KR102481870B1 - Method and apparatus for building a computing environment for artificial intelligence learning - Google Patents

Abstract

The present invention relates to a method and apparatus for constructing a computing environment for artificial intelligence learning. A method for constructing a computing environment according to an embodiment of the present invention is a method performed by a computing device, which uses resource information of a plurality of computing devices including the computing device to create a container in a computing device having available resources. Creating a worker node in the container where learning of the artificial intelligence model is performed, and adjusting hyperparameters of the artificial intelligence model using the learning result of the artificial intelligence model performed in the worker node. It may include further creating a manager node to the container.

Description

Method and apparatus for constructing a computing environment for artificial intelligence learning

The present invention relates to a method and apparatus for constructing a computing environment for artificial intelligence learning. More specifically, it relates to a method and apparatus for constructing a computing environment by a container orchestration system based on distributed computing.

In a conventional computing environment for artificial intelligence learning, it is difficult to efficiently distribute computing resources by configuring a system with a virtual machine. In addition, due to the configuration of the centralized computing system, the use of the network was limited according to the increase in the number of users.

A technical problem to be solved by the present invention is to provide a method and an apparatus for providing an integrated platform capable of efficiently managing container creation, resource allocation, and monitoring.

Another technical problem to be solved by the present invention is to provide a method and apparatus for constructing a computing environment for performing artificial intelligence learning through lightening using container technology.

Another technical problem to be solved by the present invention is to provide a method and apparatus for learning a large-scale artificial intelligence model using a research environment constructed by creating a container.

Another technical problem to be solved by the present invention is to provide a method and apparatus for sharing a research environment constructed by creating a container with other users.

The technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

A method for constructing a computing environment according to an embodiment of the present invention for solving the above technical problem is a method performed by a computing device, using resource information of a plurality of computing devices including the computing device to determine available resources Creating a container on a computing device where there exists, creating a worker node in the container where learning of an artificial intelligence model is performed, and using the learning result of the artificial intelligence model performed in the worker node, the artificial intelligence The method may further include creating a manager node in the container that performs model hyperparameter adjustment.

In one embodiment, the resource information may include CPU, GPU, and storage information of the plurality of computing devices.

In one embodiment, the container includes a first container and a second container, and the first container and the second container share an operating system, but an application running in the first container and the second container Applications running in may be isolated from each other.

In an embodiment, the generating of the container using the resource information may include displaying available resource information of each of a plurality of computing devices.

In one embodiment, the step of creating the manager node may include creating the manager node in a computing device having the maximum number of available resources.

In one embodiment, the manager node may transmit the adjusted hyperparameters to the worker node.

In an embodiment, the method may further include storing information about a computing environment built from the generated container as a file. Here, the method may further include transmitting the file to an external computing device not included in the plurality of computing devices.

In one embodiment, the method may further include transmitting a learning result of the artificial intelligence model performed on the worker node to the manager node and transmitting hyperparameters adjusted by the manager node to the worker node. Here, the transmitting to the manager node includes transmitting the learning result to the manager node even when the storage type of the manager node and the storage type of the worker node are different, and transmitting to the worker node The step may include transmitting the adjusted hyperparameter to the worker node even when the storage type of the manager node and the storage type of the worker node are different. In addition, the transmitting to the manager node includes transmitting the learning result to the manager node using the ScienceDMZ network, and the transmitting to the worker node includes the coordination using the ScienceDMZ network. and transmitting the hyperparameters to the worker node.

In an embodiment, the computing environment establishment apparatus includes a distributed system management module configuring a plurality of computing devices as one cluster, a resource management module managing resources of each of the plurality of computing devices configured in the cluster, and transmission from the resource management module. It may include a container management module that creates a container on a computing device having available resources based on the obtained resource information, and a file management module that stores computing environment information including resource information allocated to the container as a file.

In one embodiment, the distributed system management module manages information transmission between the first computing device and the second computing device included in the cluster, and the type of storage of the first computing device and the storage of the second computing device. It may be a module enabling information transmission between the first computing device and the second computing device even when the types are different.

In one embodiment, the distributed system management module may be a module for managing information transmission between a first computing device and a second computing device included in the cluster configured using the ScienceDMZ network.

In one embodiment, the file management module may be a module that allows the stored files to be viewed by the plurality of computing devices and also allows the stored files to be viewed by an external computing device of the cluster to which authority is given.

1 is a configuration diagram of a computing environment construction system according to an embodiment of the present invention.
2 is a flowchart of a method for constructing a computing environment according to another embodiment of the present invention.
3 is a flowchart of an artificial intelligence learning method using a built computing environment according to another embodiment of the present invention.
4 is a flowchart of a method for sharing a built computing environment with other users according to another embodiment of the present invention.
FIG. 5 is a diagram for describing the container described with reference to FIG. 2 in more detail.
6 is a diagram for explaining the artificial intelligence learning method described with reference to FIG. 3 in more detail.
FIG. 7 is a diagram for explaining a method of sharing the computing environment described with reference to FIG. 4 in more detail.
8 is a diagram for explaining an apparatus for constructing a computing environment according to another embodiment of the present invention.
9 is an exemplary diagram for explaining a computing device capable of implementing an apparatus for constructing a computing environment according to another embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings. However, the technical idea of the present invention is not limited to the following embodiments and can be implemented in various different forms, only the following embodiments complete the technical idea of the present invention, and in the technical field to which the present invention belongs It is provided to fully inform those skilled in the art of the scope of the present invention, and the technical spirit of the present invention is only defined by the scope of the claims.

In adding reference numerals to components of each drawing, it should be noted that the same components have the same numerals as much as possible even if they are displayed on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present invention, the detailed description will be omitted.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used in a meaning commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless explicitly specifically defined. Terminology used herein is for describing the embodiments and is not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase.

In addition, in describing the components of the present invention, terms such as first, second, A, B, (a), and (b) may be used. These terms are only used to distinguish the component from other components, and the nature, order, or order of the corresponding component is not limited by the term. When an element is described as being “connected,” “coupled to,” or “connected” to another element, that element is directly connected or connectable to the other element, but there is another element between the elements. It will be understood that elements may be “connected”, “coupled” or “connected”.

As used herein, "comprises" and/or "comprising" means that a stated component, step, operation, and/or element is the presence of one or more other components, steps, operations, and/or elements. or do not rule out additions.

Hereinafter, some embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a computing environment construction system according to an embodiment of the present invention.

Each component of the computing environment construction system disclosed in FIG. 1 represents functionally differentiated functional elements, and any one or more components may be integrated and implemented in an actual physical environment.

Referring to FIG. 1 , a computing environment construction system according to an embodiment of the present invention may include a computing environment construction device 100 and user devices 200a, 200b, 200c, 200d or less (200 for convenience of explanation). there is. Hereinafter, each component will be described in detail.

The computing environment establishment apparatus 100 may configure a plurality of computing devices as one cluster. More specifically, the computing environment establishment apparatus 100 may perform a task by using the processing capabilities of a plurality of computing devices using distributed computing technology. In particular, tasks requiring large-scale resources such as artificial intelligence model training can be performed. Here, all known distributed computing technologies may be applied to the computing environment construction apparatus 100 according to the present embodiment.

For example, the computing environment construction apparatus 100 may configure a cluster including user device 1 200a, user device 2 200b, user device 3 200c, and user device 4 200d.

Also, the computing environment construction apparatus 100 may create a container by using resource information of a plurality of computing devices included in the cluster. Here, the resource information may include information about CPUs, GPUs, and storage of a plurality of computing devices.

In some embodiments to be described later, artificial intelligence learning may be performed by a node created in a container. A detailed description related to this will be embodied later in the specification.

Also, the computing environment construction apparatus 100 may display available resource information of each of a plurality of computing devices included in the cluster. The computing environment construction device 100 may transmit the available resource information to the user device 200 .

In addition, the computing environment building apparatus 100 may store information about the computing environment built by container creation as a file. Here, the file may be a file in the form of an image.

Also, the computing environment construction apparatus 100 may store information about the computing environment built by container creation as a file, and share the stored file with other users.

The computing environment construction apparatus 100 according to some embodiments of the present invention will be described in detail later with reference to FIGS. 2 to 9 .

Next, the user device 200 may be included in the cluster configured by the computing environment building device 100 . Distributed computing technology is applied so that the user device 200 can partially perform tasks requiring large-scale resources.

The user device 200 may display a platform UI provided by the computing environment construction device 100 . Here, the user device 200 may display resource information of a plurality of computing devices included in the cluster through a platform UI provided by the computing environment construction device 100 . In addition, the user device 200 may display container information generated by the computing environment building device 100 through a platform UI provided by the computing environment building device 100 .

The user device 200 may have a web browser or dedicated application installed to output resource information or container information.

If the user device 200 that can be referred to in some embodiments of the present invention is a device capable of outputting resource information or container information of a plurality of computing devices received from the computing environment construction device 100 through a network, any Any device may be acceptable. For example, the user device 200 that may be referred to in some embodiments of the present invention includes a desktop, a workstation, a server, a laptop, a tablet, and a smart phone. It may be any one of a smart phone or a phablet, but is not limited thereto, and is not limited to, a portable multimedia player (PMP), a personal digital assistant (PDA), or an e-book terminal (E -Book Reader) and the like.

Next, a network connecting the computing environment building device 100 and the user device 200 may be configured as a scienceDMZ network. By configuring the ScienceDMZ network, large amounts of data can be transmitted between containers at high speed. Here, all known technologies related to the scienceDMZ network can be applied to the present invention. However, the present invention is not limited to the scienceDMZ network.

The computing environment construction system shown in FIG. 1 is shown as a network connecting only the computing environment construction device 100 and each user device 200, but the present invention is not limited to FIG. 1, and the user device 200 ), a network can be connected between them to form a system.

Hereinafter, a method for constructing a computing environment according to another embodiment of the present invention will be described with reference to FIG. 2 . 2 is a flowchart of a method for constructing a computing environment according to an embodiment of the present invention.

The method for constructing a computing environment according to the present embodiment may be performed by a computing device. For example, the computing device may be the computing environment construction device 100 described with reference to FIG. 1 . Also, the method according to this embodiment may be performed separately by the first computing device and the second computing device. Hereinafter, in performing each operation of the method according to the present embodiment, if the description of the subject is omitted, the subject may be interpreted as the computing device.

Referring to FIG. 2 , in step S110, a plurality of computing devices are configured as one cluster. Here, a cluster may be configured by including user devices. A method for constructing a computing environment according to an embodiment of the present invention configures a plurality of computing devices as a cluster, so that a task can be performed using the processing capabilities of the plurality of computing devices using distributed computing technology.

Next, in step S120, a container may be created using resource information of a plurality of computing devices included in the cluster. Here, the resource information may include CPU, GPU, and storage information of a plurality of computing devices included in the cluster.

In some embodiments, a container may be created on a computing device where there are available resources of multiple computing devices. Here, the container may be created in one or more computing devices among a plurality of computing devices included in the cluster. In this case, distributed computing technology may be applied. According to the present embodiment, as a container is created in a computing device capable of allocating resources, a task requiring a large amount of resources can be performed in the container.

In some embodiments, there may be multiple containers created in a cluster. At this time, the first container and the second container are created in the cluster, and the first container and the second container share an operating system, but the application process running in the first container and the application process running in the second container are isolated from each other. may have been

In this embodiment, all known container technologies can be applied. Compared to a virtual machine (VM), a container can build a more lightweight system by sharing an operating system (OS) with other containers.

In some embodiments, resource information of each of a plurality of computing devices included in the cluster may be displayed. Here, the resource information may be transmitted to and displayed in the user device. According to this embodiment, a user can check resource information of a plurality of computing devices included in a cluster.

Here, the resource information displayed on the user device may include CPU, GPU, and storage information of a plurality of computing devices. For example, the amount of CPU usage of a computing device included in the cluster may be displayed, and available CPU information according to the amount of CPU usage of the computing device may be displayed. For another example, storage usage of computing devices included in the cluster may be displayed, and available storage information according to the storage usage of the computing devices may be displayed. However, the present invention is not limited to these examples.

In some embodiments, available resource information of each of a plurality of computing devices included in the cluster may be displayed. According to this embodiment, a user may compressively check only available resource information of a plurality of computing devices included in a cluster. For example, only available CPU information according to the CPU usage of the computing device may be displayed without displaying the CPU usage of the computing device included in the cluster. For another example, only available GPU information according to the GPU usage of the computing device may be displayed without displaying the GPU usage of the computing device included in the cluster. However, the present invention is not limited to these examples.

Next, in step S130, a worker node on which learning of an artificial intelligence model is performed is created in a container.

In some embodiments, multiple worker nodes may be created in a container. As the learning of the artificial intelligence model is performed in parallel on different worker nodes, the learning efficiency of the artificial intelligence model can be improved. At this time, learning data is input to the artificial intelligence model from the generated worker node, and the output data, that is, the learning result, may be transmitted to a manager node to be described later.

Next, in step S140, a manager node that performs hyperparameter adjustment of the artificial intelligence model using the learning result of the artificial intelligence model performed in the worker node may be further created as a container.

In some embodiments, one manager node may be created for one container. According to this embodiment, by separating containers for each artificial intelligence learning module and placing a single manager node in the container, the learning efficiency of the artificial intelligence model can be increased. However, the present invention is not limited to this embodiment.

In some embodiments, a manager node may be created on a computing device with the largest available resource within a cluster that includes multiple computing devices. According to this embodiment, a manager node that adjusts and tunes hyperparameters can be created in a computing device whose resources can be maximally utilized.

In some embodiments, the manager node may send the adjusted hyperparameter to the worker node. According to this embodiment, the hyperparameters of the artificial intelligence model may be adjusted by transmitting the adjusted hyperparameters to the worker node.

All known methods may be used for adjusting hyperparameters in this embodiment. For example, adjustment of the learning rate, the C and Sigma values in the support vector model, and the K value in KNN may be included. However, the present invention is not limited to this example.

In the method for constructing a computing environment according to an embodiment of the present invention described with reference to FIG. 2, a container orchestration system based on distributed computing is disclosed. According to an embodiment, since it is not a centralized computing technology in which all systems are connected to a single switch, a problem of network performance degradation does not occur even when a plurality of users access. In addition, according to an embodiment, an environment capable of performing computing in a distributed environment is established, so that artificial intelligence model learning can be performed with high efficiency.

Hereinafter, with reference to FIGS. 3 and 4 , an example of utilizing a computing environment built by an apparatus for constructing a computing environment according to an embodiment of the present invention will be described in detail. 3 is a flowchart of an artificial intelligence learning method using a built computing environment according to another embodiment of the present invention.

Referring to FIG. 3 , in step S210, the learning result of the artificial intelligence model performed on the worker node is transmitted to the manager node. Next, in step S220, the hyperparameters adjusted by the manager node are transmitted to the worker node.

According to the embodiment described with reference to FIG. 3 , artificial intelligence model learning is performed by a worker node and a manager node created in a container.

More specifically, training data is fed into an artificial intelligence model on worker nodes. Data corresponding to the training data input to the artificial intelligence model in the worker node is output. The worker node transmits the output data, the learning result, to the manager node. The manager node receives learning results transmitted from a plurality of worker nodes. The manager node uses the received plurality of learning results to adjust the hyperparameters of the artificial intelligence model. The manager node sends the tuned hyperparameters to the worker nodes. Worker nodes apply the received tuned hyperparameters to the artificial intelligence model.

In some embodiments, even when a storage type of a manager node and a storage type of a worker node are different, information may be transmitted between nodes. That is, in information transmission between containers, information transmission may be possible between heterogeneous storages. In this embodiment, all known technologies related to information sharing and data migration between heterogeneous storages may be applied. According to this embodiment, even when a manager node and a worker node are individually located in a computing device having different types of storage included in a cluster, information transmission between the worker node and the manager node may be possible.

In some embodiments, a network network connecting manager nodes and worker nodes may be configured as a scienceDMZ network. In this embodiment, all known technologies related to scienceDMZ can be applied to the present invention. According to this embodiment, large-capacity data can be transmitted at high speed between containers using the big data highway constructed by the scienceDMZ network.

4 is a flowchart of a method for sharing a built computing environment with other users according to another embodiment of the present invention.

Referring to FIG. 4 , in step S310, information about a computing environment built from a container is stored as a file. Here, the file may be a file in the form of an image.

Also, in some embodiments, a file may mean a file created by tensorflow, pytorch, jupyterlab, jupyterhub, C, XLS, and python. . However, it is not limited to this example, and all methods capable of storing a computing environment built with containers as a single file may be included in the present invention. Here, the file may include all information about resources allocated to the container and data stored and used.

In some embodiments, information about the built computing environment may be stored in a single file. By saving the information about the computing environment as a single file, the user can easily back up the research or work environment.

Users who conduct artificial intelligence research spend a lot of time configuring a computing environment for artificial intelligence research. According to this embodiment, the continuity of the computing environment can be maintained by saving the constructed computing environment as it is and restoring it again.

Next, in step S320, the file is transferred to an external computing device not included in the cluster.

According to this embodiment, the continuity of the computing environment can be maintained by transplanting the built computing environment to a computing device not incorporated into a cluster.

In some embodiments, a file may be transferred to an external computing device authorized by a computing device included in the cluster. By not transmitting a file related to the built computing environment to an unauthorized device, leakage of technology applied to the built computing environment to the outside can be prevented.

Hereinafter, with reference to FIGS. 5 to 7 , a method for constructing a computing environment according to an embodiment of the present invention described in FIGS. 2 to 4 will be described in more detail. FIG. 5 is a diagram for describing the container described with reference to FIG. 2 in more detail.

Referring to FIG. 5 , a container may be created in a computing device having available resources among a plurality of computing devices included in a cluster by the computing environment construction apparatus 100 . At this time, a plurality of nodes may be created in the container.

Referring to the specific example shown in FIG. 5 , node 1 (10a), node 2 (10b), node 3 (10c), node 4 (10d), and node 5 (10e) are created. However, the example described with reference to FIG. 5 is only one example of the present invention, and the present invention is not limited to this example.

In some embodiments, multiple worker nodes may be created and one manager node may be created. For example, node 1 (10a) may be a manager node, and nodes 2 (10b) to 5 (10e) may be worker nodes. By having multiple worker nodes, AI models can be trained in parallel. In addition, by placing a single manager node in a single container, the learning efficiency of the artificial intelligence model can be increased.

6 is a diagram for explaining the artificial intelligence learning method described with reference to FIG. 3 in more detail.

Referring to FIG. 6 , a manager node 30 , a first worker node 20a , a second worker node 20b and a third worker node 20c are shown.

Specifically, the worker nodes 20a, 20b, and 20c input learning data 23 to the artificial intelligence model 21 and output a learning result 25. The worker node transmits the learning result 25 to the manager node 30 .

The manager node uses the received learning result 25 to adjust the hyperparameter, and transmits the adjusted hyperparameter 31 to the worker nodes 20a, 20b, and 20c.

FIG. 7 is a diagram for explaining a method of sharing the computing environment described with reference to FIG. 4 in more detail.

Referring to FIG. 7 , a computing environment 40a built by the method for constructing a computing environment described with reference to FIG. 2 is illustrated. The built computing environment 40a may be stored as one file 41 as described with reference to FIG. 4 .

In this case, a restored computing environment 40b identical to the built computing environment 40a may be constructed using the stored file 41 .

In addition, the computing environment 40a built by the first user 51 may be transmitted to the second user 53 . The computing environment 40a built by the second user 53 may be transmitted to the first user 51 . Here, the same computing environment as the computing environment 40a built by executing the transmitted file may be built.

The files described with reference to FIG. 7 may refer to files created by tensorflow, pytorch, jupyterlab, jupyterhub, C, XLS, and python. there is. However, it is not limited to this example, and all methods capable of storing a computing environment built with containers as a single file may be included in the present invention. Here, the file may include all information about resources allocated to the container and data stored and used.

So far, with reference to FIGS. 2 to 7 , a method for constructing a computing environment according to an embodiment of the present invention and its application fields have been described. Hereinafter, the computing environment construction apparatus 100 according to an embodiment of the present invention will be described in more detail with reference to FIGS. 8 to 9 .

Each component of the computing environment establishment apparatus disclosed in FIG. 8 represents functionally differentiated functional elements, and any one or more components may be integrated and implemented in an actual physical environment.

Referring to FIG. 8 , the apparatus 100 for constructing a computing environment according to an embodiment of the present invention includes a distributed system management module 110, a resource management module 120, a container management module 130, and a file management module 140. can include Hereinafter, each component will be described in detail.

The distributed system management module 110 may configure a plurality of computing devices as one cluster.

In addition, the distributed system management module 110 manages information transmission between the first computing device and the second computing device included in the cluster, and when the storage type of the first computing device and the storage type of the second computing device are different. Even so, it may enable information transmission between the first computing device and the second computing device.

Also, the distributed system management module 110 may manage information transmission between a first computing device and a second computing device included in a cluster configured using the scienceDMZ network.

Next, the resource management module 120 may manage the resources of each of the plurality of computing devices configured in a cluster. Also, the resource management module 120 may display resource information requested by the user on the user device.

Next, the container management module 130 may create a container in a computing device having available resources based on the resource information transmitted from the resource management module. Also, the container management module 130 may delete the created container.

Next, the file management module 140 may store computing environment information including resource information allocated to the container as a file. Also, the file management module 140 may allow a plurality of computing devices included in the cluster to view stored files. In addition, the file management module 140 may allow viewing of stored files even in an external computing device of a cluster to which authority is given.

An integrated platform capable of efficiently managing container creation, resource allocation, and monitoring may be provided by the computing environment construction apparatus according to an embodiment of the present invention.

9 is a hardware configuration diagram illustrating an exemplary computing device 1500 capable of implementing the computing environment construction apparatus 100 according to an embodiment of the present invention.

As shown in FIG. 9, the computing device 1500 loads one or more processors 1510, a bus 1550, a communication interface 1570, and a computer program 1591 executed by the processor 1510. It may include a memory 1530 for storing and a storage 1590 for storing the computer program 1591 . However, only components related to the embodiment of the present invention are shown in FIG. 7 . Therefore, those skilled in the art to which the present invention pertains can know that other general-purpose components may be further included in addition to the components shown in FIG. 7 .

The processor 1510 controls the overall operation of each component of the computing device 1500 . The processor 1510 includes a Central Processing Unit (CPU), a Micro Processor Unit (MPU), a Micro Controller Unit (MCU), a Graphic Processing Unit (GPU), or any type of processor well known in the art. It can be. Also, the processor 1510 may perform an operation for at least one application or program for executing a method according to embodiments of the present invention. Computing device 1500 may include one or more processors.

Memory 1530 stores various data, commands and/or information. Memory 1530 may load one or more programs 1591 from storage 1590 to execute a method according to embodiments of the present invention. The memory 1530 may be implemented as a volatile memory such as RAM, but the technical scope of the present invention is not limited thereto.

The bus 1550 provides a communication function between components of the computing device 1500 . The bus 1550 may be implemented as various types of buses such as an address bus, a data bus, and a control bus.

The communication interface 1570 supports wired and wireless Internet communication of the computing device 1500 . Also, the communication interface 1570 may support various communication methods other than internet communication. To this end, the communication interface 1570 may include a communication module well known in the art.

According to some embodiments, communication interface 1570 may be omitted.

The storage 1590 may non-temporarily store the one or more programs 1591 and various data.

The storage 1590 may be a non-volatile memory such as read only memory (ROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, or the like, a hard disk, a removable disk, or a device well known in the art. It may be configured to include any known type of computer-readable recording medium.

Computer program 1591 may include one or more instructions that when loaded into memory 1530 cause processor 1510 to perform methods/operations in accordance with various embodiments of the invention. That is, the processor 1510 may perform methods/operations according to various embodiments of the present disclosure by executing the one or more instructions.

In the above case, the computing environment construction apparatus 100 according to an embodiment of the present invention may be implemented through the computing device 1500 .

So far, various embodiments of the present invention and effects according to the embodiments have been described with reference to FIGS. 1 to 9 . Effects according to the technical idea of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

The technical idea of the present invention described with reference to FIGS. 1 to 9 so far can be implemented as computer readable code on a computer readable medium. The computer-readable recording medium may be, for example, a removable recording medium (CD, DVD, Blu-ray disc, USB storage device, removable hard disk) or a fixed recording medium (ROM, RAM, computer-equipped hard disk). can The computer program recorded on the computer-readable recording medium may be transmitted to another computing device through a network such as the Internet, installed in the other computing device, and thus used in the other computing device.

In the above, even though all the components constituting the embodiment of the present invention have been described as being combined or operated as one, the technical spirit of the present invention is not necessarily limited to these embodiments. That is, within the scope of the object of the present invention, all of the components may be selectively combined with one or more to operate.

Although actions are shown in a specific order in the drawings, it should not be understood that the actions must be performed in the specific order shown or in a sequential order, or that all depicted actions must be performed to obtain a desired result. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of the various components in the embodiments described above should not be understood as requiring such separation, and the described program components and systems may generally be integrated together into a single software product or packaged into multiple software products. It should be understood that there is

Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art can implement the present invention in other specific forms without changing the technical spirit or essential features. can understand that there is Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. The protection scope of the present invention should be construed according to the claims below, and all technical ideas within the equivalent range should be construed as being included in the scope of the technical ideas defined by the present invention.

Claims (15)

In a method performed by a computing device,
Creating a first container on a first computing device having available resources and creating a second container on a second computing device having available resources by using resource information of a plurality of computing devices including the computing device; , the available resources present in the second computing device are the maximum among the available resources of the plurality of computing devices;
generating a worker node in which learning of an artificial intelligence model is performed in the first container, wherein the plurality of computing devices are configured as one cluster;
A manager node for adjusting hyperparameters of the artificial intelligence model is further created in the second container by using a result of learning the artificial intelligence model performed in the worker node, and the second container includes the manager node. creating only nodes;
A first manager node having a first type of storage transmits an adjusted hyperparameter to a first worker node having a second type of storage, and the first worker node transmits the artificial intelligence performed on the first worker node. transmitting a result of intelligence model learning to the first manager node;
of the cluster including information about the creation of the worker node in the first container created in the first computing device and information about creation of the manager node in the second container created in the second computing device Storing information related to the computing environment as a file; and
Transmitting the stored file to an external computing device that does not belong to the cluster and is authorized by the computing device included in the cluster;
Including the step of transplanting the environment of the cluster using a computing device not incorporated into the cluster using the stored file,
How to build a computing environment. According to claim 1,
The resource information,
Including CPU, GPU and storage information of the plurality of computing devices,
How to build a computing environment. delete According to claim 1,
The step of creating the container using the resource information,
Including displaying available resource information of each of a plurality of computing devices,
How to build a computing environment. delete delete delete delete delete delete According to claim 1,
The step of transmitting to the manager node,
Using the ScienceDMZ network, transmitting the result of the learning to the manager node,
The step of transmitting to the worker node,
Using the ScienceDMZ network, transmitting the adjusted hyperparameter to the worker node,
How to build a computing environment. delete delete delete delete

Images