How to Work with AI Workbench Projects#
Overview#
- Projects are Git repositories with extra configuration files used by AI Workbench.
Projects primarily hold your code and environment configuration.
Data and models can also be versioned in the repository, but more often than not these are downloaded and uses on a machine by machine basis.
- Projects use containers for development environments.
Each project has an editable and versioned containerized development environment. You customize it with package lists, scripts, and runtime configurations.
- Projects are portable and reproducible.
AI Workbench handles the Git and container commands to build, run, and version your project. Any Git repository can become an AI Workbench project by adding the correct metadata files.
- You can version projects within the Desktop App and with Git natively.
The Desktop App has “easy buttons” for simple but end to end versioning workflows. For anything more complex, use Git in the terminal or a Git client.
Key Concepts#
- Project Container
A single container that serves as your development environment.
- Project Specification
The .spec.yaml file that defines your project metadata and configuration. A versioned file that AI Workbench reads and writes to while handling the environment configuration.
- Base Environments
Pre-configured container images from NVIDIA or custom containers.
- Mounts
Persistent storage that survives container restarts.
- Multi-Container Environments
Compose support for complex applications.
- Git-LFS
Git large file support that can be used to version data files in the project. Can be configured in the spec.yaml file.
What’s Covered?#
Topic Link |
Description |
|---|---|
Shows you how to create a new project in AI Workbench. |
|
Shows you how to understand the project specification in AI Workbench. |