Where can I download NVIDIA AI Workbench?

Currently you must be in the Early Access program and you download it from the page on DevZone. Visit here to sign up!

Is there a pricing schedule? Is this SaaS?

AI Workbench is and will remain free. It is a client application that you install on different systems. It isn’t SaaS.

What sort of GPUs are supported?

AI Workbench is independent of the GPUs on the system. The dependency is between the CUDA version in the Project container and the driver on the system.

How does it work if WSL2 isn’t already installed?

In this case, AI Workbench will enable and install it for you. This requires a reboot.

Do I install the Desktop App on a remote?

No. The Desktop App is only installed locally. To install on a remote you must use the zip file for Ubuntu 22.04.

What are the requirements for Ubuntu?

AI Workbench only supports Ubuntu 22.04 right now. The Desktop App is an AppImage file, so you should install fuse and libfuse2 before installing AI Workbench locally.

Can you only work in JupyterLab or VS Code?

JupyterLab and VS Code are the primary IDEs we are targeting at the moment. VS Code will be fully supported at GA in March of 2024.

Can you use things like torchrun?

Yes, you can use tools like torchrun from within a Workbench Project as part of the Python environment.

Can you use Jupyterlab dashboards in a Project?

Yes. That is something you could configure by listing the dependencies in requirements.txt and configuring in the postBuild.sh script.

How does GPU reservation work on a single system?

Windows does not yet support GPU reservation, but Ubuntu 22.04 does. You can configure the desired # of GPUs for a Project. If the GPUs aren’t available when you start the Project container, you will get a message to change the configuration.

Can you upload files to the remote with the Desktop App?

Yes. The files will be uploaded to whatever system the application window is attached to.

Can I install AI Workbench on a Windows remote desktop?

This isn’t currently supported by the application directly, but if you can enable WSL2 there it should be possible.

Can I use a different container from the two default bases?

Yes. Currently, there is a Bring Your Own Container (BYOC) approach (see the docs). We will have better support for this by GA in March 2024.

Can I use a different Git server like self-hosted GitLab?

Self-hosted GitLab will be supported this Beta release in January. Please see the integrations page of the docs for more details.

Can I use a RAG in a Workbench Project?

Yes, please see this example project on GitHub to get started.

How can I shutdown the AI Workbench application locally?

You can close out the application from the UI. AI Workbench will automatically shutdown locally.

Previous Deep Dive: Develop a Customized Project
Next Troubleshooting
© Copyright 2023-2024, NVIDIA. Last updated on Jan 21, 2024.