Method and device for preventing leakage of model development data for cloud platform of inclusive computing power intelligent computing center

Abstract

The application provides a model development data leakage prevention method and device for a universal computing power intelligent computing center cloud platform, and relates to the technical fields of intelligent computing centers, wisdom computing centers and computing power infrastructure. The method comprises the following steps: S1, receiving encrypted data sent by a client; S2, distributing the encrypted data to a first node in a plurality of nodes, wherein the plurality of nodes are nodes deployed by an intelligent computing center cloud platform; S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data; S4, transmitting the decrypted data to a first GPU of the first node; and S5, developing a model based on the decrypted data by using the first GPU. The application can improve the utilization rate of computing power resources and realize the wide application of universal computing power.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent computing centers, wisdom computing centers and computing power infrastructure, in particular to a model development data leakage prevention method and device for a universal computing power intelligent computing center cloud platform. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "wisdom computing centers" have emerged.

[0003] An "intelligent computing center" refers to a facility that uses large-scale heterogeneous computing power resources, including general computing power and intelligent computing power, to provide 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). An intelligent computing center covers facilities, hardware and software, and can provide full-stack capabilities from bottom-layer computing power to top-layer application enablement.

[0004] An "intelligent computing center" includes but is not limited to a "wisdom computing center".

[0005] A "wisdom computing center" is an artificial intelligence computing center that is based on artificial intelligence theory, uses artificial intelligence computing architecture, and provides computing power services, data services and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "wisdom computing centers", is the ability of computer equipment or computing / data centers to process parameters, is the ability of computer hardware and software to work together to perform certain computing needs, is the computing power that achieves the output of target results by processing parameter data, and is a new type of productivity that integrates parameter computing power, network carrying capacity and data storage power, which mainly provides services to society through computing power infrastructure.

[0007] In the model development process of the intelligent computing center, the data of the model development demand needs to be uploaded to the intelligent computing center first, and then the GPU of the intelligent computing center provides computing power resources to develop the model based on the data. In the prior art, in order to prevent the leakage of data, the data needs to be encrypted before being uploaded. After the intelligent computing center receives the encrypted data, the encrypted data is stored and decrypted by the GPU, so as to facilitate subsequent model development. However, in the prior art, the decryption of the data by the GPU occupies a large amount of computing power resources, resulting in a small amount of computing power resources for model development, which makes the utilization rate of the computing power resources very low, or the decryption of the data by the CPU and the storage of the decrypted data need a lot of time to read the decrypted data from the storage in the model development stage, which results in a very low efficiency of model development. At the same time, since the data decryption needs to consume computing power resources, the user also needs to bear the cost of renting the computing power resources when renting the computing power service, which results in a high cost of renting the computing power resources, and it is difficult to realize the wide application of the inclusive computing power.

[0008] It can be seen that the prior art has the problems of low utilization rate of computing power resources of the intelligent computing center or low efficiency of model development, and high cost of renting computing power. SUMMARY

[0009] The embodiments of the present application provide a model development data leakage prevention method and device for an inclusive computing power intelligent computing center cloud platform to solve the problems of low utilization rate of computing power resources of the intelligent computing center or low efficiency of model development, and high cost of renting computing power in the prior art.

[0010] To solve the above problems, the present application is implemented as follows:

[0011] In a first aspect, the embodiments of the present application provide a model development data leakage prevention method for an inclusive computing power intelligent computing center cloud platform, comprising:

[0012] Step S1, receiving encrypted data sent by a client;

[0013] Step S2, distributing the encrypted data to a first node in a plurality of nodes, the plurality of nodes being nodes deployed by the intelligent computing center cloud platform;

[0014] Step S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data;

[0015] Step S4, transmitting the decrypted data to a first GPU of the first node;

[0016] Step S5, developing a model based on the decrypted data by the first GPU.

[0017] In an embodiment, the step S2 comprises:

[0018] The step S21 comprises obtaining a first parameter of each node in the plurality of nodes.

[0019] The step S22 comprises determining the first node from the plurality of nodes based on the first parameter and a preset rule.

[0020] The step S23 comprises assigning the encrypted data to the first node.

[0021] The preset rule comprises one of the following:

[0022] In the case that the first parameter is the number of times of receiving encrypted data, the first node is the node in the plurality of nodes with the least number of times of receiving encrypted data.

[0023] In the case that the first parameter is a load parameter, the first node is the node in the plurality of nodes with the least load parameter.

[0024] In an embodiment, the step S4 comprises:

[0025] The step S41 comprises obtaining a load parameter of the first GPU.

[0026] The step S42 comprises transmitting the decrypted data to the first GPU in the case that the load parameter is less than or equal to a first set load threshold.

[0027] In an embodiment, the method further comprises:

[0028] The step S43 comprises transmitting the decrypted data to a second GPU of a second node in the case that the load parameter is greater than the first set load threshold, the second node being a node in the plurality of nodes other than the first node.

[0029] In an embodiment, the second node comprises a second CPU with abnormal communication transmission, or the load parameter of the second GPU is less than or equal to a second set load threshold, the second set load threshold being less than the first set load threshold.

[0030] In an embodiment, the step S4 comprises:

[0031] The step S41’ comprises transmitting the decrypted data accumulated in the memory of the first CPU to the first GPU in the case that the amount of data of the decrypted data accumulated in the memory of the first CPU reaches a preset data amount threshold.

[0032] In an embodiment, the preset data amount threshold is obtained by:

[0033] obtaining at least one of a batch size parameter, a model parameter quantity and a training parameter corresponding to a to-be-developed model;

[0034] calculating the preset data amount threshold based on at least one of the batch size parameter, the model parameter quantity and the training parameter.

[0035] In a second aspect, an embodiment of the present application further provides a model development data leakage prevention device for a universal computing power intelligent computing center cloud platform, comprising:

[0036] a receiving module configured to receive encrypted data sent by a client;

[0037] a distribution module configured to distribute the encrypted data to a first node in a plurality of nodes, the plurality of nodes being nodes deployed by the intelligent computing center cloud platform;

[0038] a decryption module configured to decrypt the encrypted data based on a first CPU of the first node to obtain decrypted data;

[0039] a transmission module configured to transmit the decrypted data to a first GPU of the first node;

[0040] a development module configured to develop a model based on the decrypted data by using the first GPU.

[0041] In a third aspect, the present application further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the model development data leakage prevention method for the universal computing power intelligent computing center cloud platform according to the first aspect.

[0042] In a fourth aspect, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the steps of the model development data leakage prevention method for the universal computing power intelligent computing center cloud platform according to the first aspect.

[0043] In a fifth aspect, the present application further provides a computer program product comprising computer instructions, wherein the computer instructions, when executed by a processor, implement the steps of the model development data leakage prevention method for the universal computing power intelligent computing center cloud platform according to the first aspect.

[0044] In the present application, a model development data leakage prevention method for a universal computing power intelligent computing center cloud platform is provided, comprising: step S1, receiving encrypted data sent by a client; step S2, distributing the encrypted data to a first node in a plurality of nodes, the plurality of nodes being nodes deployed by the intelligent computing center cloud platform; step S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data; step S4, transmitting the decrypted data to a first GPU of the first node; and step S5, developing a model based on the decrypted data through the first GPU. In this way, the encrypted data is decrypted by the first GPU, which does not occupy the GPU resources of the intelligent computing center cloud platform, so that the first GPU can provide more computing power resources for model development, greatly improving the utilization rate of computing power resources; during the model development process, the decrypted data does not need to be read from the storage, but is directly obtained by the first GPU from the decrypted data decrypted by the first CPU, thereby effectively reducing the preparation time and greatly improving the model development efficiency. At the same time, the user does not need to bear the computing power resources consumed when decrypting the encrypted data, thereby reducing the leasing cost of computing power resources and realizing the wide application of universal computing power. BRIEF DESCRIPTION OF DRAWINGS

[0045] To more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0046] Figure 1 is a flowchart of a model development data leakage prevention method for a universal computing power intelligent computing center cloud platform provided by the embodiments of the present application;

[0047] Figure 2 is a schematic diagram of a data transmission process provided by the embodiments of the present application;

[0048] Figure 3 is a structural diagram of a model development data leakage prevention device for a universal computing power intelligent computing center cloud platform provided by the embodiments of the present application;

[0049] Figure 4 is a structural diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0050] Clearly, the described embodiments are part of the present application and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of the present application.

[0051] The "computing power" described in the present application refers to: the ability of computer equipment or computing / data center to process information, the ability of computer hardware and software to cooperate to perform certain computing requirements, the computing power to achieve target result output by processing information data, and the new productivity integrating information computing power, network carrying capacity and data storage power. The computing power infrastructure mainly provides services to the society.

[0052] The "computing power" (Computational Power, CP) described in the present application refers to: a kind of ability of data center server to process data and realize result output, a comprehensive index for measuring the computing power of data center, including general computing power, supercomputing power and intelligent computing power. Commonly used unit of measurement is the number of floating point operations per second (FLOPS, 1EFLOPS=10^18FLOPS), and the larger the value represents the stronger the comprehensive computing power. According to the estimation, 1 EFLOPS is about 5 Tianhe 2A or 500,000 mainstream server CPUs or 2 million mainstream notebook computing power output. The calculation formula is: CP =CP 通用 +CP 智能 +CP 超级 .

[0053] The "carrying capacity" (Network Power, NP) described in the present application refers to: the performance of data transmission capacity of computing power facility, including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, and comprehensive ability, involving network transmission inside and between data centers, and is a comprehensive index for measuring network transmission scheduling ability.

[0054] The "storage power" (Storage Power, SP) described in the present application refers to: the comprehensive ability of data center in four aspects of data storage capacity, performance, safety and reliability, and green low carbon, which is a comprehensive index for measuring the data storage capacity of 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 exabyte (EB, 1EB=2^60bytes), the commonly used unit of measurement of performance is the number of read / write per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB), and the disaster recovery ratio is an important performance of safety and reliability.

[0055] The "computing power infrastructure" described in the application refers to 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.

[0056] The "new type of information infrastructure" described in the application refers to network infrastructure mainly including 5G network, fiber broadband network, backbone network, international communication network, satellite Internet, etc., computing power infrastructure including data center, general computing power center, intelligent computing center, supercomputing center, etc., and new technology infrastructure including artificial intelligence, blockchain, quantum computing, etc.

[0057] The "computing power" described in the application includes general computing power, intelligent computing power and supercomputing power.

[0058] The "general computing power" described in the application refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0059] The "intelligent computing power" described in the application refers to the large-scale deployment of computing platforms based on GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array) and ASIC (Application Specific Integrated Circuit) special chips for various artificial intelligence innovation applications, such as natural language processing, machine vision, etc.

[0060] The "supercomputing power" described in the application refers to the computing power provided by high-performance computing clusters such as supercomputers, which utilizes the centralized computing resources of multiple computer systems working in parallel and processes extremely complex or data-intensive problems through a dedicated operating system, mainly used for computing in cutting-edge scientific fields such as planet simulation, drug molecule design, gene analysis, etc.

[0061] The "intelligent computing center" described in the application refers to a facility that provides 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) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware and software, and can provide full-stack capabilities from bottom computing power to top application enablement.

[0062] The "intelligent computing center cloud platform" described in the application refers to a cloud computing platform based on the comprehensive services of hardware resources and software resources of the intelligent computing center.

[0063] The "intelligent computing center" described in the present application includes but is not limited to "wisdom center".

[0064] The "wisdom center" described in the present application, i.e. artificial intelligence computing center, is based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power service, data service and algorithm service required by artificial intelligence application.

[0065] The "computing power center" described in the present application refers to a facility with computing power, carrying capacity and storage capacity, mainly composed of infrastructure such as wind, fire, water and electricity and IT hardware and software equipment, including general data center, intelligent computing center, supercomputing center, etc.

[0066] The "supercomputing center" described in the present application refers to supercomputing data center, which is based on supercomputer or large-scale computing cluster, and can provide large-scale computing, storage and network services, etc. It is widely used in aerospace, national defense, oil exploration, climate modeling and genome sequencing, etc.

[0067] The "computing power resource" described in the present application refers to the technology and facilities required for the development of digital society, including but not limited to CPU, GPU, etc. Computing resources, switches, routers, etc. Network resources, storage arrays, distributed storage, etc. Storage resources, firewalls, intrusion detection systems, etc. Security resources, and supporting resources such as wind, fire, water and electricity.

[0068] The "model" described in the present application includes but is not limited to "large language model" and "multimodal large model".

[0069] The "large language model" described in the present application refers to large language model (LLM), which is a language model with large number of parameters, aiming to understand and generate human language, and is trained through a large amount of text data. It can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0070] The "multimodal large model" described in the present application refers to a model that jointly trains text, image, video, audio and other multimodal information, including but not limited to multimodal large language model.

[0071] The "computing power running task" described in the present application refers to a specific workload or job that needs certain computing power support and is executed on computing power resources, usually involving complex data processing, numerical calculation, model training or simulation, etc.

[0072] The "inclusive computing power" of the application refers to providing appropriate and effective computing power services to all social strata and groups with computing power service needs at an affordable cost based on the requirements of opportunity equality and the principle of business sustainability.

[0073] Please refer to Figure 1 , Figure 1 is a flowchart of a model development data leakage prevention method for an inclusive computing power intelligent computing center cloud platform provided by an embodiment of the application, as shown in Figure 1 , comprising the following steps:

[0074] Step S1, receiving encrypted data sent by a client.

[0075] The client is a client for users to upload data. The client can be a terminal or a server. After the user collects or processes data for model development and encrypts it, the client uploads it to the intelligent computing center cloud platform. The intelligent computing center cloud platform decrypts the encrypted data, and then develops the model based on the decrypted data.

[0076] The encrypted data is data encrypted by the client. It should be noted that the encrypted data can be real-time encrypted by the client before transmission, or all the data can be encrypted in advance and then uploaded to the intelligent computing center cloud platform through the client.

[0077] The encrypted data is the smallest unit of data, and the encrypted data cannot be further divided. If the intelligent computing center cloud platform receives multiple encrypted data, it needs to allocate each encrypted data in turn. Different encrypted data can be allocated to the same or different nodes.

[0078] Step S2, distributing the encrypted data to a first node in a plurality of nodes, wherein the plurality of nodes are nodes deployed by the intelligent computing center cloud platform.

[0079] The plurality of nodes are nodes deployed by the intelligent computing center cloud platform. Each node is created based on the computing power resources of the intelligent computing center cloud platform and can provide computing power resources to execute computing power running tasks to realize model development and other functions. As shown in Figure 2 , each node can realize data decryption and model development.

[0080] The encrypted data is distributed to the first node in the plurality of nodes, which can be received by the gateway of the intelligent computing center cloud platform from the client, and then the gateway sends the encrypted data to different nodes (including the first node).

[0081] Step S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data.

[0082] The first CPU can be a real CPU of the intelligent computing center cloud platform, or a virtual CPU created based on the CPU resources of the intelligent computing center cloud platform. It should be noted that the first CPU can provide CPU resources, thereby realizing decryption of the encrypted data by the first CPU based on the first node.

[0083] The first CPU based on the first node decrypts the encrypted data, specifically, the gateway directly allocates the encrypted data to the memory of the first CPU, and then the first CPU decrypts the encrypted data in the memory. The decryption of the encrypted data by the first CPU is real-time decryption, that is, after the encrypted data is allocated to the memory of the first CPU, the first CPU starts to decrypt the allocated encrypted data, so as to avoid long-time non-processing of the encrypted data, which causes the encrypted data to occupy the memory of the first CPU.

[0084] In the present application, the encrypted data is decrypted by the first CPU, which does not occupy the GPU resources of the intelligent computing center cloud platform, so that the intelligent computing center cloud platform can provide more GPU resources for model development, and the utilization rate of computing power resources is greatly improved.

[0085] Step S4, transmitting the decrypted data to the first GPU of the first node.

[0086] The first GPU can be a real GPU of the intelligent computing center cloud platform, or a virtual GPU created based on the GPU resources of the intelligent computing center cloud platform. It should be noted that the first GPU can provide GPU resources, thereby realizing model development based on the decrypted data by the first GPU.

[0087] The decrypted data is transmitted to the first GPU of the first node, specifically, after the first CPU decrypts to obtain the decrypted data, the decrypted data in the memory of the first CPU is transmitted to the memory of the first GPU, and then the first GPU develops the model based on the decrypted data in the memory.

[0088] In some embodiments, the decrypted data is transmitted from the memory of the first CPU to the memory of the first GPU through an expansion (Peripheral Component Interconnect Express (PCIE) interface to realize fast transmission, so that the first GPU can develop the model according to the received decrypted data.

[0089] In the present application, the first CPU directly transmits the decrypted data to the first GPU, without the need to store in the storage space of the intelligent computing center cloud platform, thereby avoiding the first GPU to obtain the decrypted data from the storage space, reducing the time loss of the data preparation process, and greatly improving the efficiency of model development.

[0090] Step S5, model development based on the decrypted data by the first GPU.

[0091] In the present application, a model development data leakage prevention method for a universal computing power intelligent computing center cloud platform is provided, comprising: step S1, receiving encrypted data sent by a client; step S2, distributing the encrypted data to a first node in a plurality of nodes, the plurality of nodes being nodes deployed by the intelligent computing center cloud platform; step S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data; step S4, transmitting the decrypted data to a first GPU of the first node; and step S5, model development based on the decrypted data by the first GPU. In this way, the encrypted data is decrypted by the first GPU, without occupying the GPU resources of the intelligent computing center cloud platform, so that the first GPU can provide more computing power resources for model development, greatly improving the utilization rate of computing power resources; in the model development process, the decrypted data is not read from the storage, but the decrypted data decrypted by the first CPU is directly obtained by the first GPU, thereby effectively reducing the preparation time and greatly improving the model development efficiency. At the same time, the user does not need to bear the computing power resources consumed when decrypting the encrypted data, thereby reducing the leasing cost of computing power resources and realizing the wide application of universal computing power.

[0092] In one embodiment, the step S2 comprises:

[0093] Step S21, obtaining a first parameter of each node in the plurality of nodes;

[0094] Step S22, determining the first node from the plurality of nodes based on the first parameter and a preset rule;

[0095] Step S23, distributing the encrypted data to the first node;

[0096] The preset rule comprises one of the following:

[0097] In the case that the first parameter is the number of times of receiving encrypted data, the first node is the node in the plurality of nodes that has the least number of times of receiving encrypted data;

[0098] In the case that the first parameter is a load parameter, the first node is the node in the plurality of nodes that has the smallest load parameter.

[0099] It should be noted that when the encrypted data is distributed to multiple nodes, there may be a case that the load of different nodes is obviously different due to uneven distribution, so that the efficiency of model development of different nodes is obviously different, which may cause the situation that the computing power resources of part of the nodes are not fully utilized, and the utilization rate of the computing power resources of part of the nodes is low.

[0100] To solve the above problems, in the present application, a preset rule is set to determine the first node to which the encrypted data needs to be distributed from multiple nodes, and then the encrypted data is distributed. In this way, each time the encrypted data is received, the preset rule is confirmed once, and then the distribution is performed, so that the multiple nodes can uniformly receive the encrypted data, the efficiency of model development of each node is close, the situation that the computing power resources of part of the nodes are not fully utilized is avoided, and the utilization rate of the computing power resources is greatly improved.

[0101] The preset rule can be set according to the number of encrypted data, or can be determined according to the load of different nodes.

[0102] In some embodiments, in the process of distributing the encrypted data to multiple nodes, the encrypted data can be distributed to different nodes in sequence, so that each node receives the encrypted data the same number of times or close to the same number of times, so that the load of each node is close, the situation that the computing power resources of part of the nodes are not fully utilized is avoided, and the utilization rate of the computing power resources is greatly improved.

[0103] In another embodiment, in the process of distributing the encrypted data to multiple nodes, the load parameters of each node in the multiple nodes are first obtained, the current load of different nodes is determined directly through the load parameters, and then the encrypted data is distributed to the node with the lowest load parameter, so that the load of each node is close, the situation that part of the nodes with low load are not allocated encrypted data is avoided, the load balance of each node is achieved, and the utilization rate of the computing power resources is improved.

[0104] In another embodiment, the first node can also be determined by the difference between the load parameters of different nodes. For example, two nodes develop models, in the case that the difference between the load parameters of the two nodes is less than or equal to a set threshold, the encrypted data is distributed in the form of the number of times; in the case that the difference between the load parameters of the two nodes is greater than the set threshold, the encrypted data is distributed to the node with lower load parameter.

[0105] In one embodiment, the step S4 comprises:

[0106] Step S41, obtaining the load parameter of the first GPU;

[0107] Step S42, in the case that the load parameter is less than or equal to the first set load threshold, the decrypted data is transmitted to the first GPU.

[0108] It should be noted that the CPU and GPU of different nodes can also realize the transmission of decrypted data through the PICE interface. In this way, if the GPU load of a certain node is large, and the GPU load of other nodes is small, the decrypted data in the CPU memory of the node can be transmitted to the GPU memory of other nodes to balance the GPU load of different nodes and improve the utilization of computing power resources.

[0109] Specifically, in the case that the load parameter is less than or equal to the first set load threshold, the load of the GPU of the first node is at a reasonable level, and there is no case of large load, at this time, the decrypted data is transmitted to the first GPU, and the model development is directly performed by the GPU of the first node.

[0110] In one embodiment, the method further comprises:

[0111] Step S43, in the case that the load parameter is greater than the first set load threshold, the decrypted data is transmitted to the second GPU of the second node, and the second node is a node other than the first node in the plurality of nodes.

[0112] In the present application, in the case that the load parameter is greater than the first set load threshold, the first GPU of the first node has a large load, at this time, the decrypted data in the first CPU memory needs to be sent to the second GPU memory to improve the utilization of computing power resources of the second GPU.

[0113] Among them, the second node can be the node with the smallest load parameter corresponding to the GPU, and the decrypted data is transmitted to the GPU memory of the node to improve the utilization of computing power resources of the node.

[0114] It should be noted that, in the case that the load parameter corresponding to the GPU of the second node is greater than or equal to the load parameter corresponding to the first GPU of the first node, or the load parameter corresponding to the GPU of all nodes is greater than the first set load threshold, it indicates that the load of all nodes is at a high level at this time, at this time, the decrypted data in the first CPU memory can not need to be transmitted across nodes, and the decrypted data in the first CPU memory is directly transmitted to the first GPU memory, and the first GPU develops the model based on the decrypted data.

[0115] In one embodiment, the second CPU of the second node communicates abnormally, or the load parameter of the second GPU is less than or equal to the second set load threshold, and the second set load threshold is less than the first set load threshold.

[0116] In the present application, the decrypted data in the CPU memory can be transmitted to the GPU memory across nodes, based on which the communication transmission abnormality problem can be solved. It should be noted that during the encrypted data distribution process, the CPU of some nodes may have communication transmission abnormality with the gateway, at which time the encrypted data cannot be obtained from the gateway by the CPU of the nodes, and the gateway can only distribute the encrypted data to the CPU memory of other nodes. This situation will cause the GPU of the nodes that cannot obtain the encrypted data to be unable to develop the model, and the load of other nodes is large. Therefore, in the present application, the node with communication transmission abnormality is taken as the second node, and the decrypted data in the first CPU memory is transmitted to the second GPU memory, so that the second GPU can develop the model, and the utilization rate of the computing power resources of the second GPU is improved.

[0117] Further, the GPU with low load can also be taken as the second GPU to improve the utilization rate of the computing power resources of the second GPU. Wherein, whether the load of the GPU is low is determined by the second set load threshold. Specifically, in the case that the load parameter of the GPU is less than or equal to the second set load threshold, the load of the GPU is low, and the GPU can be taken as the second GPU; in the case that the load parameter of the GPU is greater than the second set load threshold, the load of the GPU is normal, and the GPU is not taken as the second GPU.

[0118] Wherein, the second set load threshold is less than the first set load threshold, and the first set load threshold is used to determine whether the load of the CPU is too high, and the second set load threshold is used to determine whether the load of the CPU is too low.

[0119] In one embodiment, the step S4 comprises:

[0120] Step S41', in the case that the data amount of the decrypted data accumulated in the memory of the first CPU reaches a preset data amount threshold, the decrypted data accumulated in the memory of the first CPU is transmitted to the first GPU.

[0121] It should be noted that during the process of transmitting the encrypted data to the CPU memory, the transmission is realized through the Internet, and the transmission rate is low; and during the process of transmitting the decrypted data from the CPU memory to the GPU memory, the transmission is realized through the PCIE interface, and the transmission rate is high, which leads to that during the whole data transmission process, the GPU needs to wait for the CPU to transmit the decrypted data before developing the model, at which time the GPU occupies the computing power resources but does not develop the model, resulting in waste of the computing power resources.

[0122] In order to solve the above problems, in the present application, after the encrypted data is transmitted to the first CPU, the decrypted data is first saved in the CPU memory until the accumulated amount of decrypted data in the memory reaches the preset data amount threshold, and then the accumulated decrypted data is transmitted to the first GPU. At this time, the first GPU can receive more decrypted data and can directly start model development without waiting, thereby maximizing the utilization rate of computing power resources.

[0123] In one embodiment, the preset data amount threshold is obtained by the following method:

[0124] Obtaining at least one of a batch size parameter, a model parameter quantity and a training parameter corresponding to the to-be-developed model;

[0125] Based on at least one of the batch size parameter, the model parameter quantity and the training parameter, the preset data amount threshold is calculated.

[0126] In the present application, the preset data amount threshold is positively correlated with the batch size parameter, the model parameter quantity and the training parameter, and the preset data amount threshold can be calculated based on at least one of the batch size parameter, the model parameter quantity and the training parameter by a formula or a model. For example, the batch size parameter, the model parameter quantity and the training parameter are weighted to calculate the preset data amount threshold.

[0127] In addition, the preset data amount threshold is also limited by the GPU memory, that is, the preset data amount threshold cannot exceed the upper limit value of the GPU memory, so as to avoid program errors caused by data overflow and ensure the stability of the model development process.

[0128] Please refer to Figure 3 , Figure 3 is a structural diagram of a model development data leakage prevention device for a universal computing power intelligent computing center cloud platform provided by an embodiment of the present application, as shown in Figure 3 The model development data leakage prevention device 300 for the universal computing power intelligent computing center cloud platform comprises:

[0129] The receiving module 301 is configured to receive encrypted data sent by a client;

[0130] The distribution module 302 is configured to distribute the encrypted data to a first node in a plurality of nodes, wherein the plurality of nodes are nodes deployed by an intelligent computing center cloud platform;

[0131] The decryption module 303 is configured to decrypt the encrypted data based on a first CPU of the first node to obtain decrypted data;

[0132] The transmission module 304 is configured to transmit the decrypted data to a first GPU of the first node;

[0133] a development module 305, configured to develop a model based on the decrypted data by the first GPU.

[0134] In one embodiment, the distribution module 302 comprises:

[0135] a first obtaining unit, configured to obtain a first parameter of each node in the plurality of nodes;

[0136] a determination unit, configured to determine the first node from the plurality of nodes based on the first parameter and a preset rule;

[0137] a distribution unit, configured to distribute the encrypted data to the first node;

[0138] The preset rule comprises one of the following:

[0139] In a case where the first parameter is a number of times of receiving encrypted data, the first node is a node in the plurality of nodes with the least number of times of receiving encrypted data;

[0140] In a case where the first parameter is a load parameter, the first node is a node in the plurality of nodes with the least load parameter.

[0141] In one embodiment, the transmission module 304 comprises:

[0142] a second obtaining unit, configured to obtain a load parameter of the first GPU;

[0143] a first transmission unit, configured to transmit the decrypted data to the first GPU in a case where the load parameter is less than or equal to a first set load threshold.

[0144] In one embodiment, the transmission module 304 further comprises:

[0145] a second transmission unit, configured to transmit the decrypted data to a second GPU of a second node in a case where the load parameter is greater than the first set load threshold, the second node being a node in the plurality of nodes other than the first node.

[0146] In one embodiment, the second node comprises a second CPU with abnormal communication transmission, or a load parameter of the second GPU is less than or equal to a second set load threshold, the second set load threshold being less than the first set load threshold.

[0147] In one embodiment, the transmission module 304 comprises:

[0148] A third transmission unit is configured to transmit the decrypted data accumulated in the memory of the first CPU to the first GPU when the amount of the decrypted data accumulated in the memory of the first CPU reaches a preset data amount threshold.

[0149] In one embodiment, the preset data amount threshold is obtained by the following way:

[0150] At least one of a batch size parameter, a model parameter quantity and a training parameter corresponding to the to-be-developed model is acquired.

[0151] The preset data amount threshold is calculated based on at least one of the batch size parameter, the model parameter quantity and the training parameter.

[0152] The model development data leakage prevention device for the inclusive computing power intelligent computing center cloud platform provided by the embodiments of the present application corresponds to each process of each embodiment of the model development data leakage prevention method for the inclusive computing power intelligent computing center cloud platform described above, and can achieve the same technical effects.

[0153] It should be noted that the model development data leakage prevention device for the inclusive computing power intelligent computing center cloud platform in the embodiments of the present application can be a device, a component in an electronic device, an integrated circuit or a chip.

[0154] The present application also provides an electronic device, referring to Figure 4 , Figure 4 is a structural schematic diagram of an electronic device provided by the embodiments of the present application, and the electronic device comprises a memory 401, a processor 402 and a program or instruction stored in the memory 401 and running, which can realize the model development data leakage prevention method for the inclusive computing power intelligent computing center cloud platform when the program or instruction is executed by the processor 402. Figure 1 Any step in the corresponding model development data leakage prevention method for the inclusive computing power intelligent computing center cloud platform and the same beneficial effects can be achieved, and thus will not be repeated here.

[0155] The processor 402 can be a CPU, an ASIC, an FPGA or a GPU.

[0156] Those skilled in the art can understand that all or part of the steps of the model development data leakage prevention method for the inclusive computing power intelligent computing center cloud platform described above can be completed by program instruction related hardware, and the program can be stored in a readable medium.

[0157] The present application also provides a readable storage medium, and the readable storage medium stores a computer program, and the computer program can realize the model development data leakage prevention method for the inclusive computing power intelligent computing center cloud platform described above when executed by a processor. Figure 1Corresponding to any step in the model development data leakage prevention method embodiment of the cloud platform for the intelligent computing center of the computing power for the benefit of the people, and the same technical effect can be achieved, to avoid repetition, here will not repeat. The storage medium, such as read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.

[0158] The application also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1 Corresponding to each process of the model development data leakage prevention method embodiment of the cloud platform for the intelligent computing center of the computing power for the benefit of the people, and the same technical effect can be achieved, to avoid repetition, here will not repeat.

[0159] The terms "first", "second", and the like in the present application are used to distinguish similar objects, and do not necessarily be used to describe a specific order or sequence. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily be limited to those clearly listed steps or units, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices. In addition, "and / or" is used in the present application 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 paper, the term "include", "contain" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0161] Through the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned example methods can be implemented by means of software and necessary universal hardware platforms, and of course can also be implemented by hardware, but in many cases the former is a better implementation. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product in essence or in the part that contributes to the prior art, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions to make 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 application.

[0162] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative and not restrictive. Those skilled in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the scope protected by the claims.

Claims

1. A model development data leakage prevention method for a cloud platform of a benevolent computing power intelligent computing center, characterized in that, The method comprises the following steps: Step S1, receiving encrypted data sent by a client; Step S2, distributing the encrypted data to a first node in a plurality of nodes, wherein the plurality of nodes are nodes deployed by an intelligent computing center cloud platform; Step S3, decrypting the encrypted data based on a first CPU of the first node to obtain decrypted data; Step S4, transmitting the decrypted data to a first GPU of the first node; Step S5, developing a model based on the decrypted data by using the first GPU; The step S4 comprises: Step S41, obtaining a load parameter of the first GPU; Step S42, in a case where the load parameter is less than or equal to a first set load threshold, transmitting the decrypted data to the first GPU; The method further comprises: Step S43, in a case where the load parameter is greater than the first set load threshold, transmitting the decrypted data to a second GPU of a second node, wherein the second node is a node other than the first node in the plurality of nodes; The second node comprises a second CPU that has a communication transmission exception, or a load parameter of the second GPU is less than or equal to a second set load threshold, wherein the second set load threshold is less than the first set load threshold.

2. The method of claim 1, wherein, The step S2 comprises: Step S21, obtaining a first parameter of each node in the plurality of nodes; Step S22, determining the first node from the plurality of nodes based on the first parameter and a preset rule; Step S23, distributing the encrypted data to the first node; The preset rule comprises one of the following: In a case where the first parameter is a number of times of receiving encrypted data, the first node is a node in the plurality of nodes that has the least number of times of receiving encrypted data; In a case where the first parameter is a load parameter, the first node is a node in the plurality of nodes that has the smallest load parameter.

3. The method of claim 1, wherein, The step S4 comprises: Step S41', in a case where a data amount of the decrypted data accumulated in a memory of the first CPU reaches a preset data amount threshold, transmitting the decrypted data accumulated in the memory of the first CPU to the first GPU.

4. The method of claim 3, wherein, The preset data amount threshold is obtained by: Obtaining at least one of a batch size parameter, a model parameter quantity and a training parameter corresponding to a model to be developed; Based on at least one of the batch size parameter, the model parameter quantity and the training parameter, the preset data amount threshold is calculated.

5. A model development data leakage prevention device for a universal computing power intelligent computing center cloud platform, characterized in that, The method comprises: A receiving module, configured to receive encrypted data sent by a client; A distributing module, configured to distribute the encrypted data to a first node in a plurality of nodes, wherein the plurality of nodes are nodes deployed by an intelligent computing center cloud platform; A decrypting module, configured to decrypt the encrypted data based on a first CPU of the first node to obtain decrypted data; A transmitting module, configured to transmit the decrypted data to a first GPU of the first node; A developing module, configured to develop a model based on the decrypted data by using the first GPU; The transmitting module comprises: A second obtaining unit, configured to obtain a load parameter of the first GPU; The first transmission unit is configured to transmit the decrypted data to the first GPU when the load parameter is less than or equal to a first set load threshold. The transmission module further comprises: The second transmission unit is configured to transmit the decrypted data to a second GPU of a second node when the load parameter is greater than the first set load threshold, the second node being a node other than the first node among the plurality of nodes. The second node comprises a second CPU communication transmission exception, or a load parameter of the second GPU is less than or equal to a second set load threshold, the second set load threshold being less than the first set load threshold.

6. An electronic device, comprising: The method comprises: A processor, a memory, and a program stored on the memory and executable on the processor, the program being executed by the processor to implement the steps of the model development data leakage prevention method for the smart computing center cloud platform facing the algorithm power of the preferential treatment according to any one of claims 1 to 4.

7. A computer readable storage medium characterized by, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the model development data leakage prevention method for the smart computing center cloud platform facing the algorithm power of the preferential treatment according to any one of claims 1 to 4.

8. A computer program product, characterised in that, The computer program comprises computer instructions, and the computer instructions are executed by the processor to implement the steps of the model development data leakage prevention method for the smart computing center cloud platform facing the algorithm power of the preferential treatment according to any one of claims 1 to 4.

Images