Preview
Yandex DataSphere

A service for ML development that combines the most popular tools and scalable resources for the full cycle of machine learning: from an experiment to the launch of a finished model.

The service is at the Preview stage and is not charged.

Familiar development environment
Use the Jupyter® Notebook user-friendly interface as an environment for a variety of tasks: data analysis, model development, running complex computing, and much more.
Serverless computing
Select the appropriate computing resource configurations for specific cells and run code without having to configure servers and VMs.
Flexible choice of resources
In a running project, select the configuration and the necessary resources for specific parts of the code. Changes occur in seconds, within the same training scenario, while maintaining the result.
Popular libraries
DataSphere already has TensorFlow, PyTorch*, and other major libraries for data analysis and machine learning. Moreover, you can always install any additional packages you need.
Distributed computing
The environment enables you perform distributed computing on an Apache Spark cluster running in the Yandex Data Proc service.

Getting started

Create your project in the management console and work in the familiar Jupyter® Notebook interface using Yandex.Cloud computing resources.

Create project

Questions and answers

What is Yandex DataSphere?

DataSphere is a cloud environment that combines the most popular tools and resources for ML development. It helps you complete the full cycle of creating a model: from an experiment and development to the launch of a finished version on Yandex.Cloud computing resources.

DataSphere is a cloud environment that combines the most popular tools and resources for ML development. It helps you complete the full cycle of creating a model: from an experiment and development to the launch of a finished version on Yandex.Cloud computing resources.

What are the advantages of the Yandex DataSphere serverless environment?

  • You don’t need to spend time creating and maintaining VMs: when you create a new project or run calculations, computing resources are allocated automatically.
  • You can scale the resources you need for working with data and training your models without interrupting computing processes.
  • The service supports various computer configurations, including VMs with Nvidia V100 GPU and distributed computing on a SPARK cluster.
  • As a development environment, the service uses the popular JupyterLab interface.
  • You can start using basic packages for data analysis and machine learning immediately, including TensorFlow, Keras, NumPy, PyTorch, CatBoost, LightGBM, and more.

All this helps significantly reduce the cost of ML as compared to computing on your own hardware or other cloud platforms, as well as reduce the time for development and speed up the transfer of models from the experimental stage to commissioning.

  • You don’t need to spend time creating and maintaining VMs: when you create a new project or run calculations, computing resources are allocated automatically.
  • You can scale the resources you need for working with data and training your models without interrupting computing processes.
  • The service supports various computer configurations, including VMs with Nvidia V100 GPU and distributed computing on a SPARK cluster.
  • As a development environment, the service uses the popular JupyterLab interface.
  • You can start using basic packages for data analysis and machine learning immediately, including TensorFlow, Keras, NumPy, PyTorch, CatBoost, LightGBM, and more.

All this helps significantly reduce the cost of ML as compared to computing on your own hardware or other cloud platforms, as well as reduce the time for development and speed up the transfer of models from the experimental stage to commissioning.

Get started with DataSphere

  1. TensorFlow is a registered trademark of Google Inc.
    PyTorch is a registered trademark of Facebook, Inc.