What is NVIDIA AI Workbench?
NVIDIA AI Workbench is a developer toolkit for data science, machine learning, and AI project development. AI Workbench is free, you can install it in minutes on your local or remote computers, and it provides both a desktop application and a command-line interface (CLI).
AI Workbench enables the following:
Develop on your laptop and then move workloads to scalable GPU resources in a data center or the cloud.
Collaborate and distribute your work through GitHub and GitLab.
Write and run code in JupyterLab, VS Code, or a text editor.
Clone fully-functioning NVIDIA-provided examples from GitHub for tasks such as Retrieval-Augmented Generation (RAG), model fine-tuning, and RAPIDS data science workflows.
Develop in a containerized environment using NVIDIA’s pre-built containers, or your own fully-custom containers.
AI Workbench is supported on the following operating systems:
Windows 11 (build 22000 or later), Windows 10 (build 19044 or later)
Ubuntu 24.04, Ubuntu 22.04
macOS Monterrey (12) or later
AI Workbench is composed of the following:
Desktop app — The AI Workbench desktop application. A cross-platform application that provides the graphical user interface (GUI) for AI Workbench.
Command-Line Interface (CLI) — The AI Workbench CLI. The command line interface for users who prefer working in a terminal, or who want to script AI Workbench processes.
Service/Server — The AI Workbench service. The core component that provides the API to manage and interact with projects, and that handles backend operations and management tasks.
Credential Manager — The AI Workbench credential manager. A small application that integrates with your host’s keychain or secret store.
For more information, see AI Workbench System.
A computer that has AI Workbench installed is called a location. The first place you install AI Workbench is on your local computer, which can be a CPU-only computer. Later you can install AI Workbench on one or more remote computers, such as workstations, servers, cloud instances, and virtual machines. For more information, see AI Workbench Locations.
An AI Workbench project contains everything you need to do your work, such as code, data, models, environment configuration, metadata, and work history. AI Workbench projects are specially-formatted Git repositories, and AI Workbench uses the format and metadata to provide automation for container and version management. For more information, see AI Workbench Projects.
The environment for a project is a container that AI Workbench builds and runs. NVIDIA provides default containers that you can select from as the starting point for each new project. For more information, see Base Container Environments for AI Workbench Projects.
The NVIDIA-provided default containers include the following:
PyTorch for AI Workbench
Python Basic for AI Workbench
Python with CUDA 11.7, 12.0, and 12.2
RAPIDS with CUDA 12.0.
AI Workbench lets you connect to external systems, such as container registries and Git servers, through authentication methods like personal-access-tokens (PAT) and Oauth integrations. AI Workbench stores your credentials securely so you don’t have to log in every time. After you authenticate to an external system, you can access it from any project running in any of your locations, including remote locations. For more information, see AI Workbench Integrations.
Applications are software or user interfaces available to use in, or with, an AI Workbench project. For example:
JupyterLab — Each of the NVIDIA-provided base container options has JupyterLab installed by default.
Visual Studio Code — If VS Code is installed on your local computer, you can connect it to your Workbench project.
Gradio Apps — Some projects have simple web apps like chat apps or RAG apps installed.
For more information, see AI Workbench Applications.