AI Workbench Locations#
Overview of Locations#
Work locally or remotely in the same UI/UX#
A location is a system with NVIDIA AI Workbench installed.
local: Your laptop with the Desktop App installed, often called local Workbench.
remote: Remote system with the Workbench CLI installed, often called remote Workbench.
Working in a location is simple#
Open the AI Workbench Desktop App and select the location you want to work in.
When the location window opens, select a project to work on, or create/clone a new project.
Start working!
Note
The systems don’t need to have an NVIDIA GPU. You can work in a location that has a CPU only.
For example, your laptop can be a mac.
Local and Remote#
One UI/UX for all your locations#
Local Workbench: Desktop App installed locally is the UI to manage pretty much everything.
Laptop can be CPU only, or can also have a GPU.
The primary location that provides the user-interface.
You can develop and work here, even if the laptop is CPU-only.
Remote Workbench: CLI installed on a remote system
Remote can be a desktop, server or VM. Main point is that it’s accessed via SSH.
Accessed and managed from your local Workbench via SSH.
Provides more compute power, typically GPU, for workloads your laptop can’t handle.
Note
While you can have multiple remote locations, you should only have one local Workbench.
Transferring Projects#
Move projects between locations by syncing them through a Git platform like GitHub.com or GitLab.com.
Git in AI Workbench: Follows the typical collaboration and “right-sizing” workflow experienced developers, data scientists, and engineers use.
Workbench is explicitly designed to streamline this workflow, including handling of:
Runtime changes that need to be made, e.g. adjusting source directories for mounts.
GPU-specific configuration, e.g. CUDA_VISIBLE_DEVICES.
Underlying operating system differences, e.g. Windows vs Ubuntu vs MacOS.
Architecture differences, e.g. x86 vs ARM.
Managing Remote Locations#
Before you can work on a remote location, you must install AI Workbench on the system and then connect it to your local Workbench.
Setting Up a Remote Location#
There are two ways to setup a remote Workbench location:
Manually Connect a Remote: SSH into a remote system and run a command to install AI Workbench. Then connect to your local Workbench using SSH information.
NVIDIA Brev Integration (experimental): Create an account on Brev, configure Brev locally in Workbench, run a Workbench CLI command.
Activating a Remote Location#
Once a remote location is added to your local Workbench, you just click on the location name to activate it and start working.
All connected locations are visible in the My Locations view.
Activating a location may fail if the SSH connection has changed or the remote system is not on.
Deleting a Remote Location#
Delete a remote location by clicking the option dots on the location entry.
Do this from Manage Locations (“My Locations”).
This does not delete the remote system or the projects on it.
As long as the remote system is still running, you can re-add the remote location at any time.
FAQs#
Common questions on remote locations#
Can I use a remote Windows desktop as a remote location?#
Not directly. WSL made a recent change that supports this, but we’ve not yet implemented it by default.
You may be able to sort it out by yourself though.
Why do I need a remote location?#
Various reasons:
More compute power, i.e. GPUs.
The data you want to use is on a remote system and is too large to transfer to your local system.
You aren’t allowed to do work locally, i.e. your company’s IT policy.
How does Workbench handle the connection to a remote location?#
When you connect a remote location to your local Workbench, the following is enabled:
AI Workbench establishes a secure SSH tunnel between your local machine and the remote AI Workbench service.
Starting a remote Workbench session will automatically start the SSH tunnel.
Stopping a remote Workbench session will automatically stop the SSH tunnel.
Opening a project and starting an application in the project will automatically start another SSH tunnel.
A proxy service is used to properly route the connection to the application.
For more details, see Manually Connect a Remote.
What happens to the projects on a remote location when I delete the location?#
Nothing. Removing the location just removes the SSH connection. It doesn’t affect the actual remote system.
Can I connect to and manage remote locations with the CLI?#
Yes. You can connect to and manage remote locations with the CLI.
Consult the CLI help menu for more details: nvwb --help.