HOME > Solution > Solutions >

NVIDIA Artificial Intelligence Lab Solution

TIME:2020-04-10 14:55       

      NVIDIA artificial intelligence laboratory solution is an integrated artificial intelligence laboratory solution based on the data center virtualization infrastructure that NVIDIA has launched for the Chinese education market, plus the NVIDIA Deep Learning Academy training system. The solution uses a user-centric computing model and provides GPU resources, framework resources, and deep learning environments to users on demand. To ensure users' exclusive enjoyment of GPU resources, to ensure users have sufficient freedom and a high user experience, and to meet the needs of centralized management and data security, flexible and rapid deployment of artificial intelligence experiment environment.

      Data center virtualization infrastructure platform is a mature solution, mainly including server virtualization, storage virtualization, network virtualization and management platform. Coupled with NVIDIA's GPU virtualization, users can easily, quickly, and one-stop build a campus GPU resource management platform, provide GPU services internally, greatly simplify GPU resource management, improve data security, and improve machine hardware utilization Rate and reduce IT total cost of ownership (TCO).

      Using a hybrid cloud computing model (collective server virtualization, GPU virtualization, PC virtualization, and network disk), it not only supports the flexible allocation of the server's back-end computing resource pool, but also supports the distributed computing of the client's front-end computing resources, Grid6. 0 brings a new environment implementation for GPU virtualization, supports the construction and calculation of deep learning frameworks under GPU virtual environment, and is an important technical platform for the implementation of artificial intelligence laboratories in educational institutions.


1.1 System architecture


1.1.1 System architecture


1.1.2 vGPU architecture


1.2 Product description


1.2.1 NVIDIA GPU
       Through the server virtualization platform, the GPU computing resources are cut and allocated according to demand. Relying on the DOCK technology, various deep learning practice environments are quickly deployed. Users can access cloud-based computers, notebooks, PCs, and smart terminals through browsers. Cloud artificial intelligence laboratory portal website, to carry out the interactive experience of AI experiments, model practice and deduction products.


1.2.2 Deep learning framework


1.2.3 Nvidia Deep Learning Academy
      The NVIDIA artificial intelligence laboratory solution is built on the courses of the NVIDIA Deep Learning Academy. At the same time, the content and elements in line with Chinese education are added to the original international curriculum to help teachers and students better teach, experience and learn the knowledge of artificial intelligence.

      The NVIDIA Deep Learning Academy (DLI) provides hands-on training for AI and accelerated computing, aimed at solving real-world problems. DLI has tailored courses for developers, data scientists, and researchers. You can learn in three ways:


1.2.3.1 Online themed courses


      Each subject course includes eight hours of lessons to learn how to implement and deploy end-to-end projects. You can start online learning anytime, anywhere with a fully-accelerated GPU-accelerated workstation in the cloud.


1.2.3.2 Online elective courses


      Each elective course includes two hours of class time to explore how to apply specific technologies or develop technologies. You can start online learning anytime, anywhere with a fully-accelerated GPU-accelerated workstation in the cloud.


1.2.3.3 Instructor-led training courses
      Face-to-face training classes include eight-hour themed courses to learn how to implement and deploy end-to-end projects. It is taught by DLI certified instructors and uses cloud-configured GPUs to accelerate the practice of workstations. You can apply for training in your own organization, university campus and large-scale related conferences.


1.2.3.4 Certification


      Participants can obtain certifications to prove their abilities in related fields and provide support for their professional development. You can obtain certification in some online courses and instructor-led training courses.

 

1.2.4 Nvidia Artificial Intelligence Laboratory Platform
1.2.4.1 Interface




1.2.4.2 Function




1.2.4.3 How to use


1.3 Course content


1.3.1 Course list



1.3.2 Artificial intelligence annotation technology



1.3.3 Analysis of artificial intelligence applications



1.3.4 Python visualization and its application


                                                                                            


1.3.5 Interactive Python Programming



                                                                                            


1.3.6 Experimental effect



2. Feature highlights


2.1 Convenient management


2.1.1 Management platform
      The management platform of the artificial intelligence laboratory system adopts the B / S architecture, a simple full graphical management interface, users can access the platform through a browser on any networked terminal, and the server, user, desktop, scene, terminal can be realized on the platform , Storage and other comprehensive centralized management.

      The browser-based web console is the only management portal for the entire self-service platform. Administrators can sit on the desk and maintain the GPU resources, cloud terminals and infrastructure of the entire school through the browser, whether it is on / off / restart or backup / Restore, or reset / upgrade, do not enter the laboratory or go to the site, it is easily completed, and the work efficiency is improved by more than ten times.


2.1.2 Centralized software and hardware upgrade


      The artificial intelligence laboratory system decouples the computing performance requirements of the hardware from the local terminal. All GPU computing resources are provided by the cloud server. The local terminal can be an ordinary computer, an all-in-one computer, or a thin client.

      In addition to managing cloud resources, server virtualization software can also provide cloud restoration functions to ensure that the experimental environment is automatically restored to the preset state after the course. At the same time, based on the template and link cloning technology, the server side centrally deploys desktops. As long as a single template is configured, modified, installed and upgraded software / virtual machine hardware, hundreds of desktops can be synchronized and updated, and the administrator can complete all in a few minutes Operation, no need to go to the user site to operate one by one.

      PC virtualization includes computer management functions, and is responsible for the management of virtual system templates, computer terminal information management, access control, and peripheral device management, and unified push and update of virtual operating systems and applications through the management platform.


2.1.3 Rapid deployment of experimental courses

      Using the management platform, it is easy to make targeted NVIDIA DOCKER software environment templates. The templates can be used to create, update, and distribute experimental environments in batches. Hundreds of terminals can instantly use the corresponding user operating environment to achieve rapid deployment. Terminal application software installation and configuration changes can be centralized on the server side through the management console, which greatly simplifies the configuration and deployment of the user's operating environment.


2.2 Effective control


2.2.1 Teaching management

      For the management authority, the education desktop provides different management ends for administrators and class teachers. Administrators log in to the background software management platform to complete the management and system maintenance of all users. The teaching management terminal is a management software for class teachers. It can choose a variety of independent teaching environments, and can switch the teaching environment with one key, and switch on and off the class and class. It also allows students to choose freely without logging in to the management background. operating.


2.2.2 Effectively control the use of USB interface
      Based on the USB interface redirection and penetration technology, it effectively supports various USB devices in teaching experiments, not limited to USB storage, printers, scanners, smart cards, U-keys, touch screens, etc.

      At the same time, for various viruses transmitted by USB storage and data security and confidentiality specifications, the USB interface can be restricted for each user terminal. Even if a virus-infected mobile storage device is plugged in, it cannot be accessed and accessed on the desktop. The security of the user's desktop and data is guaranteed.


2.2.3 Complete monitoring function
      Provide performance monitoring function, real-time data statistics on indicators such as GPU, CPU, network, disk, and resource usage of each process in the server and terminal desktop, and can reflect the current resource bottleneck of physical machines and virtual machines, and managers can also monitor See the operation of the client access to the desktop, and take screenshots or videos when necessary.


2.3 Teacher Office


2.3.1 Rapid delivery of personalized DOCK environment
      The terminal desktop architecture is managed uniformly in the background, quickly delivering various teaching application desktops without the need to install them one by one on site; the system selection is more "flexible" Teachers can flexibly choose a variety of NVIDIA DOCKER environments according to personal usage habits.


2.3.2 Mobile teaching office mode
      Mobile office includes school office and home office; mobile teaching includes office preparation and classroom teaching. The desktop used by teachers is to build a variety of teaching software and office software into the cloud desktop and package it into various standard desktops. Configure and distribute as needed. If there is a network, you can use your own personal desktop. However, logging in to the virtual desktop is highly dependent on bandwidth. Now the network is developing at a high speed, and the home bandwidth can guarantee 20 megabytes, which can fully support the remote connection of home office. At the same time, teachers use other mobile devices to work, but also to ensure the requirements of broadband, teachers can remotely access their desktops anywhere in the school.

      After the teacher prepares the lesson in the office, after preparing the PPT, he only needs to come to the teacher, use the cloud terminal, log in to his desktop, and then complete the teaching directly. At the same time, the personal user password can be modified during login, ensuring the security of the personal desktop. When teachers return home, they can also work from home, they can log in to their desktops remotely, and use various software to access the resources in the school using the client. Create a brand-new teaching and office model that allows learning on demand, fragmented learning, and mobile office at any time.

      Teachers no longer need to carry laptops or USB drives back and forth between offices, homes, and classrooms. It is easy and convenient, and uses the same personal desktop.


3. Project value


Deep learning calculation model: Traveling-> INFERENCE



3.1 Artificial intelligence application experience
      In the future, with the in-depth development of technologies such as face recognition, speech recognition, semantic understanding, and deep learning, robots will have human-like thinking and interaction capabilities, empowering all aspects of economic production and social life. If students can deeply experience the magic brought by science and technology before learning related to artificial intelligence, it will be of great help to the cultivation of interest and later learning.

      AI language teachers can teach you how to write Tibetan poems, guess numbers with AI math teachers, AI English teachers learn English words and pronunciation, and can also quickly graffiti with AI art teachers, ask AI teachers many interesting questions face-to-face, and then AI sports teacher to play a game of Go ...

      In the future, with the application of smart correction, smart sparring, voice assessment, smart library, consultation management and other applications, AI will meet the personalized needs of students, solve the problem of relatively limited teacher resources, and at the same time transfer educational resources to more Where necessary, alleviate the imbalance of educational resources. With the help of AI empowerment and interoperability in the world, human beings will learn and develop at a new level of commonality.


3.2 Diversified AI Learning Environment-Training (AI Train)
      Based on NVDOCK technology, it can package a variety of deep learning environments, package various frameworks and drivers, and realize a variety of deep learning scenarios and personalized teaching experimental classrooms in a classroom. Administrators only need to provide different environments to students. For students, not only teaching-related software is available on the desktop, which improves desktop response speed, but also standardizes student attention, improves teaching quality, and has different framework learning opportunities. Experience various deep learning methods, which is very suitable for college AI laboratory.


3.3 Diversified AI Learning Environment-Inference
      The deep learning laboratory can be used as an artificial intelligence science laboratory to plan and design, and can use the existing Inference model to experience the classroom, such as machine vision, speech recognition and other artificial intelligence application scene experience. You can experience AI courses through models such as "machine games" and "visual recognition". Very suitable for the general education laboratory.


3.4 Training of AI professional teacher resources
      Through NVIDIA's Deep Learning Academy, AI professional teachers can be trained for schools. Teachers can superimpose artificial intelligence concepts and basic knowledge, deep learning techniques and algorithms, etc., to construct the school's own teaching and experimental content. The school has established a systematic AI teaching course which can also be realized.


3.5 Cloud teaching environment can be realized
      Due to the adoption of virtual desktop technology, common teaching software and lesson preparation resources can be packaged into a virtual desktop and delivered to students remotely through the network, which not only solves the problem of students installing and maintaining professional software, but also increases software and resources. Utilization rate, apportion purchase cost, also facilitate teachers and students to prepare lessons and complete teaching experiments at home.


The experiment document and the experiment environment are on the same page, creating an experiment experience of learning and practicing


Teachers can independently produce experimental documents and deploy experimental environments to create courses that are close to the needs of the school


3.6 Reuse of computing resources
       Due to the characteristics of Grid 6.0, GPU computing resources can be reused, and virtual cutting can be performed to achieve deep learning experimental classrooms and student teaching. At the same time, it can be used for large-scale computing tasks in idle time, such as integrating GPUs for model calculations at night.

       Due to the virtual desktop technology, the platform can provide heavy graphics and image application environments, such as 3Dmax, CATIA software teaching and experimental cloud environment. It can also provide more cloud application resource services for teachers and students of the school, which greatly saves the investment cost for the school.

SALES:(+86) 135 5249 6467    yj@jpgd.cn

ADD:4th Street, Shangdi, Haidian District, Beijing, China

Inquiry Us Now !

Copyright (c) 2020 jpgpu.com All Rights Reserved