TensorFlow Server Update
대표적인 딥 러닝 프레임워크인 TensorFlow와 머신러닝 패키지들이 설치된 서버(GPU 선택 가능)를 제공합니다.
The Server Carefully Selected for Deep and Machine Learning
A data analysis development environment server service that can be created quickly and used in the cloud environment without the need to install various libraries required for deep learning research and data analysis You can use various machine learning frameworks with the web analysis tool Jupyter Notebook.
- Creating a Convenient Analysis Environment
- You can easily create deep learning and machine learning environments within several minutes through a web-based console. We have carefully selected and are providing other types of analysis packages including PyTorch, Keras, Pandas, Numpy, and Scikit-Learn. Through this, data analysts can solely focus on analysis.
- Support for Intel MKL base and AVX512
- Tensorflow with Intel's MKL (Math Kernel Library) is applied to enhance the deep running and machine learning performance on Intel Xeon processor, and the built-in Tensorflow (Ubuntu 16.04 Only) that is optimized for AVX512 Vector operation when CPU intensive is selected . Offering over 200% better performance in multiple deep running, machine running workloads.
- Various Templates
- We provide Jupyter Notebook sample codes, such as basic examples of TensorFlow, regression analysis, and visualization. Even junior data analysts can experience easy code analysis through this variety of templates.
- Provides high performance and scalable GPU and CPU Intensive server
- With the optimized environment configuration provided by the Naver Cloud platform, you can choose the optimal GPU (Tesla P40) or the Intel Xeon CPU that supports single core turbo up to 3.7 GHz without complex setups.
Detailed Features
Provided Services
In addition to TensorFlow, we provide about 200 types of analysis packages, such as PyTorch, Keras, Pandas, Numpy, Scikit-Learn, and Gensim on a single virtual server. Analysts can analyze and enter visualization codes from the Jupyter Notebook. Even more convenient data analysis is possible as the Conda virtual environment management tool is enabled as default.
| Category | Content |
|---|---|
| Main Analysis Package | TensorFlow, Keras, Gensim, PyTorch, Theano, and much more (The package version distributed may change depending on the version-up speed of TensorFlow.) |
| Additional Convenience | Conda virtual environment management tool, Jupyter Notebook management script |
| OS | Ubuntu 16.04 Server (64bit) CentOS 7.3 Server (64bit) |
Management Scripts
We provide scripts for Jupyter Notebook server start, stop, restart, and account password reset. (Refer to the user manual) Through this, a basic management is possible without changing the setup files.
Reference
"Tensorflow Server" uses Tensorflow, an open source machine learning software library developed by Google Brain.
Related Services
Pricing Information
TensorFlow Server is an open source, machine learning VM (Virtual Machine) installation-type service, so no additional fee is charged other than the usage fee for the server and public IP address.
| OS | Version | Usage Fee (Month) |
|---|---|---|
| Ubuntu 16.04/CentOS 7.3 GPU Support | Tensorflow 1.13.x (Stable Latest) (The package version distributed may change depending on the version up speed of TensorFlow.) | Fees for using Server and public IP are charged Recommended Specification: For 1 GPU, vCpu 4, 24G memory, 50G HDD, 1,200,000 KRW per month + public IP usage fee of 4,032 KRW ※ See the usage fee of the GPU Server for details. |
| Ubuntu 16.04/CentOS 7.3 CPU | Tensorflow 1.13.x (Stable Latest) (The package version distributed may change depending on the version up speed of TensorFlow.) | Fees for using Server and public IP are charged Recommended Specification: For 4 vCPU and 16G memory, 205,000 KRW per month + public IP usage fee of 4,032 KRW ※ CPU Intensive type server refer to Server fee guide |
Was this page helpful?