It’s a structured Git repository that defines a containerized dev environment in a transparent way that’s easy for AI Workbench to run. The let you version your development environment along with your code.
The Basics
Projects are just Git repositories with some configuration files that follow a simple specification.
If the proper configuration files are added, any Git repository can become a Workbench Project.
AI Workbench reads the configuration files and uses the information to build and run a containerized development environment.
Project configuration files also provide information on the applications installed in the environment, and how to manage them.
The combination of the Project structure and AI Workbench automation provides streamlining, reproducibility and portability for GPU enabled development environments.
Types of Configurations
Project Configuration files: These keep track of everything required to reproduce a development environment.
Environments: Projects track the base image used to build the environment and packages that are installed on top of it.
Run time configuration: Projects track environment variables and parameters like the number of GPUs to be mounted in the container.
Application configuration: Projects capture commands and parameters for running applications inside the container.
AI Workbench follows a set of conventions based on the configuration information in a Project. This drives automation for portability and reproducibility, as well as application management.
Streamlining and Automation Driven by Simple Configuration
AI Workbench management of the Project is driven by the configuration files and user actions in the UI.
AI Workbench reads, alters and versions the configuration files based on user actions. Users can also edit them directly.
AI Workbench versions the configuration files via Git to keep configuration changes in sync with code and data changes.
Learn more about Workbench Projects
Check out some example Projects on GitHub
Install AI Workbench