KR102297576B1 - Method of managing automatically a deep learning library and apparatuses performing the same - Google Patents

Abstract

Disclosed are a method for automatically managing a deep learning library and an apparatus for performing the same. A method for automatically managing a deep learning library operating for each virtual operating environment according to an embodiment is based on first information about a plurality of deep learning libraries, the version of one or more deep learning libraries among the plurality of deep learning libraries Updating the first configuration corresponding to the version based on the first information and second information about the deep learning server on which the one or more deep learning libraries will be executed corresponding to the updated version of the one or more deep learning libraries changing to the second configuration.

Description

A method for automatically managing a deep learning library and devices for performing the same

The following embodiments relate to a method for automatically managing a deep learning library and apparatuses for performing the same.

Recently, along with the 4th industrial revolution, there has been a lot of interest and research on artificial intelligence technology. In line with this trend, various research groups and companies are developing and distributing various deep learning libraries that can use artificial intelligence technology.

Deep learning libraries are used with great interest by various users. In deep learning libraries, bug issues occur and fix very actively. Each deep learning library provider responds to many issues, and the distributed library tends to be upgraded for various reasons within a short period of time, such as redistribution after bugfixes, and version upgrades for function updates.

Each deep learning library has different specifications required for use, and even within the same deep learning library, there are many differences when the version changes. It is very unnecessary, inconvenient, and time-consuming for users to modify their development environment to match the version of the deep learning library every time.

Embodiments may provide a technique for automatically managing one or more deep learning libraries among a plurality of deep learning libraries and a configuration corresponding to one or more deep learning libraries.

However, the technical problems are not limited to the above-described technical problems, and other technical problems may exist.

A deep learning library management method of an apparatus for automatically managing a deep learning library operating for each virtual operating environment according to an embodiment is based on first information on a plurality of deep learning libraries, one or more of the plurality of deep learning libraries updating the version of the deep learning library, and configuring the first configuration corresponding to the version based on the first information and second information about the deep learning server on which the one or more deep learning libraries will be executed, the one or more deep learning libraries and changing to a second configuration corresponding to the updated version of .

The first information may include the name of each deep learning library, the latest version of each deep learning library, and information on a configuration corresponding to the latest version.

The second information may include information about the hardware of the deep learning server.

The updating includes comparing the version of the one or more deep learning libraries with the latest version of the one or more deep learning libraries in the first information to determine whether the version of the one or more deep learning libraries is the latest version; If the version of the one or more deep learning libraries is not the latest version, it may include updating the version of the one or more deep learning libraries to the latest version.

The deep learning library management method further comprises the step of searching for the one or more deep learning libraries corresponding to the input signal from among the plurality of deep learning libraries in response to an input signal of a user who will use the one or more deep learning libraries, and , the one or more deep learning libraries may be a deep learning library to be used by the user among the plurality of deep learning libraries.

The user's input signal may include information about a virtual operating environment to be used by the user, a deep learning library to be used by the user, and a version of the deep learning library to be used by the user.

The first configuration may be hardware, software, program and library compatible with the version.

The second component may be hardware, software, program and library compatible with the updated version.

The changing may include comparing the first configuration with the first information and the second information to determine whether the first configuration is a configuration compatible with the updated version, wherein the first configuration is the updated version is not compatible, changing the first configuration to the second configuration.

The deep learning library management method may further include distributing the updated version and a configuration corresponding to the updated version to a virtual operating environment in which the one or more deep learning libraries will be executed.

An apparatus for automatically managing a deep learning library operating for each virtual operating environment according to an embodiment includes a memory including instructions, and a controller for executing the instructions, wherein the controller is a control unit for a plurality of deep learning libraries. Update the version of one or more deep learning libraries among the plurality of deep learning libraries based on 1 information, and the version based on the first information and second information about the deep learning server on which the one or more deep learning libraries will be executed Change the first configuration corresponding to the second configuration corresponding to the updated version of the one or more deep learning libraries.

The first information may include the name of each deep learning library, the latest version of each deep learning library, and information on a configuration corresponding to the latest version.

The second information may include information about the hardware of the deep learning server.

The controller compares the version of the one or more deep learning libraries with the latest version of the one or more deep learning libraries in the first information to determine whether the version of the one or more deep learning libraries is the latest version, and the one or more deep learning libraries If the version of the learning library is not the latest version, the version of the one or more deep learning libraries may be updated to the latest version.

The controller searches for the one or more deep learning libraries corresponding to the input signal from among the plurality of deep learning libraries in response to an input signal of a user who will use the one or more deep learning libraries, wherein the one or more deep learning libraries include It may be a deep learning library to be used by the user among a plurality of deep learning libraries.

The user's input signal may include information about a virtual operating environment to be used by the user, a deep learning library to be used by the user, and a version of the deep learning library to be used by the user.

The first configuration may be hardware, software, program and library compatible with the version.

The second component may be hardware, software, program and library compatible with the updated version.

the controller compares the first configuration with the first information and the second information to determine whether the first configuration is a configuration compatible with the updated version, wherein the first configuration is not compatible with the updated version Otherwise, the first configuration may be changed to the second configuration.

The controller may distribute the updated version and a configuration corresponding to the updated version to a virtual operating environment in which the one or more deep learning libraries will be executed.

1 shows a schematic block diagram of a deep learning library management system according to an embodiment.
Figure 2 shows a schematic block diagram of the deep learning library management apparatus of Figure 1.
3 illustrates an example for explaining a version upgrade operation and a configuration change operation of the controller illustrated in FIG. 2 .
4A shows an example for explaining the operation of the version check module.
4B shows an example for explaining Table 1 shown in FIG. 4A.
4C shows an example for explaining Table 1_sub shown in FIG. 4A.
5A shows an example for explaining the operation of the hardware resource status check module.
5B shows an example for explaining Table 2 shown in FIG. 5A.
6A shows an example for explaining the operation of the container manager module.
6B shows an example for explaining Table 3 shown in FIG. 6A.
6C shows an example for explaining Table 3_sub indicating detailed information on containers in Table 3. Referring to FIG.
7 shows an example for explaining the operation of the container status check module.
8 shows an example for explaining the operation of the distribution module.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the patent application is not limited or limited by these embodiments. It should be understood that all modifications, equivalents and substitutes for the embodiments are included in the scope of the rights.

The terms used in the examples are used for description purposes only, and should not be construed as limiting. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present specification, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

In addition, in the description with reference to the accompanying drawings, the same components are given the same reference numerals regardless of the reference numerals, and the overlapping description thereof will be omitted. In the description of the embodiment, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the embodiment, the detailed description thereof will be omitted.

In addition, in describing the components of the embodiment, terms such as first, second, A, B, (a), (b), etc. may be used. These terms are only for distinguishing the components from other components, and the essence, order, or order of the components are not limited by the terms. When it is described that a component is “connected”, “coupled” or “connected” to another component, the component may be directly connected or connected to the other component, but another component is between each component. It will be understood that may also be "connected", "coupled" or "connected".

Components included in one embodiment and components having a common function will be described using the same names in other embodiments. Unless otherwise stated, a description described in one embodiment may be applied to another embodiment, and a detailed description in the overlapping range will be omitted.

1 shows a schematic block diagram of a deep learning library management system according to an embodiment.

In the past, a deep learning library was used through the following method.

1) The user directly checked the version of the deep learning library to be used by the user. For example, the user went to the Tensorflow official homepage and checked the version of the deep learning library the user would use.

2) In order to use the deep learning library in the user's work environment, the user directly set the hardware suitable for the version of the deep learning library to be used by the user. For example, the user installed the hardware according to the Graphic Driver, CUDA, and cudnn versions to use the GPU, and set the environment variables of the hardware.

3) In order to use the deep learning library in the user's work environment, etc., the user directly set the software, program and library required to execute the version of the deep learning library to be used by the user. At this time, the user installed the software, program, and library by checking the version of the development language and required library to be used by the user, and set environment variables of the software, program, and library. For example, the user installed python = 3.7, installed oepncv = 4.2.0.32, and then set the environment variable.

4) When the version of the deep learning library used by the user is changed, the user directly re-performs the processes 1), 2) and 3) to reset the working environment for using the deep learning library. At this time, the reasons for changing (or upgrading) the version of the deep learning library may be various. For example, the deep learning library must upgrade the version of the deep learning library due to a hotfix update that solves a critical bug issue, a version upgrade due to an overall API update, and an update due to a dependent language and library update.

Even if the GPU used for deep learning in the user's server or PC is changed, in the past, to use the deep learning library, it was necessary to check whether the changed GPU can be used in the user's existing work environment (or whether it is compatible). If the changed GPU is not compatible with the existing working environment, the user had to re-perform the steps 1), 2) and 3) to reset the working environment to use the deep learning library. For example, when changing from Nvidia 1080 Ti GPU to Nvidia 2080 Ti GPU, you need to change the hardware settings and upgrade the deep learning library version to the corresponding deep learning library version. If the version upgrade and work environment reset are not performed, there is a problem that the deep learning library cannot be used because the hardware settings and the version of the deep learning library are not compatible.

In other words, the emergence of various deep learning libraries helps a lot in deep learning research, but the user has to do detailed work for each library, such as installing programs and software, and setting up hardware. In addition, due to the rapid development of the deep learning library, the user has to frequently upgrade the version of the deep learning library, and there is an inconvenience of having to do detailed work for it every time.

Hereinafter, a deep learning library management system 10 capable of automatically managing a deep learning library operating for each virtual operating environment in order to overcome the above-described problems will be described. The virtual operating environment is a virtual operating environment for calibrating the independent deep learning development environment, and may be various deep learning development environments such as containers, conda, docker, and virtualenv. Hereinafter, for convenience of description, it is assumed that the virtual operating environment is a container.

The deep learning library management system 10 includes a user device 100 , a deep learning library information providing device 300 , and a deep learning library management device 500 .

The user device 100 may be a device used by a user. For example, the user device 100 may be various devices such as a personal computer (PC), a data server, or a portable electronic device. Portable electronic devices include a laptop computer, a mobile phone, a smart phone, a tablet PC, a mobile internet device (MID), a personal digital assistant (PDA), and an enterprise digital assistant (EDA). ), digital still camera, digital video camera, PMP (portable multimedia player), PND (personal navigation device or portable navigation device), handheld game console, e-book (e-book) may be implemented as a smart device. In this case, the smart device may be implemented as a smart watch or a smart band.

The user device 100 may generate a user's input signal based on the user's input for using the deep learning library operating for each container, and transmit the user's input signal to the deep learning library management apparatus 500 .

The deep learning library information providing apparatus 300 is a deep learning library ecosystem, and may provide various information about each of a plurality of deep learning libraries to the deep learning library management apparatus 500 . For example, the deep learning library ecosystem could be sites of various vendors providing information about deep learning libraries. Deep learning libraries can be as diverse as TensorFlow, Theano, Keras, Caffe, Torch, DL4j (deep learning 4J), MxNet, CNTK (Microsoft Cognitive Toolkit), Lasagne, Big DL, etc.

The deep learning library management device 500 is based on the information transmitted from the deep learning library information providing device 300 and the information on the deep learning server, one or more deep learning libraries and one or more deep learning libraries among a plurality of deep learning libraries. You can automatically manage the corresponding configuration. The deep learning server may be a server in which the deep learning library is executed and the deep learning library management apparatus 500 is applied.

Accordingly, the deep learning library management apparatus 500 may solve the inconvenience of the existing deep learning library use method in which the user has to manually perform the deep learning library management. The deep learning library management apparatus 500 can automatically keep the user's development environment up to date without the need for the user to directly check the version for the deep learning library and install it. The deep learning library management device 500 can help the user to focus on research without managing the deep learning library by making the deep learning library convenient for the user as it solves the existing inconvenience. The deep learning library management device 500 allows the user to use the deep learning library easily and conveniently without consuming the work cost and time required for managing the deep learning library. Therefore, users can use it conveniently and usefully without the need to directly perform detailed tasks (hardware, software, program and library management) required for the deep learning service to be used.

The deep learning library management apparatus 500 may be implemented independently of the deep learning server, but is not limited thereto. For example, the deep learning library management apparatus 500 may be implemented by being included in a deep learning server or a deep learning server.

Hereinafter, for convenience of explanation, various information on each of the plurality of deep learning libraries is named as first information, and information on the deep learning server 500 is named as second information.

Figure 2 shows a schematic block diagram of the deep learning library management apparatus of Figure 1.

The deep learning library management apparatus 500 may include a memory 510 and a controller 530 .

The memory 510 may store instructions (or programs) executable by the controller 330 . For example, the instructions may include instructions for executing an operation of the controller 530 and/or an operation of each component of the controller 330 .

The controller 530 may process data stored in the memory 510 . The controller 530 may execute computer readable codes (eg, software) stored in the memory 510 and instructions induced by the controller 530 .

The controller 530 may be a hardware-implemented data processing device having a circuit having a physical structure for executing desired operations. For example, desired operations may include code or instructions included in a program.

For example, a data processing device implemented as hardware includes a microprocessor, a central processing unit, a processor core, a multi-core processor, and a multiprocessor. , an Application-Specific Integrated Circuit (ASIC), and a Field Programmable Gate Array (FPGA).

The controller 530 may control the overall operation of the deep learning library management apparatus 500 . For example, the controller 530 may control the operation of the memory 510 .

The controller 530 may update a version of one or more deep learning libraries among a plurality of deep learning libraries based on the first information.

For example, the controller 530 may search for one or more deep learning libraries corresponding to the user's input signal from among a plurality of deep learning libraries in response to the user's input signal. The user's input signal may include information about the container and deep learning library to be used by the user. For example, the input signal may include information about the deep learning library to be used by the user, the version of the deep learning library to be used by the user, the development language to be used for development, and the container to be used by the user. Information on the container to be used by the user may be ID and/or names. For example, the user's input signal may be "DLLM <container ID or Names> <lang=python> <tensorflow=2.1> <pytorch=1.4> <etc>". At this time, the development language may be python, the deep learning library may be tensorflow and pytorch, the version of tensorflow may be version 2.1, and the version of pytorch may be version 1.4.

The controller 530 may determine whether the version of the one or more deep learning libraries is the latest version by comparing the version of the one or more deep learning libraries with the latest version of the one or more deep learning libraries in the first information. The at least one deep learning library is a deep learning library corresponding to a user's input signal among a plurality of deep learning libraries, and may be a deep learning library to be used by the user. A configuration corresponding to one or more deep learning libraries is a configuration necessary for the one or more deep learning libraries to run, and may be a configuration compatible with one or more deep learning libraries. The first information may be a name of each deep learning library, a version of each deep learning library, and information about a configuration corresponding to the version. The information about the version of each deep learning library may be all version information for each deep learning library. All version information may be information about the latest version and the old version. The information on the configuration corresponding to the version of each deep learning library may be information about hardware, software, programs and libraries compatible with each of a plurality of versions of each deep learning library.

If the version of the one or more deep learning libraries is not the latest version, the controller 530 may update the version of the one or more deep learning libraries to the latest version of the one or more deep learning libraries in the first information.

The controller 530 configures a first configuration corresponding to the version of the one or more deep learning libraries before updating based on the first information and the second information, and a second configuration corresponding to the updated version (or the latest version) of the one or more deep learning libraries. can be changed to The second information may include information about hardware of a deep learning server on which an updated version of one or more deep learning libraries will be executed.

For example, the controller 530 may compare the first configuration with the first information and the second information to determine whether the first configuration is a configuration compatible with an updated version of one or more deep learning libraries. The first configuration may be hardware, software, programs and libraries compatible with the version of the one or more deep learning libraries prior to the update (or information about the hardware, software, programs and libraries compatible with the version of the one or more deep learning libraries prior to the update). have. The second component comprises hardware, software, programs and libraries compatible with an updated version (or newer version) of one or more deep learning libraries (or hardware, software, programs compatible with an updated version (or newer version) of one or more deep learning libraries). and information about the library).

If the first configuration is not compatible with the updated version of one or more deep learning libraries, the controller 530 may change the first configuration to the second configuration based on the determination result.

The controller 530 may distribute an updated version of one or more deep learning libraries and a configuration corresponding to the updated version to a container in which the updated one or more deep learning libraries will be executed.

The controller 530 updates the deep learning library to be used by the user, and changes the configuration corresponding to the updated deep learning library to the configuration corresponding to the updated version, but is not limited thereto. For example, the controller 530 periodically updates the versions of all deep learning libraries in containers built in the deep learning server, and corresponds to the updated version of each deep learning library with a configuration corresponding to each deep learning library to be updated. configuration can be changed.

Hereinafter, for convenience of explanation, it is assumed that the deep learning library management device 500 is a deep learning server, and it is assumed that version management and configuration changes are performed for all deep learning libraries in containers built on the deep learning server.

3 illustrates an example for explaining a version upgrade operation and a configuration change operation of the controller illustrated in FIG. 2 .

The operation of the controller 530 described in FIG. 2 may be performed by the modularized detailed configuration of the deep learning server, and the operation of the detailed configuration of the deep learning server described below may be performed by the controller 530 . For example, the deep learning server is a version check module (version check module; 531), container state check module (all container state check module; 533), hardware resource state check module (hardware resource state check module; 535), container manager It may be composed of a module (container manager module; 537) and a deployment module (539).

The deep learning server may automatically perform version upgrade (or version management) by checking the version of the deep learning library for each cycle according to a preset time through the version check module 531 and the container state check module 533 . Therefore, the user does not need to directly check the version for the deep learning library and install it.

For example, the deep learning server periodically receives and stores the first information from the deep learning library information providing apparatus 300 through a version check module (531) to receive the first information for each deep learning library. can be kept up to date. The deep learning server may check the deep learning library version used for each container through the container state check module 533 . If the container's deep learning library version is not the latest version, the deep learning server upgrades the container's deep learning library version to the latest version, and includes hardware, software, and Program and/or library setup and deployment can be performed automatically.

When the version of the deep learning library is upgraded or the hardware of the deep learning server 500 is changed, the deep learning server is configured through the version check module 531, the hardware resource status check module 535 and the container status check module 533. You can keep the development environment up to date by checking the hardware settings (or version) every cycle and checking the compatibility with the upgraded deep learning library. Accordingly, the user or the server administrator does not need to directly check the hardware setting and change the setting.

For example, the deep learning server may periodically maintain the latest information on compatible hardware for each deep learning library through the version check module 531 .

The deep learning server may periodically maintain the latest information on the hardware of the current deep learning server 500 through the hardware resource state check module 535 .

When the deep learning library is upgraded or the hardware of the deep learning server is changed through the container status check module 533, the deep learning server checks the compatibility between the upgraded deep learning library and the hardware of the deep learning server and changes the hardware settings accordingly. This can be done automatically.

When the version of the deep learning library is changed or the versions of software, programs and libraries compatible with the deep learning library are changed, the deep learning server performs the software, program by cycle through the version check module 531 and the container status check module 533. You can keep the development environment up to date by checking the version of and library settings. Accordingly, there is no need for the user to directly check the software, program and library settings to install and change.

For example, the deep learning server may periodically maintain the latest information on software, programs and libraries compatible with each deep learning library through the version check module 531 .

When the deep learning library is upgraded or the version of the software, program, and library that fits the deep learning library is changed, the deep learning server checks the compatibility between the deep learning library and the software, program and library through the container status check module 533. It can automatically install and change settings for the right software, programs and libraries.

The deep learning server can distribute information about the changed configuration to a compatible configuration to the updated deep learning library and the updated deep learning library through the deep learning library management and container health check module 533. .

For example, the deep learning server checks the version of each deep learning library in the deep learning library management, the versions of hardware, software, programs and libraries compatible with each deep learning library, and performs the deep learning from the deep learning library information providing device 300 . Files necessary for learning development can be distributed and stored in deep learning library management for management.

The deep learning server may perform version control automation, hardware configuration automation, software, program, and library configuration automation through the container status check module 533, and may distribute files necessary for each container. In this case, the container state check module 533 may distribute information necessary for hardware setting first, and then distribute information necessary for software, program, and library setting. Finally, the container state check module 533 may distribute information necessary for setting up the deep learning library.

4A shows an example for explaining the operation of the version check module, FIG. 4B shows an example for explaining Table 1 shown in FIG. 4A, and FIG. 4C shows an example for explaining Table 1_sub shown in FIG. 4A An example is shown.

4A to 4C, the version check module 531 periodically accesses the deep learning library information providing device 300 and includes information about the version, install url, and version-state for each deep learning library. It is possible to confirm and receive the first information.

The version check module 531 generates and updates (updates) Table 1 based on the information on each deep learning library in the first information transmitted from the deep learning library information providing device 300, so that for each deep learning library Information can be kept up to date. Table 1 may include information listed by matching information about each deep learning library, each deep learning library, and a plurality of versions corresponding to each deep learning library, as shown in FIG. 4B .

The version check module 531 is based on the information on the configuration corresponding to each deep learning library in the first information, the version, install url for hardware, software, program and library suitable for each of a plurality of versions of each deep learning library can be checked to create and update Table 1_Sub. For example, Table 1_sub may include information listed by matching information about a plurality of versions of each deep learning library and hardware, software, programs, and libraries suitable for each version, as shown in FIG. 4C .

The deep learning library management of the deep learning server may manage the deep learning library information providing device 300 , the first information, Table 1 and Table 1_sub.

5A shows an example for explaining the operation of the hardware resource status check module, and FIG. 5B shows an example for explaining Table 2 shown in FIG. 5A.

5A and 5B, the hardware resource status check module 535 periodically checks and receives the hardware information of the deep learning server transmitted from the deep learning server hardware resource of the deep learning server. can Although the hardware resource status check module 535 checks and receives various hardware information such as GPU, CPU, Storage, and RAM of the deep learning server, the hardware that has the greatest influence on the deep learning library may be the GPU of the deep learning server.

The hardware resource status check module 535 may generate and update Table 2 based on the received hardware information of the deep learning server to keep the information on the hardware of the deep learning server up to date. Table 2 may include information listed by matching the name and type of each hardware of the deep learning server as shown in FIG. 5B.

6A shows an example for explaining the operation of the container manager module, FIG. 6B shows an example for explaining Table 3 shown in FIG. 6A, and FIG. 6C shows a table showing detailed information about containers in Table 3 An example for describing 3_sub is shown.

6A to 6B , the container manager module 537 may generate and update Table 3 based on environment information on containers built in the deep learning server. The container manager module 537 periodically checks whether the container space included in Table 3 exists in the deep learning server, and when the container is not checked in the deep learning server, the unverified container may be deleted from Table 3. Accordingly, the container manager module 537 may keep the container list of Table 3 up to date. The environment information on the container is the ID and name of the container, and Table 3 may include information listed by matching the container ID, name, and index number as shown in FIG. 6B .

6C to 6D , the container manager module 537 operates in the background inside each container and periodically checks the deep learning library, software, program and library information inside each container to generate and update Table 3_sub. can In this case, the container manager module 537 may be a container state manager module. Table 3_sub may include information listed by matching the name, version, and type of each deep learning library as shown in FIG. 6D . Accordingly, if the user arbitrarily or accidentally upgrades or changes the version of the deep learning library, software, program and library inside the container, the container manager module 537 provides information on the deep learning library, software, program and library inside each container. can manage

If the deep learning server and the stable version of the deep learning library do not match, the container manager module 537 may redistribute various information about the deep learning library to be executed based on the deep learning library.

7 shows an example for explaining the operation of the container status check module.

The container status check module 533 uses Table 3_sub for each container in Table 3 (or built in a deep learning server) to periodically determine whether information on containers has changed, and to perform distribution accordingly. can In this case, the container state check module 533 may determine whether information is changed based on Table 1, Table 1_sub, and Table 2.

For example, the container state check module 533 may compare the updated Table 1 with version information of all deep learning libraries in each container to determine whether the deep learning library has been changed to the latest stable version. When the version of the deep learning library is changed, the container state check module 533 may perform distribution of the deep learning library for version upgrade based on Table 1.

The container state check module 533 may compare the updated Table 2 with hardware information of the deep learning server in each container to determine whether the hardware of the deep learning server has been changed. When the hardware of the deep learning server is changed, the container state check module 533 may perform hardware configuration distribution for hardware compatibility check and configuration change based on Table 2 .

The container state check module 533 may compare the updated Table 1_sub with the configuration corresponding to all deep learning libraries in each container to determine whether the configuration corresponding to the deep learning library has been changed. When the configuration corresponding to the deep learning library is changed, the container state check module 533 may perform configuration configuration distribution for configuration compatibility check and configuration change based on Table 1_sub.

8 shows an example for explaining the operation of the distribution module.

The distribution module 539 may perform a distribution operation for installing and running a deep learning library suitable for the deep learning server due to a hardware change and/or a deep learning library upgrade of the deep learning server.

First, the distribution module 539 may perform a hardware distribution operation corresponding to (or matching) a deep learning library suitable for a deep learning server. For example, the distribution module 539 may check the hardware information of the current deep learning server through Table 2 to identify a deep learning library suitable for the hardware of the deep learning server. The distribution module 539 may check the hardware distribution file (Type==HW) corresponding to the deep learning library suitable for the deep learning server through the LOC of Table 1_sub of the deep learning library suitable for the deep learning server. The distribution module 539 may distribute (or install) the hardware distribution file by performing installation and setting value setting for the identified hardware distribution file. When the hardware distribution file of the LOC cannot be used, the distribution module 539 may directly distribute the hardware distribution file using the url of Table 1_sub.

Thereafter, the distribution module 539 may perform a software, program and library distribution operation corresponding to the deep learning library suitable for the deep learning server. For example, the distribution module 539 provides software, programs and library distribution files corresponding to the deep learning library suitable for the deep learning server through the LOC of Table 1_sub of the deep learning library suitable for the current deep learning server after the hardware configuration distribution is completed. (Type==SW) can be checked. The distribution module 539 may distribute (or install) the software, program, and library distribution files by performing installation and setting value setting for the identified software, program, and library distribution files. When the software, program, and library distribution file of the LOC cannot be used, the distribution module 539 may directly distribute the software, program, and library distribution file using the url of Table 1_sub.

Finally, the distribution module 539 may perform a deep learning library distribution operation suitable for the deep learning server. For example, the distribution module 539 may check a deep learning library suitable for the deep learning library through Table 2 after distribution of hardware, software, program, and library settings is completed. The distribution module 539 may check the deep learning library distribution file suitable for the deep learning server through the LOC of Table 1_sub of the deep learning library suitable for the deep learning server. The distribution module 539 may distribute (or install) the identified deep learning library distribution file. When the deep learning library distribution file of the LOC cannot be used, the distribution module 539 may directly distribute the deep learning library distribution file using the url of Table 1_sub.

As described above, although the embodiments have been described with reference to the limited drawings, those skilled in the art may apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order than the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (20)

In a method for automatically managing a deep learning library operating for each virtual operating environment,
updating a version of one or more deep learning libraries among the plurality of deep learning libraries based on first information about the plurality of deep learning libraries; and
Change the first configuration corresponding to the version to a second configuration corresponding to the updated version of the one or more deep learning libraries based on the first information and second information about the deep learning server on which the one or more deep learning libraries are to be executed step to do
A method comprising
According to claim 1,
The first information is
A method, including the name of each deep learning library, the latest version of each deep learning library, and information about a configuration corresponding to the latest version.
3. The method of claim 2,
The second information includes information about the hardware of the deep learning server, the method.
4. The method of claim 3,
The updating step is
comparing the version of the one or more deep learning libraries with the latest version of the one or more deep learning libraries in the first information to determine whether the version of the one or more deep learning libraries is the latest version; and
If the version of the one or more deep learning libraries is not the latest version, updating the version of the one or more deep learning libraries to the latest version
A method comprising
5. The method of claim 4,
Searching for the one or more deep learning libraries corresponding to the input signal from among the plurality of deep learning libraries in response to an input signal of a user who will use the one or more deep learning libraries
further comprising,
The one or more deep learning libraries,
A method, which is a deep learning library to be used by the user from among the plurality of deep learning libraries.
6. The method of claim 5,
The user's input signal is
A method comprising information about a virtual operating environment to be used by the user, a deep learning library to be used by the user, and a version of the deep learning library to be used by the user.
According to claim 1,
wherein the first configuration is hardware, software, programs and libraries compatible with the version.
In claim 7,
wherein the second component is hardware, software, program and library compatible with the updated version.
9. The method of claim 8,
The changing step is
comparing the first configuration with the first information and the second information to determine whether the first configuration is a configuration compatible with the updated version; and
if the first configuration is not compatible with the updated version, changing the first configuration to the second configuration;
A method comprising
According to claim 1,
Distributing the updated version and a configuration corresponding to the updated version to a virtual operating environment in which the one or more deep learning libraries will be executed
A method further comprising:
In a device for automatically managing a deep learning library operating for each virtual operating environment,
a memory containing instructions; and
a controller for executing the instructions
including,
The controller is
Update the version of one or more deep learning libraries among the plurality of deep learning libraries based on the first information about the plurality of deep learning libraries, and the first information and the one or more deep learning libraries for a deep learning server to be executed changing the first configuration corresponding to the version to a second configuration corresponding to the updated version of the one or more deep learning libraries based on the second information.
12. The method of claim 11,
The first information is
The device, including the name of each deep learning library, the latest version of each deep learning library, and information about the configuration corresponding to the latest version.
13. The method of claim 12,
The second information includes information about the hardware of the deep learning server, the device.
14. The method of claim 13,
The controller is
Compare the version of the one or more deep learning libraries with the latest version of the one or more deep learning libraries in the first information to determine whether the version of the one or more deep learning libraries is the latest version, and of the one or more deep learning libraries If the version is not the latest version, updating the version of the one or more deep learning libraries to the latest version.
15. The method of claim 14,
The controller is
Searching for the one or more deep learning libraries corresponding to the input signal from among the plurality of deep learning libraries in response to an input signal of a user who will use the one or more deep learning libraries,
The one or more deep learning libraries,
The device, which is a deep learning library to be used by the user among the plurality of deep learning libraries.
16. The method of claim 15,
The user's input signal is
The device comprising information about the virtual operating environment to be used by the user, the deep learning library to be used by the user, and the version of the deep learning library to be used by the user.
12. The method of claim 11,
wherein the first configuration is hardware, software, programs and libraries compatible with the version.
In claim 17,
and the second configuration is hardware, software, programs and libraries compatible with the updated version.
19. The method of claim 18,
The controller is
comparing the first configuration with the first information and the second information to determine whether the first configuration is a configuration compatible with the updated version, and if the first configuration is not compatible with the updated version, changing the first configuration to the second configuration.
12. The method of claim 11,
The controller is
Distributing the updated version and a configuration corresponding to the updated version to a virtual operating environment in which the one or more deep learning libraries will be executed.

Images