Computing power infrastructure is an important part of new information infrastructure, and has characteristics such as multi-element ubiquity, intelligence and agility, safety and reliability, and green and low carbon. It is of great significance to help promote industrial transformation and upgrading, empower technological innovation and progress, meet people's needs for a better life, and achieve efficient social governance.

[0005] At present, in a scenario requiring model training, an intelligent computing center usually allocates processor resources to a model, and then processes and calculates the model based on the processor resources to achieve training of the model.

[0006] However, there is no perfect measurement unit for the computing power of the intelligent computing center in the related art, and the computing power of the intelligent computing center cannot be conveniently measured. Only a rough estimate of the computing power of the intelligent computing center can be made, which leads to the situation that the computing power allocated by the intelligent computing center for model training is redundant or insufficient in the related art.

[0007] Therefore, there are problems of low accuracy of processor resource allocation for model training and poor accuracy of computing power measurement of the intelligent computing center in the related art. SUMMARY

[0008] Embodiments of the present disclosure provide a method and device for calculating computing power of a 1-degree computing power measurement intelligent computing center to solve the problem of low accuracy of processor resource allocation for model training and poor accuracy of computing power measurement of the intelligent computing center in the related art.

[0009] To solve the above problems, the present disclosure is implemented as follows:

[0010] In a first aspect, the embodiments of the present disclosure provide a method for measuring computing power of an intelligent computing center by 1-degree computing power, applied to a first intelligent computing center, comprising:

[0011] Step S1, obtaining an initial half-precision floating-point computing power number, an initial GPU bandwidth and an initial GPU capacity of a processor of the first intelligent computing center;

[0012] Step S2, calculating a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, a second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and a third ratio of the initial GPU capacity to a preset GPU capacity, the preset half-precision floating-point computing power number being a half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth being a GPU bandwidth of the preset processor, and the preset GPU capacity being a GPU capacity of the preset processor;

[0013] Step S3, setting a weighted sum of the first ratio, the second ratio and the third ratio, a product of a preset floating-point operation number unit and a preset time unit as an initial 1-degree computing power value of the first intelligent computing center.

[0014] In one embodiment, after the step S3, the method further comprises:

[0015] Step S4, obtaining a model parameter of an initial model;

[0016] Step S5, calculating a required 1-degree computing power value of the initial model based on the model parameter, the required 1-degree computing power value being used to represent computing power required by the initial model for training;

[0017] Step S6, performing model training on the initial model based on the required 1-degree computing power value and the initial 1-degree computing power value.

[0018] In one embodiment, the model parameter comprises a size of the initial model and a size of a training data set;

[0019] The step S5 comprises:

[0020] Step S51, calculating the required 1-degree computing power value of the initial model based on the size of the initial model and the size of the training data set.

[0021] In one embodiment, the step S6 comprises:

[0022] Step S61, calculating a residual 1-degree computing power value of the first intelligent computing center based on the initial 1-degree computing power;

[0023] Step S62, in a case where the remaining 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, allocating a processor resource based on the required 1-degree computing power value;

[0024] Step S63, performing model training on the initial model based on the processor resource.

[0025] In one embodiment, the step S6 further comprises:

[0026] Step S64, in a case where the remaining 1-degree computing power value is less than the required 1-degree computing power value of the initial model, sending the initial model, the model parameter and the required 1-degree computing power value to a second intelligent computing center;

[0027] Step S65, receiving a target model sent by the second intelligent computing center, the target model being obtained by the second intelligent computing center based on the required 1-degree computing power value to allocate the processor resource and based on the processor resource and the model parameter to train the initial model.

[0028] In one embodiment, the step S61 comprises:

[0029] Step S611, obtaining a training 1-degree computing power value of the first intelligent computing center for training other models;

[0030] Step S312, setting a difference between the initial 1-degree computing power value and the training 1-degree computing power value as the remaining 1-degree computing power value.

[0031] In one embodiment, the processor resource comprises a target half-precision floating-point computing power number, a target video memory bandwidth and a target video memory capacity corresponding to the required 1-degree computing power value; and the step S62 comprises:

[0032] Step S621, in a case where the remaining 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, calculating the target half-precision floating-point computing power number, the target video memory bandwidth and the target video memory capacity corresponding to the required 1-degree computing power value;

[0033] Step S622, allocating the target half-precision floating-point computing power number, the target video memory bandwidth and the target video memory capacity.

[0034] In a second aspect, the embodiments of the present disclosure further provide a device for calculating the computing capacity of an intelligent computing center by 1-degree computing power metering, comprising:

[0035] A first obtaining module is configured to obtain an initial half-precision floating-point computing power number, an initial video memory bandwidth and an initial video memory capacity of a processor of a first intelligent computing center;

[0036] The first calculation module is configured to calculate a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, calculate a second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and calculate a third ratio of the initial GPU capacity to a preset GPU capacity, the preset half-precision floating-point computing power number being a half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth being a GPU bandwidth of the preset processor, and the preset GPU capacity being a GPU capacity of the preset processor;

[0037] The setting module is configured to set a weighted sum of the first ratio, the second ratio and the third ratio, a product of a preset floating-point operation number unit and a preset time unit as an initial 1-degree computing power value of the first intelligent computing center.

[0038] In a third aspect, the embodiments of the present disclosure further provide an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps in the method for calculating the computing power of the 1-degree computing center by the 1-degree computing center are implemented.

[0039] In a fourth aspect, the embodiments of the present disclosure further provide a computer readable storage medium for storing a program, and when the program is executed by a processor, the steps in the method for calculating the computing power of the 1-degree computing center by the 1-degree computing center are implemented.

[0040] In a fifth aspect, the embodiments of the present disclosure further provide a computer program product including computer instructions, and when the computer instructions are executed by a processor, the steps in the method for calculating the computing power of the 1-degree computing center by the 1-degree computing center are implemented.

[0041] In the embodiments of the present disclosure, the initial half-precision floating-point computing power number, the initial GPU bandwidth and the initial GPU capacity of the processor of the first intelligent computing center are obtained; a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, a second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and a third ratio of the initial GPU capacity to a preset GPU capacity are calculated, the preset half-precision floating-point computing power number is the half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth is the GPU bandwidth of the preset processor, and the preset GPU capacity is the GPU capacity of the preset processor; and a weighted sum of the first ratio, the second ratio and the third ratio, a preset floating-point operation number unit and a preset time unit are multiplied to obtain an initial 1-degree computing power value of the first intelligent computing center. In this way, the initial 1-degree computing power value of the first intelligent computing center is calculated based on the initial half-precision floating-point computing power number, the initial GPU bandwidth and the initial GPU capacity of the first intelligent computing center, the initial 1-degree computing power value can be used to measure the computing power of the intelligent computing center, the processor resource allocation accuracy for training the model is improved, and the accuracy of measuring the computing power of the intelligent computing center is improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings needed in the description of the embodiments of the present disclosure will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] FIG. 1 is a flowchart of a method for measuring the computing power of an intelligent computing center by 1-degree computing power according to an embodiment of the present disclosure;

[0044] FIG. 2 is a floating-point operation analysis diagram of a BERT model according to an embodiment of the present disclosure;

[0045] FIG. 3 is a structural diagram of a device for measuring the computing power of an intelligent computing center by 1-degree computing power according to an embodiment of the present disclosure;

[0046] FIG. 4 is a structural diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present disclosure will be described clearly and completely with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only some embodiments of the present disclosure, not all embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present disclosure.

[0048] The "computing power" described in the present disclosure is the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to jointly perform certain computing requirements, the computing power that achieves the target result output by processing information data, and a new type of productivity integrating information computing power, network carrying capacity, and data storage power. The computing power infrastructure mainly provides services to society.

[0049] The "computing power (CP)" described in the present disclosure is a kind of ability of a data center server to process data and achieve result output, and is a comprehensive index for measuring the computing power of a data center, including general computing power, supercomputing power, and intelligent computing power. The commonly used unit of measurement is the number of floating point operations per second (FLOPS, wherein, 1EFLOPS=10^18FLOPS), and the larger the value is, the stronger the comprehensive computing power is. According to the estimation, 1EFLOPS is about the computing power output of 5 Tianhe 2A or 500,000 mainstream server CPUs or 2 million mainstream notebooks, and the calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 In the embodiments of the present disclosure, the computing power adopts the half-precision floating-point computing power number (FP16) of a graphics card.

[0050] The "carrying power" (Network Power, NP) described in the present disclosure is a representation of the data transmission capacity of the computing power facility, including the comprehensive ability of network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc. The carrying power involves network transmission inside and between data centers, and is a comprehensive index for measuring the network transmission scheduling ability. In the embodiments of the present disclosure, the carrying power adopts the memory bandwidth. In the embodiments of the present disclosure, the storage power adopts the memory bandwidth.

[0051] The "storage power" (Storage Power, SP) described in the present disclosure is the comprehensive ability of a data center in four aspects of data storage capacity, performance, safety and reliability, and green low carbon, and is a comprehensive index for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and built-in storage devices of servers. The commonly used unit of measurement of storage capacity is exabytes (EB, 1EB=2^60bytes), the commonly used unit of measurement of performance is the number of read / write operations per second per unit capacity (Input / Output Operations Per Second / TB, IOPS / TB), and the disaster recovery ratio is an important performance of safety and reliability. In the embodiments of the present disclosure, the storage power adopts the memory capacity.

[0052] The "computing power infrastructure" described in the present disclosure is a new type of information infrastructure integrating information computing power, network carrying capacity, and data storage power, which can realize centralized computing, storage, transmission, and application of information, and has the characteristics of multi-element, ubiquitous, intelligent, agile, safe, reliable, green, and low carbon, which is of great significance to promote industrial transformation and upgrading, empower China's scientific and technological innovation, meet people's better life, and realize efficient social governance.

[0053] The "degree" described in the present disclosure is a unit of measurement of computing power, and its nature and use are equivalent to "meters", "seconds", "kilograms", "amperes", "candela", "moles", "kelvin", and the like. The "degree" described in the present disclosure is illustrative rather than restrictive, and other units can also be used to measure computing power. Those skilled in the art can make many form extensions under the inspiration of the present disclosure without departing from the scope of the present disclosure and the protection scope of the claims.

[0054] The "1 degree of computing power" described in the present disclosure refers to 1 unit of computing power, which refers to the ability of an intelligent computing center to meet model training within a preset time range. The larger the value, the stronger the ability to meet model training within a preset time range. Alternatively, it refers to the computing power provided by an intelligent computing center during model training. The larger the value, the stronger the computing power within a preset time range, or the larger the value, the less time an intelligent computing center needs to complete the same model training effect, i.e., the higher the efficiency of model training. The nature and use of "1 degree of computing power" described in the present disclosure are equivalent to "1 degree of electricity". Exemplarily, the unit of "1 degree of computing power" can be the number of floating point operations within a preset time, such as "1 TFLOPS h", i.e., 10 trillion floating point operations per second x hours; "312 TFLOPS h", i.e., 312 trillion floating point operations per second x hours; "624 TFLOPS h", i.e., 624 trillion floating point operations per second x hours; "1024 TFLOPS h", i.e., 1024 trillion floating point operations per second x hours; "10000 TFLOPS h", i.e., 10000 trillion floating point operations per second x hours; or "any number of TFLOPS h". The number and "degree" are merely illustrative rather than restrictive, and other units can also be used to measure the computing power of an intelligent computing center. Those skilled in the art can make many form extensions under the inspiration of the present disclosure without departing from the scope of the present disclosure and the protection scope of the claims.

[0055] The "computing power" described in the present disclosure includes general computing power CP 通用 , intelligent computing power CP 智能 , and super computing power CP 超级 .

[0056] The "general computing power" described in the present disclosure refers to the computing power provided by a server based on a central processing unit (CPU) chip, which is used to support basic general computing such as cloud computing and edge computing.

[0057] The "intelligent computing power" described in the present disclosure refers to the large-scale deployment of intelligent computing centers based on graphics processing units (GPUs), field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), and other special chips for various artificial intelligence innovation applications, such as natural language processing, machine vision, and the like.

[0058] The "super computing power" described in the present disclosure is mainly the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for computing in cutting-edge scientific fields, such as planet simulation, drug molecule design, gene analysis, and the like.

[0059] The "intelligent computing center" described in the present disclosure refers to a facility that provides the required computing power, data, and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training, and model inference scenarios) by using large-scale heterogeneous computing resources, including general computing power (CPU: Central Processing Unit) and intelligent computing power (GPU: Graphics Processing Unit, FPGA: Field Programmable Gate Array, ASIC: Application Specific Integrated Circuit, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from bottom computing power to top application enablement.

[0060] The "computing power resource" described in the present disclosure refers to the technology and facilities required for the development of a digital society, including but not limited to computing resources such as CPUs (Central Processing Unit) and GPUs (GPU: Graphics Processing Unit), network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and support and guarantee resources such as wind, fire, water, and electricity.

[0061] Referring to FIG. 1, which is a flowchart of a method for calculating the computing power of a 1-degree intelligent computing center according to an embodiment of the present disclosure, the method is applied to a first intelligent computing center. As shown in FIG. 1, the method comprises the following steps:

[0062] In step S1, the initial half-precision floating-point computing power number, the initial GPU bandwidth, and the initial GPU capacity of the processor of the first intelligent computing center are obtained.

[0063] In step S2, the first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, the second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and the third ratio of the initial GPU capacity to a preset GPU capacity are calculated. The preset half-precision floating-point computing power number is the half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth is the GPU bandwidth of the preset processor, and the preset GPU capacity is the GPU capacity of the preset processor.

[0064] In step S3, the weighted sum of the first ratio, the second ratio, and the third ratio, the product of a preset floating-point operation number unit and a preset time unit are set as the initial 1-degree computing power value of the first intelligent computing center.

[0065] In the embodiment of the present disclosure, the initial half-precision floating-point computing power number, the initial GPU bandwidth, and the initial GPU capacity of the processor of the first intelligent computing center are obtained. The first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, the second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and the third ratio of the initial GPU capacity to a preset GPU capacity are calculated. The preset half-precision floating-point computing power number is the half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth is the GPU bandwidth of the preset processor, and the preset GPU capacity is the GPU capacity of the preset processor. The weighted sum of the first ratio, the second ratio, and the third ratio, the product of a preset floating-point operation number unit and a preset time unit are set as the initial 1-degree computing power value of the first intelligent computing center. In this way, the initial 1-degree computing power value of the first intelligent computing center is calculated based on the initial half-precision floating-point computing power number, the initial GPU bandwidth, and the initial GPU capacity of the processor of the first intelligent computing center. The initial 1-degree computing power value can be used to measure the computing power of the intelligent computing center, thereby improving the accuracy of processor resource allocation for model training and improving the accuracy of measuring the computing power of the intelligent computing center.

[0066] In the embodiments of the present disclosure, the preset half-precision floating-point computing power number is the half-precision floating-point computing power number of the preset processor, and the preset display memory bandwidth is the display memory bandwidth of the preset processor. In the process of calculating the initial 1-degree computing power value of different intelligent computing centers, the same processor is used as the preset processor, so that the calculated 1-degree computing power value can accurately measure the data processing of different processors.

[0067] In some embodiments, the computing power of the processor is directly used as the unit of measuring the computing power of the intelligent computing center. However, the unit of the computing power is the number of floating-point operations per second, and in the actual data processing process of the processor, the speed of the processor processing data is also limited by the display memory bandwidth. Through the floating-point operation analysis of the Bidirectional Encoder Representations from Transformer (BERT) model shown in FIG. 2, it can be determined that the speed of the processor processing data is also limited by the display memory bandwidth and the display memory capacity.

[0068] It should be noted that the computing device computing power is mainly measured according to the computing power device market distribution of each region in the past six years, and is measured from three categories of products, namely general servers, artificial intelligence (AI) servers, and supercomputers, to measure the basic computing power, intelligent computing power, and supercomputing power. The basic computing power mainly focuses on the server computing power of each region, and uses single-precision floating-point number (FP32) computing power to measure the computing power performance; the intelligent computing power mainly focuses on the AI server computing power of each region, and uses the mainstream half-precision floating-point computing power number (FP16) to measure the computing power performance; the supercomputing power is mainly based on the international well-known ranking list TOP500 and reference to the relevant data of supercomputing manufacturers, and uses double-precision floating-point number (FP64) computing power to measure the supercomputing power performance. Therefore, in the embodiments of the present disclosure, the half-precision floating-point computing power number is used to calculate the corresponding initial 1-degree computing power value for different intelligent computing centers.

[0069] Based on the above reasons, the initial 1-degree computing power value of the first intelligent computing center in the present disclosure is calculated by the initial half-precision floating-point computing power number, the initial display memory bandwidth, and the initial display memory capacity of the processor of the first intelligent computing center, so that the initial 1-degree computing power value can more accurately measure the computing power of the intelligent computing center.

[0070] The training scene of a typical BERT (transformer) model is calculated according to the test content of the total time consumption in the model training process in Ivanov A, Dryden N, Ben-Nun T, et al. Data movement is all you need: A case study on optimizing transformers [J]. Proceedings of Machine Learning and Systems, 2021, 3:711-732. The time consumption mainly limited by the GPU computing capability and the time consumption of the process mainly limited by the GPU memory are respectively calculated, and the proportion of each time consumption is about 6:4.

[0071] Exemplarily, the processor NVIDIA A800 80GB SXM is preset as the processor, and the value of the computing power generated in 1 hour is 1 degree of computing power value, and the unit is TFLOPS·h (trillion floating point operations per second × hour). According to the half-precision floating point computing power number, the memory capacity, and the memory bandwidth ratio 6:1:3 [that is, the initial 1 degree of computing power value = (the first ratio × 0.6 + the second ratio × 0.1 + the third ratio × 0.3) × the preset floating point operation number unit × the preset time unit)], the initial 1 degree of computing power value of different models of processors is calculated as shown in the following table:

[0072] In the table, “TFLOPS” is trillion floating point operations per second.

[0073] The initial 1 degree of computing power value of the processor of different intelligent computing centers can be calculated by the above-mentioned manner.

[0074] In the embodiments of the present disclosure, the metering conversion formula of the heterogeneous processor (for example, GPU) and the acceleration card is provided, the computing capability of the processor, the memory bandwidth of the processor, and the weight of the typical work load are considered, so that the unified computing power settlement standard (that is, 1 degree of computing power value) can be used between different intelligent computing centers. This solves the trouble of complex negotiation and calculation between intelligent computing centers in the related art, promotes the unified scheduling of computing power resources, and realizes the connection and integration of computing power resources of different sources and different geographical locations through network technology, and realizes the sharing and optimized utilization of computing power.

[0075] Further, the user can flexibly use various heterogeneous computing power equipment through the intelligent computing center and the artificial intelligence basic software without worrying about the conversion problem of the specific computing power unit, effectively enhancing the flexibility and use optimization of resource configuration, so that various intelligent computing resources can be utilized in the most efficient way, thereby improving the resource utilization rate of the entire intelligent computing center.

[0076] Further, using 1 degree of computing power as a standard unit of measurement helps the widespread collaborative construction and operation of intelligent computing centers, reducing the construction and operation costs between intelligent computing centers. Each intelligent computing center can not only co-construct and share computing power resources, but also can realize accurate matching of computing power demand and supply according to the 1 degree of computing power standard, and fully exert the advantages of regional intelligent computing resources.

[0077] Further, through the 1 degree of computing power value determined computing power service, users do not need to complexly evaluate the detailed specifications of various hardware devices, and can quickly obtain the corresponding computing power resources. This significantly improves the user experience, while also reduces the threshold for users to use computing power services, and promotes the efficient development of artificial intelligence (AI) related research and development work.

[0078] Further, by calculating the initial 1 degree of computing power value, users can simply and intuitively evaluate and compare the actual effect of using different processors (such as GPUs) and accelerator cards in large model training and inference, eliminating the need for complex analysis of various processor hardware and accelerator card specific performance indicators, making it more efficient and convenient for users to choose.

[0079] In the embodiments of the present disclosure, the preset floating point operation frequency unit can be set according to user needs, for example, it can be set to trillion floating point operations per second (TFLOPS), petaflop floating point operations per second (PFLOPS), or exaflop floating point operations per second (EFLOPS) and the like.

[0080] In the embodiments of the present disclosure, the preset time unit can be set according to user needs, for example, it can be set to per second, per minute, or per hour and the like.

[0081] In one embodiment, after the step S3, the method further comprises:

[0082] Step S4, obtaining the model parameters of the initial model;

[0083] Step S5, calculating the demand 1 degree of computing power value of the initial model based on the model parameters, the demand 1 degree of computing power value being used to represent the computing power required for training the initial model;

[0084] Step S6, performing model training on the initial model based on the demand 1 degree of computing power value and the initial 1 degree of computing power value.

[0085] In the embodiments of the present disclosure, the initial model is a model that needs to be trained, which can be a model uploaded by a user to the first intelligent computing center, can be a model designed according to user demand, or can be a model sent by another intelligent computing center to the first intelligent computing center and requesting the first intelligent computing center to train the model.

[0086] In some embodiments, when the initial model is a model uploaded by a user to the first intelligent computing center or a model designed according to user demand, the first intelligent computing center directly sends the trained model to the terminal corresponding to the user after training the initial model; or, the first intelligent computing center retains the trained model and sends the trained model to the terminal corresponding to the user after receiving a request instruction from the terminal corresponding to the user.

[0087] In some embodiments, when the initial model is a model sent by another intelligent computing center to the first intelligent computing center, the first intelligent computing center directly sends the trained model to the other intelligent computing center after training the initial model.

[0088] In the embodiments of the present disclosure, the model parameter is a parameter of the initial model, which is used to represent related information of the initial model, such as the size of the initial model, the structure of the initial model, and the training data set that needs to be trained by the initial model.

[0089] In the embodiments of the present disclosure, the demand 1-degree computing power value of the initial model is calculated based on the model parameter, the demand 1-degree computing power value can represent the computing power required for training the initial model, and the initial model is trained based on the demand 1-degree computing power value and the initial 1-degree computing power value, so that the first intelligent computing center can complete the training of the initial model without allocating too many processor resources, thereby avoiding waste of processor resources.

[0090] In one embodiment, the model parameter includes the size of the initial model and the size of the training data set.

[0091] The step S5 includes:

[0092] Step S51, calculating the demand 1-degree computing power value of the initial model based on the size of the initial model and the size of the training data set.

[0093] In the embodiments of the present disclosure, the demand 1-degree computing power value of the initial model is calculated based on the size of the initial model and the size of the training data set, so that the processor resources allocated based on the demand 1-degree computing power value can meet the size of the initial model and the size of the training data set.

[0094] In some embodiments, the demand 1st FLOPS value can be a product value of the size of the initial model, the size of the training data set, and a preset coefficient. Wherein, the preset coefficient can be determined according to the computing performance of the intelligent computing center, so that different intelligent computing centers can allocate the same processor resources according to the demand 1st FLOPS value.

[0095] In some embodiments, the model parameters can also include the number of iterations of training, and the demand 1st FLOPS value is calculated based on the size of the initial model, the size of the training data set, and the number of iterations of training, so that the processor resources allocated based on the demand 1st FLOPS value can meet the size of the initial model, the size of the training data set, and the number of iterations of training.

[0096] For example, the demand 1st FLOPS value of the initial model can be calculated by the following formula: FLOPS 模型 =k x model parameter quantity x token training number

[0097] Wherein, FLOPS 模型 is the demand 1st FLOPS value, k is a preset coefficient, model parameter quantity represents the size of the initial model, and token training number represents the number of iterations. In some examples, k=8.

[0098] It can be understood that model training includes forward pass and backward pass, counted in unit of unit, 1 unit for one forward pass, and 2 units (output gradient + parameter gradient) for one backward pass, and an additional forward pass is needed to reduce intermediate activation memory using activation recomputation technology. Generally, activation recomputation technology needs to be used, otherwise a single flagship card cannot train a 7B model. Therefore, one complete training: forward pass + backward pass + activation recomputation = 1+2+1=4 units; that is, for each token and each model parameter, 4 units of calculation are needed. Each unit of calculation is a matrix operation, which requires one multiplication and one addition for one matrix operation, a total of 2 floating point operations. Therefore, a total of 8 times (model parameter quantity x total token quantity) calculations are needed, so the above calculation formula is obtained: FLOPS 模型 =8 x model parameter quantity x total token quantity.

[0099] For example, the initial model is llama2-7b model, the total number of parameters is 6.74B, the activation recomputation is enabled, and the total number of tokens of the training data is assumed to be 1T, then: FLOPS 模型= 8 x 6.74B x 1T = 8 x (6.74 x 10^9) x (1 x 10^12) = 53.92 x 10^21 = 5392 x 10^7 TFLOPS h = 53,920,000 PFLOPS h.

[0100] If the training is performed using an A100 card (with a computing power of 312 x 10^12 FLOPS), the corresponding 1-degree algorithmic value of the corresponding computing power consumption is expected to be: 8 x (6.74 x 10^9) x (1 x 10^12) / (312 x 10^12) ≈ 172.8 x 10^6 card x seconds = 48,000 card x hours. Wherein, 1 card x hour = 312 x 10^12 FLOPS h.

[0101] Further, according to the test content of the total time consumption in the model training process in FIG. 2, the time consumption mainly limited by the GPU computing power and the time consumption of the process mainly limited by the GPU memory and bandwidth are respectively counted, and the respective time consumption ratio is about 6:4. Therefore, the corresponding actual training 1-degree algorithmic value is 48,000 / 0.6 = 80,000 card x hours.

[0102] For example, with the processor NVIDIA A800 80GB SXM as the preset processor, the value of the algorithm generated by running for 1 hour is the 1-degree algorithmic value. Therefore, even if the llama2-7b model is trained using 1T token, a total of 80,000 1-degree algorithmic values are consumed.

[0103] It should be noted that after the required 1-degree algorithmic value is calculated, the first intelligent computing center can determine whether to perform model training according to the required 1-degree algorithmic value. Wherein, in the case that the remaining 1-degree algorithmic value of the first intelligent computing center is greater than or equal to the required 1-degree algorithmic value of the initial model, the remaining algorithmic value of the first intelligent computing center can meet the training requirement of the initial model, and at this time, the first intelligent computing center can perform model training on the initial model; and in the case that the remaining 1-degree algorithmic value of the first intelligent computing center is less than the required 1-degree algorithmic value of the initial model, the remaining algorithmic value of the first intelligent computing center cannot meet the training requirement of the initial model, and at this time, the first intelligent computing center can request scheduling algorithmic value from other intelligent computing centers to realize the training of the initial model.

[0104] Specifically, in one embodiment, the step S6 comprises:

[0105] Step S61, calculating the remaining 1-degree algorithmic value of the first intelligent computing center based on the initial 1-degree algorithmic value;

[0106] Step S62, in the case that the remaining 1-degree algorithmic value is greater than or equal to the required 1-degree algorithmic value of the initial model, allocating processor resources based on the required 1-degree algorithmic value;

[0107] Step S63, model training is performed on the initial model based on the processor resource.

[0108] In the embodiments of the present disclosure, by obtaining the residual 1-degree computing power value of the first intelligent computing center, and in the case that the residual 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, the processor resource is allocated based on the required 1-degree computing power value to perform model training on the initial model, which realizes training of the initial data locally by the first intelligent computing center.

[0109] Meanwhile, in one embodiment, the step S6 further includes:

[0110] Step S64, in the case that the residual 1-degree computing power value is less than the required 1-degree computing power value of the initial model, the initial model, the model parameter and the required 1-degree computing power value are sent to the second intelligent computing center;

[0111] Step S65, receiving the target model sent by the second intelligent computing center, the target model being obtained by the second intelligent computing center based on the required 1-degree computing power value to allocate the processor resource, and based on the processor resource and the model parameter to train the initial model.

[0112] In the embodiments of the present disclosure, in the case that the residual 1-degree computing power value is less than the required 1-degree computing power value of the initial model, the initial model, the model parameter and the required 1-degree computing power value are sent to the second intelligent computing center, and the target model sent by the second intelligent computing center is received, the target model being obtained by the second intelligent computing center based on the required 1-degree computing power value to allocate the processor resource and the model parameter to train the initial model. In this way, in the case that the residual 1-degree computing power value is less than the required 1-degree computing power value of the initial model, the first intelligent computing center does not directly train the initial model, but dispatches the computing power of the second intelligent computing center, and the second intelligent computing center trains the initial model, and then sends it to the first intelligent computing center, which realizes the training of the initial model in the case that the computing power of the first intelligent computing center is insufficient.

[0113] In the embodiments of the present disclosure, the residual 1-degree computing power value of the second intelligent computing center needs to be greater than or equal to the required 1-degree computing power value to train the initial model. If the residual 1-degree computing power value of the second intelligent computing center needs to be less than the required 1-degree computing power value, the second intelligent computing center should send an indication of insufficient computing power to the first intelligent computing center to prompt the first intelligent computing center to request other intelligent computing centers to train the initial model.

[0114] For example, llama2-7b model training requires 80000 1-degree computing power values. If the training needs to be completed within 1 month, 80000 / (30*24)≈111*A100 cards are required, that is, the first intelligent computing center needs to provide 111 A100 cards to complete the initial model training. If the first intelligent computing center can only provide 60 A100 cards. In order to complete the task within the specified time, 51 A100 cards can be rented from the second intelligent computing center, which is equivalent to 1-degree computing power value of 51*30*24=36,720 1-degree computing power values. In the formula, "30" means 30 days, and "24" means 24 hours.

[0115] At the same time, the 1-degree computing power value can also estimate the number of cards required for different GPU to achieve computing power scheduling.

[0116] For example, the second intelligent computing center only provides H800 cards, and in order to complete the equivalent computing task within 1 month, 36,720 / (2.4949*30*24)≈20*H800 cards need to be rented, that is, the second intelligent computing center needs to allocate 20 H800 cards for model training at this time. In the formula, "30" means 30 days, and "24" means 24 hours.

[0117] In an embodiment, the step S61 comprises:

[0118] Step S611, obtaining a training 1-degree computing power value of the first intelligent computing center for training other models;

[0119] Step S612, setting the difference between the initial 1-degree computing power value and the training 1-degree computing power value as the remaining 1-degree computing power value.

[0120] In the embodiments of the present disclosure, the training 1-degree computing power value of the first intelligent computing center for training other models is obtained, and the difference between the initial 1-degree computing power value and the training 1-degree computing power value is set as the remaining 1-degree computing power value. In this way, the remaining 1-degree computing power value is calculated by the initial 1-degree computing power value and the training 1-degree computing power value, and the first intelligent computing center can be determined whether the initial model can be trained by the remaining 1-degree computing power value.

[0121] In the embodiments of the present disclosure, the initial 1-degree computing power value is the maximum 1-degree computing power value corresponding to the processor of the first intelligent computing center, that is, the processor resources that can be called when the first intelligent computing center processes all training models do not exceed the processor resources corresponding to the initial 1-degree computing power value.

[0122] It should be noted that there are differences between the processors used by different intelligent computing centers. In order to allocate the same processor resources according to the 1-degree computing power value for different intelligent computing centers, the initial 1-degree computing power value corresponding to each processor needs to be calculated first, and then the processor resources corresponding to the required 1-degree computing power value are allocated to different training models based on the initial 1-degree computing power value.

[0123] In one embodiment, the processor resources include the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity corresponding to the required 1-degree computing power value; and the step S62 includes:

[0124] Step S621, in the case where the remaining 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, calculating the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity corresponding to the required 1-degree computing power value;

[0125] Step S622, allocating the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity.

[0126] In the embodiments of the present disclosure, in the case where the remaining 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity corresponding to the required 1-degree computing power value are calculated; and the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity are allocated for model training of the initial model. In this way, the allocated target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity can meet the training requirements of the initial model, thereby ensuring the training of the initial model while reducing the waste of processor resources.

[0127] In the embodiments of the present disclosure, a fourth ratio of the target half-precision floating-point computing power number to the preset half-precision floating-point computing power number of the preset processor is calculated, a fifth ratio of the target video memory bandwidth to the preset video memory bandwidth of the preset processor is calculated, and a sixth ratio of the target video memory capacity to the preset video memory capacity of the preset processor is calculated. The weighted sum of the fourth ratio, the fifth ratio, and the sixth ratio is the required 1-degree computing power value. The weighted value of the third ratio is the same as the weighted value of the first ratio, and the weighted value of the fourth ratio is the same as the weighted value of the second ratio. For example, the initial 1-degree computing power value=(the first ratio*0.6+the second ratio*0.1+the third ratio*0.3)*preset floating-point operation frequency unit*preset time unit, and the required 1-degree computing power value=(the fourth ratio*0.6+the fifth ratio*0.1+the sixth ratio*0.3)*preset floating-point operation frequency unit*preset time unit.

[0128] Please refer to FIG. 3, which is a structural diagram of an apparatus for measuring computing capability of an intelligent computing center by 1-degree computing power, according to an embodiment of the present disclosure. As shown in FIG. 3, the apparatus 300 for measuring computing capability of an intelligent computing center by 1-degree computing power comprises:

[0129] A first obtaining module 301 is configured to obtain an initial half-precision floating-point computing power number, an initial GPU bandwidth and an initial GPU capacity of a processor of a first intelligent computing center.

[0130] A first calculating module 302 is configured to calculate a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, a second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and a third ratio of the initial GPU capacity to a preset GPU capacity. The preset half-precision floating-point computing power number is a half-precision floating-point computing power number of a preset processor. The preset GPU bandwidth is a GPU bandwidth of the preset processor. The preset GPU capacity is a GPU capacity of the preset processor.

[0131] A setting module 303 is configured to set a weighted sum of the first ratio, the second ratio and the third ratio, a product of a preset floating-point operation number unit and a preset time unit as an initial 1-degree computing power value of the first intelligent computing center.

[0132] In one embodiment, after the setting module 303, the apparatus 300 for measuring computing capability of an intelligent computing center by 1-degree computing power further comprises:

[0133] A second obtaining module is configured to obtain a model parameter of an initial model.

[0134] A second calculating module is configured to calculate a required 1-degree computing power value of the initial model based on the model parameter, where the required 1-degree computing power value is used to represent computing power required by the initial model for training.

[0135] A training module is configured to perform model training on the initial model based on the required 1-degree computing power value and the initial 1-degree computing power value.

[0136] In one embodiment, the model parameter comprises a size of the initial model and a size of a training data set.

[0137] The second calculating module comprises:

[0138] A first calculating unit is configured to calculate the required 1-degree computing power value of the initial model based on the size of the initial model and the size of the training data set.

[0139] In one embodiment, the training module comprises:

[0140] a second calculation unit, configured to calculate a remaining 1-degree computing power value of the first intelligent computing center based on the initial 1-degree computing power;

[0141] an allocation unit, configured to, in a case where the remaining 1-degree computing power value is greater than or equal to a required 1-degree computing power value of the initial model, allocate a processor resource based on the required 1-degree computing power value;

[0142] a training unit, configured to perform model training on the initial model based on the processor resource.

[0143] In an embodiment, the training module further includes:

[0144] a sending unit, configured to, in a case where the remaining 1-degree computing power value is less than the required 1-degree computing power value of the initial model, send the initial model, the model parameter, and the required 1-degree computing power value to a second intelligent computing center;

[0145] a receiving unit, configured to receive a target model sent by the second intelligent computing center, the target model being obtained by the second intelligent computing center based on the required 1-degree computing power value to allocate the processor resource and based on the processor resource and the model parameter to train the initial model.

[0146] In an embodiment, the second calculation unit includes:

[0147] an obtaining sub-unit, configured to obtain a training 1-degree computing power value of the first intelligent computing center for training other models;

[0148] a first calculation sub-unit, configured to set a difference between the initial 1-degree computing power value and the training 1-degree computing power value as the remaining 1-degree computing power value.

[0149] In an embodiment, the processor resource includes a target half-precision floating-point computing power number, a target video memory bandwidth, and a target video memory capacity corresponding to the required 1-degree computing power value; and the training unit includes:

[0150] a second calculation sub-unit, configured to, in a case where the remaining 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, calculate the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity corresponding to the required 1-degree computing power value;

[0151] a training sub-unit, configured to allocate the target half-precision floating-point computing power number, the target video memory bandwidth, and the target video memory capacity.

[0152] The device for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement provided in the embodiments of the present disclosure corresponds to each process of each embodiment of the method for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement, and can achieve the same technical effects. To avoid repetition, details are not described herein.

[0153] It should be noted that the device for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement in the embodiments of the present disclosure can be a device, or a component, an integrated circuit, or a chip in an electronic device.

[0154] The embodiments of the present disclosure further provide an electronic device, referring to FIG. 4, which is a structural schematic diagram of an electronic device provided in the embodiments of the present disclosure. The electronic device includes a memory 401, a processor 402, and a program or instruction stored in the memory 401 and running thereon. When the program or instruction is executed by the processor 402, any step in the embodiments of the method for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement corresponding to FIG. 1 and achieving the same beneficial effects can be implemented. Details are not described herein.

[0155] The processor 402 can be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD).

[0156] Those skilled in the art can understand that all or part of the steps of the above-mentioned method for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement can be completed by program instruction related hardware, and the program can be stored in a readable medium.

[0157] The embodiments of the present disclosure further provide a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, any step in the above-mentioned embodiments of the method for calculating the computing capability of the intelligent computing center by 1-degree computing power measurement corresponding to FIG. 1 can be implemented, and the same technical effects can be achieved. To avoid repetition, details are not described herein. The storage medium can be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0158] The embodiment of the present disclosure further provides a computer program product comprising computer instructions which, when executed by a processor, implement the steps in the method for calculating the computing power of a 1-degree computing center through intelligent computing, as shown in FIG. 1, and achieve the same technical effects. To avoid repetition, details are not described here.

[0159] The terms "first", "second", and the like in the embodiments of the present disclosure are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device comprising a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices. In addition, "and / or" is used in the present disclosure to represent at least one of the connected objects, for example, A and / or B and / or C, which represents 7 cases including A alone, B alone, C alone, A and B both exist, B and C both exist, A and C both exist, and A, B and C all exist.

[0160] It should be noted that in this document, the term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of additional identical elements in the process, method, article or device that includes the element.

[0161] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and necessary general hardware platforms, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present disclosure or the part that contributes to the related art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal (which can be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) execute the methods of various embodiments of the present disclosure.

[0162] The embodiments of the present disclosure are described above with reference to the accompanying drawings, but the present disclosure is not limited to the specific embodiments described above, and the specific embodiments described above are merely illustrative, but not restrictive, and a person of ordinary skill in the art can make many forms under the inspiration of the present disclosure without departing from the purpose of the present disclosure and the scope protected by the claims.

Claims

1. A method for calculating the computing capacity of an intelligent computing center by 1-degree computing power, applied to a first intelligent computing center, the method comprising: Step S1, obtaining an initial half-precision floating-point computing power number, an initial GPU bandwidth and an initial GPU capacity of a processor of the first intelligent computing center; Step S2, calculating a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, a second ratio of the initial GPU bandwidth to a preset GPU bandwidth, and a third ratio of the initial GPU capacity to a preset GPU capacity, the preset half-precision floating-point computing power number being a half-precision floating-point computing power number of a preset processor, the preset GPU bandwidth being a GPU bandwidth of the preset processor, and the preset GPU capacity being a GPU capacity of the preset processor; Step S3, setting a weighted sum of the first ratio, the second ratio and the third ratio, a product of a preset floating-point operation number unit and a preset time unit as an initial 1-degree computing power value of the first intelligent computing center.

2. The method of claim 1, wherein, After the step S3, the method further comprises: Step S4, obtaining a model parameter of an initial model; Step S5, calculating a required 1-degree computing power value of the initial model based on the model parameter, the required 1-degree computing power value being used to represent the computing power required for training of the initial model; Step S6, performing model training on the initial model based on the required 1-degree computing power value and the initial 1-degree computing power value.

3. The method of claim 2, wherein, The model parameter comprises a size of the initial model and a size of a training data set; The step S5 comprises: Step S51, calculating the required 1-degree computing power value of the initial model based on the size of the initial model and the size of the training data set.

4. The method of claim 2 or 3, wherein, The step S6 comprises: Step S61, calculating a residual 1-degree computing power value of the first intelligent computing center based on the initial 1-degree computing power; Step S62, in a case where the residual 1-degree computing power value is greater than or equal to the required 1-degree computing power value of the initial model, allocating a processor resource based on the required 1-degree computing power value; Step S63, performing model training on the initial model based on the processor resource.

5. The method of claim 4, wherein, The step S6 further comprises: Step S64, in a case where the residual 1-degree computing power value is less than the required 1-degree computing power value of the initial model, sending the initial model, the model parameter and the required 1-degree computing power value to a second intelligent computing center; Step S65, receiving a target model sent by the second intelligent computing center, the target model being obtained by the second intelligent computing center based on the required 1-degree computing power value to allocate the processor resource and based on the processor resource and the model parameter to train the initial model.

6. The method of claim 4, wherein, The step S61 comprises: Step S611, obtaining a training 1-degree computing power value of the first intelligent computing center for training of other models; Step S612, setting a difference between the initial 1-degree computing power value and the training 1-degree computing power value as the residual 1-degree computing power value.

7. The method of claim 4, wherein, The processor resource comprises a target half-precision floating-point computing power number, a target GPU bandwidth and a target GPU capacity corresponding to the required 1-degree computing power value; The step S62 comprises: Step S621, in the case where the remaining 1-degree computing power value is greater than or equal to the demand 1-degree computing power value of the initial model, calculating the target half-precision floating-point computing power number, the target video memory bandwidth and the target video memory capacity corresponding to the demand 1-degree computing power value; Step S622, allocating the target half-precision floating-point computing power number, the target video memory bandwidth and the target video memory capacity.

8. An apparatus for calculating the computing power of an intelligent computing center by 1-degree computing power measurement, comprising: a first obtaining module, configured to obtain an initial half-precision floating-point computing power number, an initial video memory bandwidth and an initial video memory capacity of a processor of a first intelligent computing center; a first calculating module, configured to calculate a first ratio of the initial half-precision floating-point computing power number to a preset half-precision floating-point computing power number, a second ratio of the initial video memory bandwidth to a preset video memory bandwidth, and a third ratio of the initial video memory capacity to a preset video memory capacity, the preset half-precision floating-point computing power number being a half-precision floating-point computing power number of a preset processor, the preset video memory bandwidth being a video memory bandwidth of the preset processor, and the preset video memory capacity being a video memory capacity of the preset processor; a setting module, configured to set a product of a weighted sum of the first ratio, the second ratio and the third ratio, a preset floating-point operation number unit and a preset time unit as an initial 1-degree computing power value of the first intelligent computing center.

9. An electronic device comprising: A processor, a memory and a program stored on the memory and executable on the processor, the program, when executed by the processor, implements the steps of the method for calculating the computing power of an intelligent computing center by 1-degree computing power measurement according to any one of claims 1 to 7.

10. A computer-readable storage medium, having a computer program stored thereon, the computer program, when executed by a processor, implements the steps of the method for calculating the computing power of an intelligent computing center by 1-degree computing power measurement according to any one of claims 1 to 7.

11. A computer program product, comprising computer instructions, the computer instructions, when executed by a processor, implement the steps of the method for calculating the computing power of an intelligent computing center by 1-degree computing power measurement according to any one of claims 1 to 7.

Images